issue_owner_repo listlengths 2 2 | issue_body stringlengths 0 261k ⌀ | issue_title stringlengths 1 925 | issue_comments_url stringlengths 56 81 | issue_comments_count int64 0 2.5k | issue_created_at stringlengths 20 20 | issue_updated_at stringlengths 20 20 | issue_html_url stringlengths 37 62 | issue_github_id int64 387k 2.46B | issue_number int64 1 127k |
|---|---|---|---|---|---|---|---|---|---|
[
"langchain-ai",
"langchain"
] | ### Feature request
Is possible utilizate another languaje, like spanish?
### Motivation
Another Languaje
### Your contribution
I don't | Another Languaje | https://api.github.com/repos/langchain-ai/langchain/issues/11272/comments | 3 | 2023-10-02T00:27:49Z | 2024-02-08T16:23:26Z | https://github.com/langchain-ai/langchain/issues/11272 | 1,921,054,820 | 11,272 |
[
"langchain-ai",
"langchain"
] | ### System Info
I'm trying to use langchain with cache using gptcache, using pydantic models
(prior to this message I'm building prompt of mixed human and system messages)
```
chain = create_structured_output_chain(Job, llm, prompt, verbose=False)
res = chain.run({'job': job})
```
when not using cache everything seems to work but when getting from cache I'm seeing this error:
```
File "/Users/orlevy/Projects/tfg/llm-service/venv/lib/python3.9/site-packages/langchain/schema/output.py", line 62, in set_text
values["text"] = values["message"].content
KeyError: 'message'
```
When I digged a little further it seems like this code, is not generating the same response (we are losing function call response) :
<img width="661" alt="image" src="https://github.com/langchain-ai/langchain/assets/21291210/cfeccb45-4806-4f85-a01c-67b42a882c8d">
modified code a bit(not sure if this is actually a good idea) but this one seems to work for chat:
```
res = get(prompt, cache_obj=_gptcache)
if res:
return [
ChatGeneration(
message=AIMessage(content='',
additional_kwargs=generation_dict['message']['additional_kwargs']
),
generation_info=dict(finish_reason=generation_dict.get("finish_reason")),
) for generation_dict in json.loads(res)
]
```
would really appreciate
your help with this
### Who can help?
_No response_
### Information
- [ ] The official example notebooks/scripts
- [X] My own modified scripts
### Related Components
- [X] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
1. use chat with chain messages
2. use cachegpt (used postgres)
3. make one call
4. cache should be populated
5. from that point on when using cache we see error
### Expected behavior
I should be able to use cache | Chat cache results in KeyError: 'message' | https://api.github.com/repos/langchain-ai/langchain/issues/11268/comments | 7 | 2023-10-01T17:48:12Z | 2024-02-16T16:08:26Z | https://github.com/langchain-ai/langchain/issues/11268 | 1,920,886,661 | 11,268 |
[
"langchain-ai",
"langchain"
] | I'm using create_conversational_retrieval_agent to create an agent and also passing in a system prompt. However , the prompt is passed in as user message, is this expected behavior?
```
conv_agent_executor = create_conversational_retrieval_agent(
llm, tools, verbose=False,
system_message=SystemMessage(
content="Your name is Votum, an expert legal assistant with extensive knowledge about Indian law. Your task is to respond to the given query in a factually correct and concise manner unless asked for a detailed explanation. Feel free to use any tools available to look up relevant information, only if necessary")
)
```
Server Logs:
```
<|im_start|>system
You are a helpful assistant.<|im_end|>
<|im_start|>user
Your are an expert legal assistant with extensive knowledge about Indian law. Your task is to respond to the given query in a factually correct and concise manner unless asked for a detailed explanation. Feel free to use any tools available to look up relevant information, only if necessary
Answer the following questions as best you can. You have access to the following APIs:
search: Call this tool to interact with the search API. What is the search API useful for? search(query: str) -> str - Useful for retrieving documents related to Indian law. Parameters: {"title": "SearchInput", "type": "object", "properties": {"query": {"title": "Query", "description": "should be a search query in string format", "type": "string"}}, "required": ["query"]}
Use the following format:
Question: the input question you must answer
Thought: you should always think about what to do
Action: the action to take, should be one of [search]
Action Input: the input to the action
Observation: the result of the action
... (this Thought/Action/Action Input/Observation can be repeated zero or more times)
Thought: I now know the final answer
Final Answer: the final answer to the original input question
Begin!
```
### Suggestion:
_No response_ | SystemMessage not updating system prompt | https://api.github.com/repos/langchain-ai/langchain/issues/11262/comments | 2 | 2023-10-01T13:18:15Z | 2024-02-07T16:24:28Z | https://github.com/langchain-ai/langchain/issues/11262 | 1,920,763,561 | 11,262 |
[
"langchain-ai",
"langchain"
] | ### System Info
langchain version v0.0.305
Python 3.11
### Who can help?
@hwchase17 @agola11
### Information
- [ ] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [X] LLMs/Chat Models
- [ ] Embedding Models
- [X] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
```python
claudev2 = BedrockChat(
client=bedrock_inference,
model_id="anthropic.claude-v2",
)
from langchain.prompts import PromptTemplate
from langchain.chat_models import ChatOpenAI
from langchain_experimental.smart_llm import SmartLLMChain
prompt = PromptTemplate.from_template(hard_question)
chain = SmartLLMChain(
llm=claudev2,
prompt=prompt,
n_ideas=2,
)
response = chain.run({} )
print(response)
```
Results in error:
```
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
[/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb](https://file+.vscode-resource.vscode-cdn.net/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb) Cell 10 line 1
[5](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=4) prompt = PromptTemplate.from_template(hard_question)
[7](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=6) chain = SmartLLMChain(
[8](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=7) ideation_llm=claudev2,
[9](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=8) critique_llm=claudeinstant,
(...)
[15](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=14) #verbose=True,
[16](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=15) )
---> [18](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=17) response = chain.run({} )
[20](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=19) print(response)
File [~/mambaforge/envs/bedrock/lib/python3.11/site-packages/langchain/chains/base.py:507](https://file+.vscode-resource.vscode-cdn.net/Users/austinmw/Desktop/bedrocktesting/~/mambaforge/envs/bedrock/lib/python3.11/site-packages/langchain/chains/base.py:507), in Chain.run(self, callbacks, tags, metadata, *args, **kwargs)
505 if len(args) != 1:
506 raise ValueError("`run` supports only one positional argument.")
--> 507 return self(args[0], callbacks=callbacks, tags=tags, metadata=metadata)[
508 _output_key
509 ]
511 if kwargs and not args:
512 return self(kwargs, callbacks=callbacks, tags=tags, metadata=metadata)[
513 _output_key
514 ]
File [~/mambaforge/envs/bedrock/lib/python3.11/site-packages/langchain/chains/base.py:312](https://file+.vscode-resource.vscode-cdn.net/Users/austinmw/Desktop/bedrocktesting/~/mambaforge/envs/bedrock/lib/python3.11/site-packages/langchain/chains/base.py:312), in Chain.__call__(self, inputs, return_only_outputs, callbacks, tags, metadata, run_name, include_run_info)
310 except BaseException as e:
311 run_manager.on_chain_error(e)
--> 312 raise e
313 run_manager.on_chain_end(outputs)
314 final_outputs: Dict[str, Any] = self.prep_outputs(
315 inputs, outputs, return_only_outputs
316 )
File [~/mambaforge/envs/bedrock/lib/python3.11/site-packages/langchain/chains/base.py:306](https://file+.vscode-resource.vscode-cdn.net/Users/austinmw/Desktop/bedrocktesting/~/mambaforge/envs/bedrock/lib/python3.11/site-packages/langchain/chains/base.py:306), in Chain.__call__(self, inputs, return_only_outputs, callbacks, tags, metadata, run_name, include_run_info)
299 run_manager = callback_manager.on_chain_start(
300 dumpd(self),
301 inputs,
302 name=run_name,
303 )
304 try:
305 outputs = (
--> 306 self._call(inputs, run_manager=run_manager)
307 if new_arg_supported
308 else self._call(inputs)
309 )
310 except BaseException as e:
311 run_manager.on_chain_error(e)
File [~/mambaforge/envs/bedrock/lib/python3.11/site-packages/langchain_experimental/smart_llm/base.py:166](https://file+.vscode-resource.vscode-cdn.net/Users/austinmw/Desktop/bedrocktesting/~/mambaforge/envs/bedrock/lib/python3.11/site-packages/langchain_experimental/smart_llm/base.py:166), in SmartLLMChain._call(self, input_list, run_manager)
164 ideas = self._ideate(stop, run_manager)
165 self.history.ideas = ideas
--> 166 critique = self._critique(stop, run_manager)
167 self.history.critique = critique
168 resolution = self._resolve(stop, run_manager)
File [~/mambaforge/envs/bedrock/lib/python3.11/site-packages/langchain_experimental/smart_llm/base.py:293](https://file+.vscode-resource.vscode-cdn.net/Users/austinmw/Desktop/bedrocktesting/~/mambaforge/envs/bedrock/lib/python3.11/site-packages/langchain_experimental/smart_llm/base.py:293), in SmartLLMChain._critique(self, stop, run_manager)
290 callbacks = run_manager.handlers if run_manager else None
291 if llm:
292 critique = self._get_text_from_llm_result(
--> 293 llm.generate_prompt([prompt], stop, callbacks), step="critique"
294 )
295 _colored_text = get_colored_text(critique, "yellow")
296 _text = "Critique:\n" + _colored_text
File [~/mambaforge/envs/bedrock/lib/python3.11/site-packages/langchain/chat_models/base.py:475](https://file+.vscode-resource.vscode-cdn.net/Users/austinmw/Desktop/bedrocktesting/~/mambaforge/envs/bedrock/lib/python3.11/site-packages/langchain/chat_models/base.py:475), in BaseChatModel.generate_prompt(self, prompts, stop, callbacks, **kwargs)
467 def generate_prompt(
468 self,
469 prompts: List[PromptValue],
(...)
472 **kwargs: Any,
473 ) -> LLMResult:
474 prompt_messages = [p.to_messages() for p in prompts]
--> 475 return self.generate(prompt_messages, stop=stop, callbacks=callbacks, **kwargs)
File [~/mambaforge/envs/bedrock/lib/python3.11/site-packages/langchain/chat_models/base.py:365](https://file+.vscode-resource.vscode-cdn.net/Users/austinmw/Desktop/bedrocktesting/~/mambaforge/envs/bedrock/lib/python3.11/site-packages/langchain/chat_models/base.py:365), in BaseChatModel.generate(self, messages, stop, callbacks, tags, metadata, run_name, **kwargs)
363 if run_managers:
364 run_managers[i].on_llm_error(e)
--> 365 raise e
366 flattened_outputs = [
367 LLMResult(generations=[res.generations], llm_output=res.llm_output)
368 for res in results
369 ]
370 llm_output = self._combine_llm_outputs([res.llm_output for res in results])
File [~/mambaforge/envs/bedrock/lib/python3.11/site-packages/langchain/chat_models/base.py:355](https://file+.vscode-resource.vscode-cdn.net/Users/austinmw/Desktop/bedrocktesting/~/mambaforge/envs/bedrock/lib/python3.11/site-packages/langchain/chat_models/base.py:355), in BaseChatModel.generate(self, messages, stop, callbacks, tags, metadata, run_name, **kwargs)
352 for i, m in enumerate(messages):
353 try:
354 results.append(
--> 355 self._generate_with_cache(
356 m,
357 stop=stop,
358 run_manager=run_managers[i] if run_managers else None,
359 **kwargs,
360 )
361 )
362 except BaseException as e:
363 if run_managers:
File [~/mambaforge/envs/bedrock/lib/python3.11/site-packages/langchain/chat_models/base.py:507](https://file+.vscode-resource.vscode-cdn.net/Users/austinmw/Desktop/bedrocktesting/~/mambaforge/envs/bedrock/lib/python3.11/site-packages/langchain/chat_models/base.py:507), in BaseChatModel._generate_with_cache(self, messages, stop, run_manager, **kwargs)
503 raise ValueError(
504 "Asked to cache, but no cache found at `langchain.cache`."
505 )
506 if new_arg_supported:
--> 507 return self._generate(
508 messages, stop=stop, run_manager=run_manager, **kwargs
509 )
510 else:
511 return self._generate(messages, stop=stop, **kwargs)
[/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb](https://file+.vscode-resource.vscode-cdn.net/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb) Cell 10 line 3
[366](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=365) if stop:
[367](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=366) params["stop_sequences"] = stop
--> [369](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=368) completion = self._prepare_input_and_invoke(
[370](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=369) prompt=prompt, stop=stop, run_manager=run_manager, **params
[371](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=370) )
[373](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=372) message = AIMessage(content=completion)
[374](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=373) return ChatResult(generations=[ChatGeneration(message=message)])
[/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb](https://file+.vscode-resource.vscode-cdn.net/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb) Cell 10 line 2
[223](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=222) provider = self._get_provider()
[224](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=223) params = {**_model_kwargs, **kwargs}
--> [225](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=224) input_body = LLMInputOutputAdapterv2.prepare_input(provider, prompt, params)
[226](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=225) body = json.dumps(input_body)
[227](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=226) accept = "application/json"
[/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb](https://file+.vscode-resource.vscode-cdn.net/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb) Cell 10 line 7
[73](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=72) input_body = {**model_kwargs}
[74](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=73) if provider == "anthropic":
---> [75](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=74) input_body["prompt"] = _human_assistant_format_v2(prompt)
[76](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=75) elif provider == "ai21":
[77](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=76) input_body["prompt"] = prompt
[/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb](https://file+.vscode-resource.vscode-cdn.net/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb) Cell 10 line 4
[47](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=46) count += 1
[48](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=47) else:
---> [49](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=48) raise ValueError(ALTERNATION_ERROR)
[51](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=50) if count % 2 == 1: # Only saw Human, no Assistant
[52](vscode-notebook-cell:/Users/austinmw/Desktop/bedrocktesting/bedrock_test.ipynb#X14sZmlsZQ%3D%3D?line=51) input_text = input_text + ASSISTANT_PROMPT # SILENT CORRECTION
ValueError: Error: Prompt must alternate between '
Human:' and '
Assistant:'.
```
### Expected behavior
No error | SmartLLMChain errors with Bedrock Anthropic models and n_ideas>1 | https://api.github.com/repos/langchain-ai/langchain/issues/11261/comments | 8 | 2023-10-01T13:06:07Z | 2024-05-11T16:07:02Z | https://github.com/langchain-ai/langchain/issues/11261 | 1,920,758,064 | 11,261 |
[
"langchain-ai",
"langchain"
] | ### Feature request
Hi, does langchain have a feature similar to LlamaIndex's [Sub Question Query Engine](https://gpt-index.readthedocs.io/en/latest/examples/query_engine/sub_question_query_engine.html)? If not, I'd like to submit a feature request!
### Motivation
This capability is very useful!
### Your contribution
Not sure. | Sub Question Query Engine | https://api.github.com/repos/langchain-ai/langchain/issues/11260/comments | 10 | 2023-10-01T12:37:44Z | 2024-05-11T16:06:57Z | https://github.com/langchain-ai/langchain/issues/11260 | 1,920,745,638 | 11,260 |
[
"langchain-ai",
"langchain"
] | ### System Info
|Name|Detail|
|--------|--------|
| Python | `3.10.13` |
| LangChain | `0.0.301` |
| System | MacOS `12.2.1` |
### Bug Detail
Playing around with a recursive function to decrement value `k` and ran into an issue whereby `self` is overwritten and bound to instance of `openai.error.InvalidRequestError` when the exception is caught. Note in the code below, the `AttributeError` exception is thrown when `self.save()` is executed upon a successful `run()` or `try` where `openai.error.InvalidRequestError` is not thrown.
### Who can help?
_No response_
### Information
- [ ] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
**The code below is a gross oversimplification, but it illustrates how I found this bug:**
```python
class SomeModel(models.Model):
query = models.CharField(max_length=255)
response = models.TextField()
def save(self, *args, **kwargs):
# Unmodified Django save method
return super().save(*args, **kwargs)
def run(self, kb, k=5):
try:
llm = ChatOpenAI()
chain = load_qa_chain(llm=llm, chain_type='stuff')
self.response = chain.run(input_documents=[], human_input=self.query)
self.save() # AttributeError: 'InvalidRequestError' object has no attribute 'save'
except (openai.error.InvalidRequestError,) as e:
if k < 2:
raise e
new_k_val = k - 1
print('K value too high. Trying with %s' % new_k_val)
return self.run(kb, k=new_k_val)
```
### Expected behavior
Where exception happens, `self` should not be bound to another instance/instance of `openai.error.InvalidRequestError` | ChatOpenAI: Exception InvalidRequestError changes bound "self" | https://api.github.com/repos/langchain-ai/langchain/issues/11258/comments | 2 | 2023-10-01T10:23:41Z | 2023-10-01T11:52:12Z | https://github.com/langchain-ai/langchain/issues/11258 | 1,920,688,423 | 11,258 |
[
"langchain-ai",
"langchain"
] | ### Feature request
Is it possible to integrate any language model (LM) trained on source code or opensource codebases on various programming languages e.g. :
- [StarCoder](https://github.com/bigcode-project/starcoder)
- [replit/replit-code-v1-3b](https://huggingface.co/replit/replit-code-v1-3b)
- [Bloom](https://bigscience.huggingface.co/blog/bloom)
- etc...
as an [LLM Model](https://python.langchain.com/en/latest/modules/models.html) or an [Agent](https://python.langchain.com/en/latest/modules/agents.html) with [LangChain](https://github.com/hwchase17/langchain), and [chain](https://python.langchain.com/en/latest/modules/chains.html) it in a complex usecase?
It would be really cool if, during a general query, the LangChain system has the capability to understand code context (possibly when chained to a vector DB or any data source that contains code or a full code repo), _kind of like chatting with your codebase_!
Any help / hints on the same would be appreciated!
### Motivation
Initially inspired by this [Issue #173](https://github.com/nomic-ai/gpt4all/issues/173)
Raised a similar Issue long back with StarCoder [Issue #53](https://github.com/bigcode-project/starcoder/issues/53)
### Your contribution
n/a | Implement a custom LangChain class wrapper (LLM model/Agent) for StarCoder model | https://api.github.com/repos/langchain-ai/langchain/issues/11252/comments | 11 | 2023-09-30T17:57:16Z | 2024-03-13T19:58:25Z | https://github.com/langchain-ai/langchain/issues/11252 | 1,920,348,709 | 11,252 |
[
"langchain-ai",
"langchain"
] | ### System Info
I am following this example: https://python.langchain.com/docs/integrations/memory/dynamodb_chat_message_history
When attempting to import DynamoDBChatMessageHistory I am receiving an error
`from langchain.memory.chat_message_histories import DynamoDBChatMessageHistory`
`---------------------------------------------------------------------------
PydanticUserError Traceback (most recent call last)
/var/folders/jq/w80pdyz11zbbhbvz7p8795gh0000gn/T/ipykernel_9395/3590968237.py in <module>
----> 1 from langchain.memory.chat_message_histories import DynamoDBChatMessageHistory
2
3 # history = DynamoDBChatMessageHistory(table_name="SessionTable", session_id="0")
4
5 # history.add_user_message("hi!")
/usr/local/lib/python3.9/site-packages/langchain/__init__.py in <module>
4 from typing import Optional
5
----> 6 from langchain.agents import MRKLChain, ReActChain, SelfAskWithSearchChain
7 from langchain.cache import BaseCache
8 from langchain.chains import (
/usr/local/lib/python3.9/site-packages/langchain/agents/__init__.py in <module>
1 """Interface for agents."""
----> 2 from langchain.agents.agent import (
3 Agent,
4 AgentExecutor,
5 AgentOutputParser,
/usr/local/lib/python3.9/site-packages/langchain/agents/agent.py in <module>
13 from pydantic import BaseModel, root_validator
14
---> 15 from langchain.agents.tools import InvalidTool
16 from langchain.base_language import BaseLanguageModel
17 from langchain.callbacks.base import BaseCallbackManager
/usr/local/lib/python3.9/site-packages/langchain/agents/tools.py in <module>
2 from typing import Optional
3
----> 4 from langchain.callbacks.manager import (
5 AsyncCallbackManagerForToolRun,
6 CallbackManagerForToolRun,
/usr/local/lib/python3.9/site-packages/langchain/callbacks/__init__.py in <module>
1 """Callback handlers that allow listening to events in LangChain."""
2
----> 3 from langchain.callbacks.aim_callback import AimCallbackHandler
4 from langchain.callbacks.clearml_callback import ClearMLCallbackHandler
5 from langchain.callbacks.comet_ml_callback import CometCallbackHandler
/usr/local/lib/python3.9/site-packages/langchain/callbacks/aim_callback.py in <module>
2 from typing import Any, Dict, List, Optional, Union
3
----> 4 from langchain.callbacks.base import BaseCallbackHandler
5 from langchain.schema import AgentAction, AgentFinish, LLMResult
6
/usr/local/lib/python3.9/site-packages/langchain/callbacks/base.py in <module>
5 from uuid import UUID
6
----> 7 from langchain.schema import (
8 AgentAction,
9 AgentFinish,
/usr/local/lib/python3.9/site-packages/langchain/schema.py in <module>
145
146
--> 147 class ChatGeneration(Generation):
148 """Output of a single generation."""
149
/usr/local/lib/python3.9/site-packages/langchain/schema.py in ChatGeneration()
152
153 @root_validator
--> 154 def set_text(cls, values: Dict[str, Any]) -> Dict[str, Any]:
155 values["text"] = values["message"].content
156 return values
/usr/local/lib/python3.9/site-packages/pydantic/deprecated/class_validators.py in root_validator(pre, skip_on_failure, allow_reuse, *__args)
220 if __args:
221 # Ensure a nice error is raised if someone attempts to use the bare decorator
--> 222 return root_validator()(*__args) # type: ignore
223
224 if allow_reuse is True: # pragma: no cover
/usr/local/lib/python3.9/site-packages/pydantic/deprecated/class_validators.py in root_validator(pre, skip_on_failure, allow_reuse, *__args)
226 mode: Literal['before', 'after'] = 'before' if pre is True else 'after'
227 if pre is False and skip_on_failure is not True:
--> 228 raise PydanticUserError(
229 'If you use `@root_validator` with pre=False (the default) you MUST specify `skip_on_failure=True`.'
230 ' Note that `@root_validator` is deprecated and should be replaced with `@model_validator`.',
PydanticUserError: If you use `@root_validator` with pre=False (the default) you MUST specify `skip_on_failure=True`. Note that `@root_validator` is deprecated and should be replaced with `@model_validator`.
For further information visit https://errors.pydantic.dev/2.1.1/u/root-validator-pre-skip`
### Who can help?
_No response_
### Information
- [X] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [X] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
Attempt to import DynamoDBChatMessageHistory
from langchain.memory.chat_message_histories import DynamoDBChatMessageHistory
### Expected behavior
no errors, successful import | Error importing DynamoDBChatMessageHistory | https://api.github.com/repos/langchain-ai/langchain/issues/11250/comments | 2 | 2023-09-30T15:34:02Z | 2024-01-30T00:43:41Z | https://github.com/langchain-ai/langchain/issues/11250 | 1,920,300,925 | 11,250 |
[
"langchain-ai",
"langchain"
] | ### System Info
langchain==0.0.305
MacOS M1 Silicon
### Who can help?
_No response_
### Information
- [X] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [X] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
I have a basic methon for loading the pdf via mathpix
```
def load_pdf(path):
loader = MathpixPDFLoader(path)
pages = loader.load()
return pages;
pages = load_pdf("example_data/DOF-230519AFYXBECVECGY-0028747724.pdf")
```
but if I run this, I gen only this:
Status: None, waiting for processing to complete
Status: None, waiting for processing to complete
Status: None, waiting for processing to complete
Status: None, waiting for processing to complete
Status: None, waiting for processing to complete
Status: None, waiting for processing to complete
Status: None, waiting for processing to complete
...
in backend of mathpix I see the document aldready parsed.
Any idea on this?
### Expected behavior
Expect the document to return. | MathpixPDFLoader never finishes | https://api.github.com/repos/langchain-ai/langchain/issues/11249/comments | 6 | 2023-09-30T12:12:22Z | 2024-03-13T20:04:02Z | https://github.com/langchain-ai/langchain/issues/11249 | 1,920,235,293 | 11,249 |
[
"langchain-ai",
"langchain"
] | ### System Info
version 0.300
<img width="1397" alt="Screenshot 2023-09-30 at 5 36 41 AM" src="https://github.com/langchain-ai/langchain/assets/17947802/124b1f2a-db02-497f-9cd0-dbad1033b6c8">
\\nObservation: requests_put is not a valid tool, try one of [requests_get, requests_post].\\nThought:I need to use requests_post to update the pet\'s ship date.\\n\\nAction:
This keep happening time and time again!
### Who can help?
hierarchical planner, requests agent
https://github.com/openchatai/OpenCopilot/blob/ecd2f51bfc1f989133a44c4cff15d858b0d2fd30/llm-server/api_caller/planner.py#L319
### Information
- [ ] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [X] Agents / Agent Executors
- [X] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
--
### Expected behavior
The agent should be able to use the `put` tool when it needs to use it | Hierarchical Planner - Observation: requests_put is not a valid tool, try one of [requests_get, requests_post]. | https://api.github.com/repos/langchain-ai/langchain/issues/11244/comments | 5 | 2023-09-30T02:45:21Z | 2024-01-30T00:42:35Z | https://github.com/langchain-ai/langchain/issues/11244 | 1,920,074,483 | 11,244 |
[
"langchain-ai",
"langchain"
] | ### Feature request
We all want to have our clients to interact with the database using LLM, langchain with create_sql_agent, are great, the only problem I see, there is no way to limit user access to the database and fence him to his data.
For example we have table called AI_Product with many columns and one of them is storeid, for normal user, he should have limited access to this table, I also don't trust the AI to limit that, there should be a way to programmatically do that.
So let's look at AI_Product
```
id, name, price, store_id
1, water, 2, 10001
2, pepsi, 4, 10001
3, cokezero , 5, 10002
```
For the above table, the user only have rights to access storeid= 10001, for security and true data we need a way to limit what this user can see.
One solution could be is to allow include_tables to have sql
`include_tables = ['AI_Product']
`
to
`include_tables = ['(select * from AI_Product where store_id = '10001') AI_Product']
`
The other one, is allow the SQL query to come back before execution and allow me to modify it
So if I asked , "How many products do we have?"
```
agent: sql:"select count(*) from AI_Product"
Then we can modify the sql to
protected_sql = sql.replace("AI_Product", "(select * from AI_Product where store_id = '10001') AI_Product")
#sql = "select count(*) from (select * from AI_Product where store_id = '10001') AI_Product"
agent.run (protected_sql)
```
### Motivation
Find a way to protect the data while allow end-user unlimited access to it.
### Your contribution
I'm not familiar with the internal code, but I can help if gudied | Limit Data and Enforcing Security for SQLDatabase & Agents | https://api.github.com/repos/langchain-ai/langchain/issues/11243/comments | 17 | 2023-09-29T22:31:23Z | 2024-03-25T16:06:18Z | https://github.com/langchain-ai/langchain/issues/11243 | 1,919,979,559 | 11,243 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
while RAG we create a db and then we ask a query to the db in this way,
`index = VectorstoreIndexCreator(
vectorstore_cls=DocArrayInMemorySearch
).from_loaders([loader])`
`index.query(llm=chat, question=query)`
we get some response.
Now lets say wanted to create dataset with the generations ive got from gpt to fintune llama2 llm.
the main problem comes here, now how to know what is the reference we got from the index search as it is totaly abstracted!!
i need the exact querry that gone into llm(gpt) in this process.
Or else can i have any other meathod that solves my problem(tried similarity search and fed reference+query manualy the quality of the output droped so i need index search it self.
### Suggestion:
_No response_ | Issue: < Urgent Help needed related to VectorstoreIndexCreator > | https://api.github.com/repos/langchain-ai/langchain/issues/11242/comments | 2 | 2023-09-29T21:53:43Z | 2023-10-31T07:58:55Z | https://github.com/langchain-ai/langchain/issues/11242 | 1,919,948,621 | 11,242 |
[
"langchain-ai",
"langchain"
] | ### Feature request
[LanguageParser](https://github.com/langchain-ai/langchain/blob/e9b51513e9bab738c741a348ffbf0f36ce19faf6/libs/langchain/langchain/document_loaders/parsers/language/language_parser.py#L21C7-L21C21) is a parser for Document Loaders that, given source code, splits each top-level function or class into separate documents. As stated in its documentation:
> This approach can potentially improve the accuracy of QA models over source code.
>
> Currently, the supported languages for code parsing are Python and JavaScript.
We would like to add support for additional languages, such as C, C++, Rust, Ruby, Perl, and so on.
### Motivation
There is an [open request](https://github.com/langchain-ai/langchain/issues/10996) for "CPP" (presumably C++) support. By integrating a generic parsing library (such as [tree-sitter](https://github.com/tree-sitter/py-tree-sitter#py-tree-sitter)), we could make LanguageParser work with many more languages, and thus be more generally useful.
### Your contribution
We intend to submit a pull request for this issue no later than mid-November, and likely sooner. | Add support for a variety of languages to LanguageParser | https://api.github.com/repos/langchain-ai/langchain/issues/11229/comments | 33 | 2023-09-29T16:07:57Z | 2024-06-26T09:34:16Z | https://github.com/langchain-ai/langchain/issues/11229 | 1,919,554,598 | 11,229 |
[
"langchain-ai",
"langchain"
] | ### System Info
Python = 3.8.16
langchain==0.0.304
langchainhub==0.1.13
### Who can help?
@hwchase17 @agola11
### Information
- [ ] The official example notebooks/scripts
- [X] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [X] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
1. install python 3.8.16
2. install langchain==0.0.304 and langchainhub==0.1.13
3. Run the following script :
from langchain import hub
# LANGCHAIN_HUB_API_KEY is set on my .env file
# pull a chat prompt
try:
prompt = hub.pull("efriis/my-first-prompt")
except Exception as e:
print(e)
raise e
### Expected behavior
I'm uncertain about the expected behavior, as I've never been able to successfully execute hub.pull. | Encountering a "type object is not subscriptable" error when invoking hub.pull. | https://api.github.com/repos/langchain-ai/langchain/issues/11226/comments | 12 | 2023-09-29T15:55:14Z | 2024-02-27T11:46:37Z | https://github.com/langchain-ai/langchain/issues/11226 | 1,919,535,609 | 11,226 |
[
"langchain-ai",
"langchain"
] | ### Issue with current documentation:
What is the difference between OpenAIFunctionsAgent and OpenAIMultiFunctionsAgent? The document do not clarify it.
### Idea or request for content:
_No response_ | DOC: What is the difference between OpenAIFunctionsAgent and OpenAIMultiFunctionsAgent? | https://api.github.com/repos/langchain-ai/langchain/issues/11225/comments | 4 | 2023-09-29T15:51:08Z | 2024-02-12T16:12:59Z | https://github.com/langchain-ai/langchain/issues/11225 | 1,919,529,912 | 11,225 |
[
"langchain-ai",
"langchain"
] | When using async `LLMChain` calls, the logger prints out a ton of info statements that looks like this:
```
2023-09-29 08:24:51,038:INFO - message='OpenAI API response' path=https://api.openai.com/v1/chat/completions processing_ms=7 request_id=0cde8ecb8d402ec3db4ec674f0450ade response_code=200
```
I tried setting the logging level on the execution level in a bunch of ways, but to no avail. Do you have any recommendation on how to get rid of those logs? | How to disable logger streaming for async LLMChain calls? | https://api.github.com/repos/langchain-ai/langchain/issues/11223/comments | 1 | 2023-09-29T15:26:52Z | 2023-09-29T15:41:29Z | https://github.com/langchain-ai/langchain/issues/11223 | 1,919,488,448 | 11,223 |
[
"langchain-ai",
"langchain"
] | ### System Info
while trying to use BedrockChat , it threw my error below , the updated prompt which am using after the recent changes in claude prompt is below .
prompt_template = """
Human:
You are a helpful, respectful, and honest assistant, dedicated to providing valuable and accurate information.
Guidance for answers below
Answer the question only using the in the context given below, and not with the prior knowledge.
If you don't see answer in the context just Reply "Sorry , the answer is not in the context so I don't know".
Now read this context and answer the question.
{context}
Based on the provided context above and information from the retriever source, provide a detailed answer to the below question
{question}
If the information is not available in the context , respond with "don't know."
Assistant:
"""
Error is below
raise ValueError(ALTERNATION_ERROR)
ValueError: Error: Prompt must alternate between '
Human:' and '
Assistant:'.
### Who can help?
_No response_
### Information
- [ ] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
prompt_template = """
Human:
You are a helpful, respectful, and honest assistant, dedicated to providing valuable and accurate information.
Guidance for answers below
Answer the question only using the in the context given below, and not with the prior knowledge.
If you don't see answer in the context just Reply "Sorry , the answer is not in the context so I don't know".
Now read this context and answer the question.
{context}
Based on the provided context above and information from the retriever source, provide a detailed answer to the below question
{question}
If the information is not available in the context , respond with "don't know."
Assistant:
"""
modelId = 'anthropic.claude-v2'
# Langchain chain for invoking Endpoint
llm = BedrockChat(
model_id=modelId,
client=bedrock_client,
model_kwargs={"max_tokens_to_sample": 1000}
)
### Expected behavior
I would have expected the code to run with the prompt. | raise ValueError(ALTERNATION_ERROR) ValueError: Error: Prompt must alternate between ' Human:' and ' Assistant:'. | https://api.github.com/repos/langchain-ai/langchain/issues/11220/comments | 16 | 2023-09-29T13:18:59Z | 2024-05-11T16:06:52Z | https://github.com/langchain-ai/langchain/issues/11220 | 1,919,260,245 | 11,220 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
I fetch the result from pinecone using Pinecone's from_existing_index function.
My single Index contains multiple namespaces in pinecone.
Now I have 2 cases:
1. I want to get the result of particular namespace.
2. secondly I want to fetch the result from all the index's namespace.
How is it possible, Please Guide me accordingly?
### Suggestion:
_No response_ | Issue: How can I fetch the response from Pinecone using namespace with similarity_search | https://api.github.com/repos/langchain-ai/langchain/issues/11212/comments | 2 | 2023-09-29T08:39:49Z | 2024-01-30T00:41:52Z | https://github.com/langchain-ai/langchain/issues/11212 | 1,918,850,778 | 11,212 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
I have custom memory with memory key as 'context' as follows which i would like to use in the underling chain as readonly memory
```
memory = ScopeContextMemory(memory_key='context')
readonlymemory = ReadOnlySharedMemory(memory_key='context', memory=memory)
```
i have the following template stored in a file
```
Given an input question, first create a syntactically correct {dialect} query to run, then look at the results of the query and return the answer.
Use the following format:
Question: "Question here"
SQLQuery: "SQL Query to run"
SQLResult: "Result of the SQLQuery"
Answer: "Final answer here"
Only use the following tables:
{table_info}
Use the following context for the query generation:
{context}
```
i have prompt initialised as follows
```
with open('prompts/sql_default_prompt.txt','r') as fp:
DEFAULT_PROMPT = PromptTemplate(input_variables=['input','table_info','dialect', 'context'],
template=fp.read())
```
NOTE: 'context' is added as one of the input variable
Ive initialised SQL Database chain as follows
```
llm=OpenAI(temperature=0)
chain = SQLDatabaseChain.from_llm(llm=llm,
db=db,
verbose=True,
use_query_checker=True,
prompt=DEFAULT_PROMPT,
return_intermediate_steps=True,
memory=memory,
top_k=3)
```
when i tried to run im getting an exception on missing input keys 'context'
```
chain("List Items under the shopping cart for Davis")
---------------------------------------------------------------------------
ValueError
169 missing_keys = set(self.input_keys).difference(inputs)
170 if missing_keys:
--> 171 raise ValueError(f"Missing some input keys: {missing_keys}")
ValueError: Missing some input keys: {'context'}
```
### Suggestion:
Ive tried looking at the documention on memory and its usage in chain , tools and agents but could not figure out what is going wrong here. any pointers on the cause is appreciated and how can i remediate the thing | Issue: SQLDatabaseChain fails to run when initialised with custom read only memory | https://api.github.com/repos/langchain-ai/langchain/issues/11211/comments | 8 | 2023-09-29T07:59:22Z | 2023-09-29T14:32:37Z | https://github.com/langchain-ai/langchain/issues/11211 | 1,918,785,813 | 11,211 |
[
"langchain-ai",
"langchain"
] | ### Feature request
Hi,
I am working on Document QA, and using ConversationalRetrievalChain, with map-reduce as chain type.
In the final result of ConversationalRetrievalChain, I want to get the individual output from each map-chain.
### Motivation
This is a separate issue, and I think there should be a parameter that we can pass, to get the output of these chains as well.
### Your contribution
Not as of now. | Return output from map chain in the final result of ConversationalRetrievalChain. | https://api.github.com/repos/langchain-ai/langchain/issues/11209/comments | 4 | 2023-09-29T07:23:10Z | 2024-02-07T16:24:38Z | https://github.com/langchain-ai/langchain/issues/11209 | 1,918,733,653 | 11,209 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
How do you stream a langchain response from Flask to React?
I am having issues with it. I successfully implemented it with Azure OpenAI API only but langchain doesn't give me the right implementation.
### Suggestion:
_No response_ | How do you stream a langchain response from Flask to React? | https://api.github.com/repos/langchain-ai/langchain/issues/11208/comments | 3 | 2023-09-29T04:59:39Z | 2024-04-27T16:28:23Z | https://github.com/langchain-ai/langchain/issues/11208 | 1,918,600,924 | 11,208 |
[
"langchain-ai",
"langchain"
] | ### System Info
Following the steps of indexing from [ https://python.langchain.com/docs/modules/data_connection/indexing ](url) you'll find the following error "redis.exceptions.ResponseError: my_docs: no such index".
You'll get this exception while using redis as retriever:
### Who can help?
_No response_
### Information
- [x] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [X] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
The error is here:
If you look for the index you'll get (empty list or set).
This line makes impossible to save in the wanted format, and there's another thing: The index is not created for some reason.
I'll try to fix, but I'm not sure if it's possible for me at the moment, so I'm reporting this, I hope it helps.
### Expected behavior
Expected behavior inside Redis: "docs:indexname_:12ss2sadd" | Documents not being correctly indexed in vector database. ["redis.exceptions.ResponseError: my_docs: no such index"] | https://api.github.com/repos/langchain-ai/langchain/issues/11197/comments | 3 | 2023-09-28T19:57:36Z | 2023-10-24T17:51:26Z | https://github.com/langchain-ai/langchain/issues/11197 | 1,918,183,041 | 11,197 |
[
"langchain-ai",
"langchain"
] | ### System Info
Lambda Python 3.11 runtime
### Who can help?
_No response_
### Information
- [ ] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [X] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
As of Bedrock GA today, I get this error:
"/var/task/langchain/llms/bedrock.py", line 237, in _prepare_input_and_invoke raise ValueError(f"Error raised by bedrock service: {e}") ValueError: Error raised by bedrock service: An error occurred (ValidationException) when calling the InvokeModel operation: The requested operation is not recognized by the service.
### Expected behavior
It should invoke the model. I'm guessing with today's GA release of Bedrock, an underlying API call changed. | Bedrock InvokeModel failes on GA release | https://api.github.com/repos/langchain-ai/langchain/issues/11189/comments | 8 | 2023-09-28T18:17:28Z | 2024-02-15T16:09:15Z | https://github.com/langchain-ai/langchain/issues/11189 | 1,918,043,819 | 11,189 |
[
"langchain-ai",
"langchain"
] | ### System Info
Amazon Linux, langchain version 0.0.304, Docker Image
### Who can help?
@hwchase17
@agola11
### Information
- [ ] The official example notebooks/scripts
- [X] My own modified scripts
### Related Components
- [X] LLMs/Chat Models
- [ ] Embedding Models
- [X] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
Steps to reproduce:
1. Initialize ChatAnthropic model, with keyword "anthropic_api_key", as documented: https://api.python.langchain.com/en/latest/chat_models/langchain.chat_models.anthropic.ChatAnthropic.html#langchain.chat_models.anthropic.ChatAnthropic
2. Call chat model via __call__ method, e.g.:
"""
chat = ChatAnthropic(..., anthropic_api_key=os.environ["ANTHROPIC_API_KEY"])
messages = [....]
response = chat(messages)
"""
### Expected behavior
A normal response for a chat model is expected, but instead we get:
```
Traceback (most recent call last):
File "/var/task/app/src/utils/error_route.py", line 27, in custom_route_handler
return await original_route_handler(request)
File "/var/lang/lib/python3.10/site-packages/fastapi/routing.py", line 273, in app
raw_response = await run_endpoint_function(
File "/var/lang/lib/python3.10/site-packages/fastapi/routing.py", line 190, in run_endpoint_function
return await dependant.call(**values)
File "/var/task/app/app.py", line 184, in chat
reply = chat(data["question"])
File "/var/task/app/src/app_modules/chat.py", line 192, in __call__
res = self.chatbot(messages)
File "/var/lang/lib/python3.10/site-packages/langchain/chat_models/base.py", line 595, in __call__
generation = self.generate(
File "/var/lang/lib/python3.10/site-packages/langchain/chat_models/base.py", line 348, in generate
raise e
File "/var/lang/lib/python3.10/site-packages/langchain/chat_models/base.py", line 338, in generate
self._generate_with_cache(
File "/var/lang/lib/python3.10/site-packages/langchain/chat_models/base.py", line 490, in _generate_with_cache
return self._generate(
File "/var/lang/lib/python3.10/site-packages/langchain/chat_models/anthropic.py", line 184, in _generate
response = self.client.completions.create(**params)
File "/var/lang/lib/python3.10/site-packages/anthropic/_utils/_utils.py", line 250, in wrapper
return func(*args, **kwargs)
File "/var/lang/lib/python3.10/site-packages/anthropic/resources/completions.py", line 223, in create
return self._post(
File "/var/lang/lib/python3.10/site-packages/anthropic/_base_client.py", line 949, in post
return cast(ResponseT, self.request(cast_to, opts, stream=stream, stream_cls=stream_cls))
File "/var/lang/lib/python3.10/site-packages/anthropic/_base_client.py", line 748, in request
return self._request(
File "/var/lang/lib/python3.10/site-packages/anthropic/_base_client.py", line 766, in _request
request = self._build_request(options)
File "/var/lang/lib/python3.10/site-packages/anthropic/_base_client.py", line 397, in _build_request
headers = self._build_headers(options)
File "/var/lang/lib/python3.10/site-packages/anthropic/_base_client.py", line 373, in _build_headers
headers = httpx.Headers(headers_dict)
File "/var/lang/lib/python3.10/site-packages/httpx/_models.py", line 71, in __init__
self._list = [
File "/var/lang/lib/python3.10/site-packages/httpx/_models.py", line 75, in <listcomp>
normalize_header_value(v, encoding),
File "/var/lang/lib/python3.10/site-packages/httpx/_utils.py", line 53, in normalize_header_value
return value.encode(encoding or "ascii")
AttributeError: 'SecretStr' object has no attribute 'encode'
```
Corresponding httpx code:
```
def normalize_header_value(
value: typing.Union[str, bytes], encoding: typing.Optional[str] = None
) -> bytes:
"""
Coerce str/bytes into a strictly byte-wise HTTP header value.
"""
if isinstance(value, bytes):
return value
return value.encode(encoding or "ascii")
``` | Pydantic type conversion to SecreteStr for ChatAnthropic anthropic_api_key keyword value raises attribute error in Anthropic client | https://api.github.com/repos/langchain-ai/langchain/issues/11182/comments | 4 | 2023-09-28T16:37:45Z | 2023-09-28T18:21:58Z | https://github.com/langchain-ai/langchain/issues/11182 | 1,917,900,561 | 11,182 |
[
"langchain-ai",
"langchain"
] | ### Feature request
Bedrock added support for Cohere models and the Bedrock llm code must be updated with the right prompt formatting.
The prompt must be passed as `"prompt"` field and not as the code below does as `"input"`
```
{
"modelId": "cohere.command-text-v14",
"contentType": "application/json",
"accept": "*/*",
"body": {
"prompt": "Write a LinkedIn post about starting a career in tech:",
"max_tokens": 100,
"temperature": 0.8,
"return_likelihood": "GENERATION"
}
}
```
https://github.com/langchain-ai/langchain/blob/44489e70291be95b5a95c2e0476f205202945778/libs/langchain/langchain/llms/bedrock.py#L71C1-L89C26
### Motivation
Getting support for all FM exposed by Bedrock via Langchain
### Your contribution
I can submit a PR | Add support for Bedrock cohere models | https://api.github.com/repos/langchain-ai/langchain/issues/11181/comments | 2 | 2023-09-28T16:10:03Z | 2024-01-30T00:43:13Z | https://github.com/langchain-ai/langchain/issues/11181 | 1,917,854,905 | 11,181 |
[
"langchain-ai",
"langchain"
] | ### System Info
LangChain version: 0.0.304
### Who can help?
@eyurt
### Information
- [X] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [X] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
```python
from langchain.document_loaders.web_base import WebBaseLoader
loader = WebBaseLoader("https://www.google.com")
docs = loader.load()
```
Result: RequestException is raised with the message "Invalid URL 'h': No scheme supplied. Perhaps you meant https://h?"
### Expected behavior
The page contents correctly loads to documents. | WebBaseLoader interprets incorrectly the web_path parameter | https://api.github.com/repos/langchain-ai/langchain/issues/11180/comments | 2 | 2023-09-28T15:45:13Z | 2023-09-28T15:48:19Z | https://github.com/langchain-ai/langchain/issues/11180 | 1,917,815,779 | 11,180 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
full_chain = {
"source1": {"question": lambda x: x["question"]} | query_chain | db.run,
"source2": (lambda x: x['question']) | vectorstore.as_retriever(),
"question": lambda x: x['question'],
} | qa_prompt | ChatOpenAI(streaming=True, callbacks=[QueueCallback(q)], temperature=0, model_name='gpt-4')
how can I add chat history to this?
### Suggestion:
_No response_ | Issue: <Please write a comprehensive title after the 'Issue: ' prefix> | https://api.github.com/repos/langchain-ai/langchain/issues/11173/comments | 3 | 2023-09-28T13:22:51Z | 2024-01-30T00:36:52Z | https://github.com/langchain-ai/langchain/issues/11173 | 1,917,547,206 | 11,173 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
full_chain = {
"source1": {"question": lambda x: x["question"]} | query_chain | db.run,
"source2": (lambda x: x['question']) | vectorstore.as_retriever(),
"question": lambda x: x['question'],
} | qa_prompt | ChatOpenAI(streaming=True, callbacks=[QueueCallback(q)], temperature=0, model_name='gpt-4')
how can I add chat history to this please?
### Suggestion:
_No response_ | how to add chat history to runnablesequence instance | https://api.github.com/repos/langchain-ai/langchain/issues/11171/comments | 6 | 2023-09-28T12:54:18Z | 2024-01-30T00:40:13Z | https://github.com/langchain-ai/langchain/issues/11171 | 1,917,489,116 | 11,171 |
[
"langchain-ai",
"langchain"
] | ### Feature request
When I use llama cpp for my case, I have [problems](https://github.com/ggerganov/llama.cpp/issues/2619). I've tried use "-esp 1e-5". It works well for me.
But when I use ```from langchain.llms import LlamaCpp``` I get the same problems. However I dont know the way to pass parameters
Thank you~~
### Motivation
Relevant with encode - decode
### Your contribution
[Llama cpp](https://github.com/ggerganov/llama.cpp) | norm_eps parameters in LlamaCpp | https://api.github.com/repos/langchain-ai/langchain/issues/11169/comments | 2 | 2023-09-28T11:48:48Z | 2024-01-30T00:41:57Z | https://github.com/langchain-ai/langchain/issues/11169 | 1,917,344,258 | 11,169 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
I am writing a few apps that use LangChain and deploying them to AWS Lambda. My previous 2 lambdas are running smoothly, but the one I'm trying to deploy this week is giving me the following stack trace when the lambda is run:
``` "errorMessage": "[Errno 38] Function not implemented",
"errorType": "OSError",
"requestId": "23ef7a93-3d00-475c-b495-421810f93d92",
"stackTrace": [
" File \"/var/task/datadog_lambda/wrapper.py\", line 214, in __call__\n self.response = self.func(event, context, **kwargs)\n",
" File \"/var/task/app.py\", line 4, in lambda_handler\n total_indexed_pages = get_latest_changes()\n",
" File \"/var/task/ddtrace/contrib/aws_lambda/patch.py\", line 115, in __call__\n self.response = self.func(*args, **kwargs)\n",
" File \"/var/task/app.py\", line 4, in lambda_handler\n total_indexed_pages = get_latest_changes()\n",
" File \"/var/task/index_latest_docs.py\", line 137, in get_latest_changes\n raise e\n",
" File \"/var/task/index_latest_docs.py\", line 126, in get_latest_changes\n Pinecone.from_documents(docs, embedding=embeddings, index_name=PINECONE_INDEX_NAME, namespace=PINECONE_NAMESPACE)\n",
" File \"/var/task/langchain/schema/vectorstore.py\", line 420, in from_documents\n return cls.from_texts(texts, embedding, metadatas=metadatas, **kwargs)\n",
" File \"/var/task/langchain/vectorstores/pinecone.py\", line 416, in from_texts\n pinecone.add_texts(\n",
" File \"/var/task/langchain/vectorstores/pinecone.py\", line 138, in add_texts\n async_res = [\n",
" File \"/var/task/langchain/vectorstores/pinecone.py\", line 139, in <listcomp>\n self._index.upsert(\n",
" File \"/var/task/pinecone/core/utils/error_handling.py\", line 17, in inner_func\n return func(*args, **kwargs)\n",
" File \"/var/task/pinecone/index.py\", line 150, in upsert\n return self._upsert_batch(vectors, namespace, _check_type, **kwargs)\n",
" File \"/var/task/pinecone/index.py\", line 237, in _upsert_batch\n return self._vector_api.upsert(\n",
" File \"/var/task/pinecone/core/client/api_client.py\", line 776, in __call__\n return self.callable(self, *args, **kwargs)\n",
" File \"/var/task/pinecone/core/client/api/vector_operations_api.py\", line 956, in __upsert\n return self.call_with_http_info(**kwargs)\n",
" File \"/var/task/pinecone/core/client/api_client.py\", line 838, in call_with_http_info\n return self.api_client.call_api(\n",
" File \"/var/task/pinecone/core/client/api_client.py\", line 421, in call_api\n return self.pool.apply_async(self.__call_api, (resource_path,\n",
" File \"/var/task/pinecone/core/client/api_client.py\", line 107, in pool\n self._pool = ThreadPool(self.pool_threads)\n",
" File \"/var/lang/lib/python3.11/multiprocessing/pool.py\", line 930, in __init__\n Pool.__init__(self, processes, initializer, initargs)\n",
" File \"/var/lang/lib/python3.11/multiprocessing/pool.py\", line 196, in __init__\n self._change_notifier = self._ctx.SimpleQueue()\n",
" File \"/var/lang/lib/python3.11/multiprocessing/context.py\", line 113, in SimpleQueue\n return SimpleQueue(ctx=self.get_context())\n",
" File \"/var/lang/lib/python3.11/multiprocessing/queues.py\", line 341, in __init__\n self._rlock = ctx.Lock()\n",
" File \"/var/lang/lib/python3.11/multiprocessing/context.py\", line 68, in Lock\n return Lock(ctx=self.get_context())\n",
" File \"/var/lang/lib/python3.11/multiprocessing/synchronize.py\", line 167, in __init__\n SemLock.__init__(self, SEMAPHORE, 1, 1, ctx=ctx)\n",
" File \"/var/lang/lib/python3.11/multiprocessing/synchronize.py\", line 57, in __init__\n sl = self._semlock = _multiprocessing.SemLock
```
On line 126 in my index_latest_docs.py file, I have the following line:
`Pinecone.from_documents(docs, embedding=embeddings, index_name=PINECONE_INDEX_NAME, namespace=PINECONE_NAMESPACE)`
This is causing it to not run in AWS Lambda because Lambda cannot handle multiprocessing, which the function seems to be doing internally.
I don't know if this is a newly introduced issue, because I'm using the same exact function in my other lambda apps, and they seem to be running fine - but maybe the containers were built with a different version than this more recent one.
Any help would be greatly appreciated
### Suggestion:
_No response_ | Issue: Pinecone.from_documents cannot run on AWS Lambda because of multiprocessing | https://api.github.com/repos/langchain-ai/langchain/issues/11168/comments | 13 | 2023-09-28T11:38:04Z | 2024-05-11T23:12:40Z | https://github.com/langchain-ai/langchain/issues/11168 | 1,917,319,585 | 11,168 |
[
"langchain-ai",
"langchain"
] | ### System Info
langchain==0.0.299
Python 3.10.13
OS: Windows 10
### Who can help?
_No response_
### Information
- [ ] The official example notebooks/scripts
- [X] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [x] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [X] Agents / Agent Executors
- [X] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
```python
from langchain.agents import initialize_agent
from langchain.agents.agent_types import AgentType
from langchain.chat_models import AzureChatOpenAI
from langchain.agents.tools import Tool
from langchain.agents import load_tools
from langchain.agents import create_csv_agent
from langchain.memory import ConversationBufferMemory
llm = AzureChatOpenAI(
model="gpt-4",
temperature=0,
openai_api_version="2023-05-15",
verbose=True,
)
tools = load_tools(["wikipedia", "llm-math"], llm=llm)
csv_agent = create_csv_agent(
llm=llm,
path=["sample_02.csv"], # some sales data
verbose=True,
handle_parsing_errors=True,
agent_type=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
memory=ConversationBufferMemory(),
)
csv_tool = Tool(
name="CSV Agent",
func=csv_agent.run,
description="Useful for interacting with CSV data.",
)
tools.append(csv_tool)
agent = initialize_agent(
llm=llm,
tools=tools,
agent_type=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
verbose=True,
handle_parsing_errors=True,
memory=ConversationBufferMemory(),
)
agent.run("What is the sum of sales")
```
### Expected behavior
I am getting exception:
```
File "C:\Users\X\lib\site-packages\langchain\agents\mrkl\output_parser.py", line 52, in parse raise OutputParserException( langchain.schema.output_parser.OutputParserException: Could not parse LLM output: 'some info about data comes here'
```
Expected: No exception and agent should have same behaviour as if I would use the CSV agent standalone | [BUG] OutputParserException when using CSV agent as a tool for another agent | https://api.github.com/repos/langchain-ai/langchain/issues/11167/comments | 5 | 2023-09-28T11:15:57Z | 2023-10-05T13:33:50Z | https://github.com/langchain-ai/langchain/issues/11167 | 1,917,278,742 | 11,167 |
[
"langchain-ai",
"langchain"
] | ### System Info
Langchain 0.0.302
Pydantic 2.4.1
reproducing code:
```
from langchain.chains import create_extraction_chain_pydantic
from langchain.chat_models import ChatOpenAI
from pydantic import BaseModel, model_validator
from typing import Any
class TestData(BaseModel):
company_name: str
llm = ChatOpenAI(temperature=0, model="gpt-3.5-turbo")
chain = create_extraction_chain_pydantic(pydantic_schema=TestData, llm=llm)
extracted_subsidiary_list = chain.run('')
```
Exception:
**RuntimeError: no validator found for <class '__main__.TestData'>, see `arbitrary_types_allowed` in Config**
Tried adding: (didn't help)
```
@model_validator(mode='after')
def dummy_validator(self):
return self
@model_validator(mode='before')
@classmethod
def dummy_validator1(cls, data: Any) -> Any:
return data
```
### Who can help?
@agola11
### Information
- [ ] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [X] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
```
from langchain.chains import create_extraction_chain_pydantic
from langchain.chat_models import ChatOpenAI
from pydantic import BaseModel, model_validator
from typing import Any
class TestData(BaseModel):
company_name: str
llm = ChatOpenAI(temperature=0, model="gpt-3.5-turbo")
chain = create_extraction_chain_pydantic(pydantic_schema=TestData, llm=llm)
extracted_subsidiary_list = chain.run('')
```
### Expected behavior
Doesn't throw exception | Runtime Exception in create_extraction_chain_pydantic (RuntimeError: no validator found) | https://api.github.com/repos/langchain-ai/langchain/issues/11166/comments | 6 | 2023-09-28T11:09:25Z | 2024-01-30T00:38:09Z | https://github.com/langchain-ai/langchain/issues/11166 | 1,917,264,832 | 11,166 |
[
"langchain-ai",
"langchain"
] | ### System Info
Hi Team,
The issue is while calling the ChatAnthropic function with the right anthropic secret key. The type SecretStr is not getting resolved. Discussed over here:https://github.com/langchain-ai/langchain/pull/10724
### Who can help?
@josh
### Information
- [ ] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [X] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [X] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
To recreate the error :
from langchain.chat_models import ChatAnthropic
claude_chat = ChatAnthropic(anthropic_api_key="xy-key")
### Expected behavior
It should initialize the claude chat | Argument 'value' has incorrect type (expected str, got SecretStr) (type=type_error) in chatAnthropic- Induced in 0.0.301 | https://api.github.com/repos/langchain-ai/langchain/issues/11165/comments | 4 | 2023-09-28T11:04:07Z | 2023-10-03T12:20:10Z | https://github.com/langchain-ai/langchain/issues/11165 | 1,917,252,474 | 11,165 |
[
"langchain-ai",
"langchain"
] | ### Feature request
Add `hierarchy` key to Document metadata returned by `ConfluenceLoader`.
So, for example if a page is located under the pages Page A/Page B, the `hierarchy` should be equal to `Page A/Page B`
This could be achieved by calling the confluence endpoint using `expand = "ancestors"`.
For example [this](https://github.com/langchain-ai/langchain/blob/cfa2203c626a2287d60c1febeb3e3a68b77acd77/libs/langchain/langchain/document_loaders/confluence.py#L282) would become
```python
pages = self.paginate_request(
self.confluence.get_all_pages_from_space,
space=space_key,
limit=limit,
max_pages=max_pages,
status="any" if include_archived_content else "current",
expand=f"{content_format.value},ancestors",
)
```
### Motivation
Page Hierarchy is very important in a confluence-like space and it would definitely be relevant when searching for documents in the Vector Store.
### Your contribution
I am willing to submit a PR to add this feature. | Add `hierarchy` key to Document metadata in `ConfluenceLoader` | https://api.github.com/repos/langchain-ai/langchain/issues/11163/comments | 2 | 2023-09-28T10:33:12Z | 2024-01-30T00:37:05Z | https://github.com/langchain-ai/langchain/issues/11163 | 1,917,192,653 | 11,163 |
[
"langchain-ai",
"langchain"
] | ### Feature request
It looks like most of the models of the `Completion` API have been marked as legacy and will be discontinued in early 2024.
On Azure OpenAI I already no longer have access to non-chat models and I can't use the completion API with the newest get-35-turbo and gpt4 (0613).
Shall we raise a warning to the user when instantiating `langchain.llms.OpenAI` and `langchain.llms.AzureOpenAI` with deprecated models?
References:
- https://openai.com/blog/gpt-4-api-general-availability
- https://platform.openai.com/docs/deprecations
### Your contribution
I can contribute to this if the feature makes sense. | Raise a warning when using OpenAI and Azure OpenAI Completion API | https://api.github.com/repos/langchain-ai/langchain/issues/11159/comments | 1 | 2023-09-28T09:27:24Z | 2024-01-30T00:40:21Z | https://github.com/langchain-ai/langchain/issues/11159 | 1,917,083,704 | 11,159 |
[
"langchain-ai",
"langchain"
] | template = """Use the following pieces of context to answer the question at the end. If the answer can't be determined using only the information in the provided context simply output "NO ANSWER", just say that you don't know, don't try to make up an answer. Use three sentences maximum. Keep the answer as concise as possible.
{context}
Question: {question}
Helpful Answer:"""
QA_CHAIN_PROMPT = PromptTemplate.from_template(template)
qa = RetrievalQA.from_chain_type(
llm=streaming_llm,
chain_type="stuff",
retriever=compression_retriever,
callback_manager=manager,
chain_type_kwargs={
"prompt": QA_CHAIN_PROMPT
}
) | ISSUE: how can i get answer only based on vector store? If the context is not available in vector store | https://api.github.com/repos/langchain-ai/langchain/issues/11157/comments | 9 | 2023-09-28T08:16:34Z | 2024-02-14T16:10:19Z | https://github.com/langchain-ai/langchain/issues/11157 | 1,916,961,643 | 11,157 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
I am trying to run a PlanAndExecute agent. The load_agent_executor method call returns:
"pydantic.v1.errors.ConfigError: field "retry_chain" not yet prepared so type is still a ForwardRef, you might need to call OutputFixingParser.update_forward_refs()."
This is my code:
`import os
import dotenv
from langchain.chat_models import ChatOpenAI
from langchain_experimental.plan_and_execute import PlanAndExecute, load_agent_executor, load_chat_planner
from langchain.llms import OpenAI
from langchain.agents.tools import Tool
from langchain.chains.llm_math.base import LLMMathChain
from langchain.tools import DuckDuckGoSearchRun
from langchain.output_parsers import OutputFixingParser
dotenv.load_dotenv()
apikey = os.getenv("OPEN_AI_API_KEY", default="")
search = DuckDuckGoSearchRun()
llm = OpenAI(temperature=0, openai_api_key=apikey)
llm_math_chain = LLMMathChain.from_llm(llm=llm, verbose=True)
tools = [
Tool(
name="Search",
func=search.run,
description="useful for when you need to answer questions about current events"
),
Tool(
name="Calculator",
func=llm_math_chain.run,
description="useful for when you need to answer questions about math"
),
]
model = ChatOpenAI(temperature=0)
planner = load_chat_planner(model)
executor = load_agent_executor(model, tools, verbose=True)
agent = PlanAndExecute(planner=planner, executor=executor, verbose=True)`
### Suggestion:
_No response_ | Issue: load_agent_executor returning ConfigError | https://api.github.com/repos/langchain-ai/langchain/issues/11156/comments | 5 | 2023-09-28T08:11:27Z | 2024-01-30T00:40:24Z | https://github.com/langchain-ai/langchain/issues/11156 | 1,916,953,343 | 11,156 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
Python 3.10.7
langchain 0.0.301
I started exploring langchain features such as stuff mode, map reduce mode and refine mode.
i would like to know more about the map reduce and refine modes further.
my understanding is that if the token limit of a document exceeds then the document goes through the spit and passes in the next call.
Using the following code
```
import os
from langchain.document_loaders import WebBaseLoader
from langchain.chat_models import ChatOpenAI
from langchain.chains.combine_documents.stuff import StuffDocumentsChain
from langchain.prompts import PromptTemplate
from langchain.chains import LLMChain, ReduceDocumentsChain, MapReduceDocumentsChain
from langchain.text_splitter import CharacterTextSplitter
os.environ['OPENAI_API_KEY'] = "xxxxx-xxxxxxxxxxxxxxxxxxxxxxxxx"
loader = WebBaseLoader("https://lilianweng.github.io/posts/2023-06-23-agent/")
docs = loader.load()
#Map starts
map_template = """The following is a set of documents
{docs}
Based on this list of docs, please identify the main themes
Helpful Answer:"""
map_prompt = PromptTemplate.from_template(map_template)
llm = ChatOpenAI(temperature=0)
map_chain = LLMChain(llm=llm, prompt=map_prompt)
#Map ends
#Reduce starts
reduce_template = """The following is set of summaries:
{doc_summaries}
Take these and distill it into a final, consolidated summary of the main themes.
Helpful Answer:"""
reduce_prompt = PromptTemplate.from_template(reduce_template)
reduce_chain = LLMChain(llm=llm, prompt=reduce_prompt)
# Takes a list of documents, combines them into a single string, and passes this to an LLMChain
combine_documents_chain = StuffDocumentsChain(
llm_chain=reduce_chain, document_variable_name="doc_summaries"
)
reduce_documents_chain = ReduceDocumentsChain(
combine_documents_chain=combine_documents_chain,
collapse_documents_chain=combine_documents_chain,
token_max=4000,
)
#Reduce ends
#combine Map and Reduce starts
map_reduce_chain = MapReduceDocumentsChain(
llm_chain=map_chain,
reduce_documents_chain=reduce_documents_chain,
document_variable_name="docs",
return_intermediate_steps=False,
)
#combine Map and Reduce ends
text_splitter = CharacterTextSplitter.from_tiktoken_encoder(
chunk_size=1000, chunk_overlap=0
)
split_docs = text_splitter.split_documents(docs)
print(map_reduce_chain.run(split_docs[:8]))
#print(split_docs[:8])
```
as we can see i am cnsidering the first 8 documents of the public dataset, which is having 8,104 Tokens composed of 34312 Characters
As per the 4000 tokens limit the dataset shall break in to three parts and process the operation, but i am facing with the issue
> Retrying langchain.chat_models.openai.ChatOpenAI.completion_with_retry.<locals>._completion_with_retry in 4.0 seconds as it raised RateLimitError: Rate limit reached for default-gpt-3.5-turbo in organization org-xxxxxxxxxxxxxxxxx on requests per min. Limit: 3 / min. Please try again in 20s. Contact us through our help center at help.openai.com if you continue to have issues. Please add a payment method to your account to increase your rate limit. Visit https://platform.openai.com/account/billing to add a payment method..
could you please help me to under stand further how map reduce or refine actual workings.
### Suggestion:
_No response_ | Issue: While using Map Reduce Max number of limit exceed even though not exceed. | https://api.github.com/repos/langchain-ai/langchain/issues/11155/comments | 14 | 2023-09-28T06:11:23Z | 2024-02-15T16:09:20Z | https://github.com/langchain-ai/langchain/issues/11155 | 1,916,784,254 | 11,155 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
I am using langchain version 0.0.281 and using Milvus as the vector store. For now, I have a Milvus-lite instance on my device. I tried using paritions with base vectorstore operations like `add_documents` and `similarity_search` and its working fine i.e it inserts and searches from the mentioned partition only. But when I try to use partitions with ConversationalRetrievalChain and even RetrievalQA it doesn't work i.e it retrieves documents through all partitions. When using `from_llm` classmethod I tried giving the partition names like below:
`QA = ConversationalRetrievalChain.from_llm(
llm=LLM,
retriever=VECTOR_DB.vec_db.as_retriever(partition_names=partition_names),
memory=MEMORY,
return_source_documents=True
)`
But this doesn't ensure that the vector store will only search the mentioned partitions, rather the vectorestore searches through all the documents present regardless of partition. Is there anyway I can limit the retrieval of documents to only those present in mentioned partition.
### Suggestion:
_No response_ | Issue: Using Milvus partitions with ConversationalRetrievalChain | https://api.github.com/repos/langchain-ai/langchain/issues/11153/comments | 2 | 2023-09-28T04:59:34Z | 2023-09-28T09:13:26Z | https://github.com/langchain-ai/langchain/issues/11153 | 1,916,701,031 | 11,153 |
[
"langchain-ai",
"langchain"
] | ### System Info
when using PlanAndExecutor with guidance prompt of using tools, it does remember previously steps, which will make the pipeline fail as it doesn't know what's previously steps' ouput...
<img width="822" alt="image" src="https://github.com/langchain-ai/langchain/assets/35717759/726a42c3-e4c9-4acf-9808-66645eee090c">
### Who can help?
@hwchase17 @agola11
### Information
- [ ] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [X] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [X] Agents / Agent Executors
- [ ] Tools / Toolkits
- [X] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
1: define guidence tools prompt:
combo_template = """
First understand the problem and devise a plan to solve the problem.
Please output the plan starting with the header 'Plan:' and then followed by a numbered list of steps. Do not use past conversation history when you are planning the steps.
Please make the plan the minimum number of steps required to accurately complete the task.
These are guidance on when to use a tool to solve a task, follow them strictly:
first use "reverse index search" tool to get table name
if previously step not find anyting，then use "embedding knn search" tool to get table name
if previously step still can't find the answer, say "I can't find the answer for this question."
otherwise use "db Querying Tool" tool to search table info using table name which previously step return
DO NOT CREATE STEPS THAT ARE NOT NEEDED TO SOLVE A TASK.
Once you have answers for the question, stop and provide the final answers. The final answers should be a combination of the answers to all the questions, not just the last one.
Please make sure you have a plan to answer all the questions in the input, not just the last one.
Please use these to construct an answer to the question , as though you were answering the question directly. Ensure that your answer is accurate and doesn’t contain any information not directly supported by the summary and quotes.
If there are no data or information in this document that seem relevant to this question, please just say "I can’t find any relevant quotes".
"""
2: load chat planner with llm(llama2 13b):
planner = load_chat_planner(sm_llm)
system_message_prompt = SystemMessagePromptTemplate.from_template(combo_template)
human_message_prompt = planner.llm_chain.prompt.messages[1]
planner.llm_chain.prompt = ChatPromptTemplate.from_messages([system_message_prompt, human_message_prompt])
executor = load_agent_executor(sm_llm, custom_tool_list, verbose=False)
agent = PlanAndExecute(planner=planner, executor=executor, verbose=False, max_iterations=1)
3: run agent
output = agent({"input":"最近一个月温度合格的派车单数量"})
### Expected behavior
base on each steps to get into next step with previously steps' output | P | https://api.github.com/repos/langchain-ai/langchain/issues/11152/comments | 4 | 2023-09-28T04:26:55Z | 2024-02-10T16:16:27Z | https://github.com/langchain-ai/langchain/issues/11152 | 1,916,677,818 | 11,152 |
[
"langchain-ai",
"langchain"
] | ### Issue with current documentation:
[This link ](https://python.langchain.com/docs/modules/model_io/prompts/prompt_templates/custom_prompt_template.html)references a code sample for a Custom Prompt (awesome stuff!)
Pydantic updated their library and the code sample or library dependencies do not work anymore. They have a process to help migrate code
Details here: https://docs.pydantic.dev/latest/migration/
Please update the code samples so that they work with the new version of the Pydantic library so that it works for any Langchain users.
### Idea or request for content:
Updated code samples only | DOC: Custom Prompt example code does not work. Requires pydantic V2 | https://api.github.com/repos/langchain-ai/langchain/issues/11150/comments | 1 | 2023-09-28T03:40:36Z | 2024-01-30T00:35:57Z | https://github.com/langchain-ai/langchain/issues/11150 | 1,916,646,925 | 11,150 |
[
"langchain-ai",
"langchain"
] | hi team,
Can I feed different QA prompt into conversationRetrievalQA?
I know to set the specific prompt according to knowledge vector store searched.
For example, If document A be searched I would set the prompt-A for it, document B for prompt-B. | converstaionRetrievalQA with different prompt | https://api.github.com/repos/langchain-ai/langchain/issues/11149/comments | 3 | 2023-09-28T03:37:06Z | 2024-02-10T16:16:32Z | https://github.com/langchain-ai/langchain/issues/11149 | 1,916,644,707 | 11,149 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
I followed the tutorial at [Code Understanding](https://python.langchain.com/docs/use_cases/code_understanding), loaded a small directory of test files into the db, and asked the question:
`Ask a question: what ways would you simplify e2e_btc_mint_2auto.test.js?`
The retriever, which was setup as follows, actually gave every file in the directory except `e2e_btc_mint_2auto.test.js` to ChatGPT! Why is ChromaDB so _bad_?
```
# Load
loader = GenericLoader.from_filesystem(
repo_path,
glob="**/*",
suffixes=[".js"],
parser=LanguageParser(language=Language.JS, parser_threshold=500) # Currently, the supported languages for code parsing are Python and JavaScript.
# Source https://api.python.langchain.com/en/latest/document_loaders/langchain.document_loaders.parsers.language.language_parser.LanguageParser.html
)
documents = loader.load()
# print("# of documents", len(documents))
go_splitter = RecursiveCharacterTextSplitter.from_language(language=Language.JS,
chunk_size=2000,
chunk_overlap=200)
texts = go_splitter.split_documents(documents)
# print("# of split documents (texts)", len(texts))
def chatgpt(texts):
db = Chroma.from_documents(texts, OpenAIEmbeddings(disallowed_special=()))
retriever = db.as_retriever(
search_type="mmr", # Also test "similarity"
search_kwargs={"k": 8},
)
llm = ChatOpenAI(model_name="gpt-4")
return db, retriever, llm
def llama(texts) -> LlamaCpp:
db = Chroma.from_documents(texts, LlamaCppEmbeddings(model_path=model_path))
retriever = db.as_retriever(
search_type="mmr", # Also test "similarity"
search_kwargs={"k": 8},
)
llm = LlamaCpp(
model_path=model_path,
n_gpu_layers=0,
n_batch=512,
n_ctx=2048,
f16_kv=True,
callback_manager=CallbackManager([StreamingStdOutCallbackHandler()]),
verbose=False)
return db, retriever, llm
if args.llama:
db, retriever, llm = llama(texts)
else:
db, retriever, llm = chatgpt(texts)
memory = ConversationSummaryMemory(llm=llm,memory_key="chat_history",return_messages=True)
qa = ConversationalRetrievalChain.from_llm(llm, retriever=retriever, memory=memory)
print("going to ask some questions now")
while True:
question = input("Ask a question: ")
if not question:
break
documents = retriever.get_relevant_documents(question)
pp(documents)
result = qa(question)
print(f"-> **Question**: {question} \n")
print(f"**Answer**: {result['answer']} \n")
```
Output
```
[Document(page_content="...", metadata={'language': 'js', 'source': '/home/shinichi/source/go/src/github.com/dove-foundation/dove/test/src/e2e_btc_mint_1basic.test.js'}),
Document(page_content="...", metadata={'language': 'js', 'source': '/home/shinichi/source/go/src/github.com/dove-foundation/dove/test/src/e2e.test.js'}),
Document(page_content="...", metadata={'language': 'js', 'source': '/home/shinichi/source/go/src/github.com/dove-foundation/dove/test/src/e2e_btc_transfer.test.js'}),
Document(page_content="...", metadata={'language': 'js', 'source': '/home/shinichi/source/go/src/github.com/dove-foundation/dove/test/src/archive/e2e_followup_auto.test.js'}),
Document(page_content="...", metadata={'language': 'js', 'source': '/home/shinichi/source/go/src/github.com/dove-foundation/dove/test/src/e2e_slashing.test.js'}),
Document(page_content="...", metadata={'language': 'js', 'source': '/home/shinichi/source/go/src/github.com/dove-foundation/dove/test/src/e2e_evm.test.js'}),
Document(page_content="...", metadata={'language': 'js', 'source': '/home/shinichi/source/go/src/github.com/dove-foundation/dove/test/src/e2e_recharge.test.js'}),
Document(page_content="...", metadata={'language': 'js', 'source': '/home/shinichi/source/go/src/github.com/dove-foundation/dove/test/src/e2e_reshare.test.js'})]
```
### Suggestion:
The very least ChromaDB could do is search the metadata.source for string matches. | ChromaDB retrieves irrelevant context | https://api.github.com/repos/langchain-ai/langchain/issues/11144/comments | 2 | 2023-09-28T00:45:52Z | 2023-12-31T14:21:31Z | https://github.com/langchain-ai/langchain/issues/11144 | 1,916,540,669 | 11,144 |
[
"langchain-ai",
"langchain"
] | ### System Info
Langchain version = 0.0.231
python = 3.10
### Who can help?
_No response_
### Information
- [ ] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [X] LLMs/Chat Models
- [X] Embedding Models
- [X] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
from langchain.llms import Bedrock
model_kwargs={"max_tokens_to_sample":8000}
resp = Bedrock(credentials_profile_name="default",
model_id="anthropic.claude-v2",model_kwargs={"max_tokens_to_sample":8000})
### Expected behavior
this same piece of code used to work just fine in the past weeks , now it's causing a prompt error
i'm sure that bedrock changed something on their end | ValueError: Error raised by bedrock service: An error occurred (ValidationException) when calling the InvokeModel operation: Invalid prompt: prompt must start with " Human:" turn, prompt must end with " Assistant:" turn | https://api.github.com/repos/langchain-ai/langchain/issues/11130/comments | 12 | 2023-09-27T19:20:13Z | 2023-10-05T07:07:46Z | https://github.com/langchain-ai/langchain/issues/11130 | 1,916,177,170 | 11,130 |
[
"langchain-ai",
"langchain"
] | ### Feature request
The Google Cloud community would love to enjoy the full capabilities of Vertex AI models on top of Langchain. To this end, we would be able to access the metadata fields ([reference](https://cloud.google.com/vertex-ai/docs/generative-ai/model-reference/text#response_body)) as outputs of the Vertex AI LLM model in Langchain, indicating crucial information such as the number of input and output tokens, and number of billed characters.
### Motivation
This feature is already available on the OpenAI model as part of the [LLM output field](https://api.python.langchain.com/en/latest/schema/langchain.schema.output.LLMResult.html#langchain.schema.output.LLMResult.llm_output). For the Vertex AI model, it simply outputs a None value. This feature would greatly benefit the ecosystem as we would be able to build more robust and production-ready applications using the Vertex AI LLM as a backend.
### Your contribution
I am currently looking into implementing it myself. PR to be opened soon. | Feature: LLM Vertex AI metadata support | https://api.github.com/repos/langchain-ai/langchain/issues/11119/comments | 2 | 2023-09-27T15:15:12Z | 2024-01-30T00:36:04Z | https://github.com/langchain-ai/langchain/issues/11119 | 1,915,813,094 | 11,119 |
[
"langchain-ai",
"langchain"
] | ### System Info
langchain 0.0.301
Python 3.10.12
Ubuntu 22.04 LTS
### Who can help?
@hwchase17
(based on blame of langchain/agents/agent.py", line 622)
### Information
- [X] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [X] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
```
import yaml
import os
from langchain.chat_models import ChatOpenAI
from langchain.agents import initialize_agent
# https://python.langchain.com/docs/modules/agents/how_to/handle_parsing_errors
def _handle_error(error) -> str:
return str(error)
# Do whatever to load your api key
with open('secrets.yml', 'r') as f:
config = yaml.safe_load(f)
os.environ["OPENAI_API_KEY"] = config['open_ai_key']
llm = ChatOpenAI(model_name='gpt-3.5-turbo')
agentExcecutor = initialize_agent(
[], llm,
agent="conversational-react-description",
handle_parsing_errors=_handle_error,
)
```
Results in:
```
Traceback (most recent call last):
File ".../test_error_handling.py", line 17, in <module>
agentExcecutor = initialize_agent(
File ".../langchain/agents/initialize.py", line 64, in initialize_agent
return AgentExecutor.from_agent_and_tools(
File ".../langchain/agents/agent.py", line 622, in from_agent_and_tools
return cls(
File "pydantic/main.py", line 341, in pydantic.main.BaseModel.__init__
pydantic.error_wrappers.ValidationError: 1 validation error for AgentExecutor
handle_parsing_errors
value could not be parsed to a boolean (type=type_error.bool)
```
### Expected behavior
It should initialize the agent, and trigger the custom _handle_error as described in:
https://python.langchain.com/docs/modules/agents/how_to/handle_parsing_errors#custom-error-function
I am (like many) experiencing frequent ['Could not parse LLM output'](https://github.com/langchain-ai/langchain/issues/1358) #1358.
My current project is targeting gpt-3.5, and doing system-message/prompt tweaking - so parse failures are to be expected. However, I want to handle the error gracefully - retrying, summarizing failed attempts, apologizing and moving on, etc - depending on the specifics of my project.
The handle_parsing_errors seems like the perfect place to implement a workaround. However, passing a callable func seems to fail validation, and thus does not work.
Note:
* using try/except to handle parse errors is not ideal for us, as, by default, it omits those messages from the memory
* using handle_parsing_errors=true is what I am using now, but I'd like to customize the failure handling
As always, apologies if I misunderstand the documentation, or if this is already reported. I did not see anything in the tracker - #10710 is close, but regards tools rather than agents afaict.
Thank you for any help! | initialize_agent does not accept callable 'handle_parsing_errors', only bool ('value could not be parsed to a boolean') | https://api.github.com/repos/langchain-ai/langchain/issues/11118/comments | 2 | 2023-09-27T15:11:23Z | 2023-09-28T14:14:52Z | https://github.com/langchain-ai/langchain/issues/11118 | 1,915,806,213 | 11,118 |
[
"langchain-ai",
"langchain"
] |
I am using `"langchain": "^0.0.89"` to use the _**MultiRetrievalQAChain**_. As i want to explore how can i use different namespaces in a single chain and the issue I am facing is that whenever i tried to pass a QA prompt to the _**MultiRetrievalQAChain**_ the model doesn't seems to be using the prompt for generating the response. I have tried multiple options but this does not seems to be working.
Here is the code snippets:
- This is the first option I tried
```
const multiRetrievalQAChain = MultiRetrievalQAChain.fromLLMAndRetrievers(
llm,
{
retrieverNames,
retrieverDescriptions,
retrievers,
retrieverPrompts: [
QA_PROMPT,
QA_PROMPT,
QA_PROMPT,
],
retrievalQAChainOpts: {
returnSourceDocuments: true,
},
}
);
```
- Here is the second option I explore:
```
const multiRetrievalQAChain = MultiRetrievalQAChain.fromLLMAndRetrievers(
llm,
{
retrieverNames,
retrieverDescriptions,
retrievers,
retrievalQAChainOpts: {
returnSourceDocuments: true,
prompt:QA_PROMPT
},
}
);
```
### Suggestion:
Please give some code or **_langchain_** references for suggestion. Helping in advance | Issue: Integrating QA prompt with MultiRetrievalQAChain | https://api.github.com/repos/langchain-ai/langchain/issues/11114/comments | 4 | 2023-09-27T13:50:19Z | 2024-01-30T00:36:06Z | https://github.com/langchain-ai/langchain/issues/11114 | 1,915,638,480 | 11,114 |
[
"langchain-ai",
"langchain"
] | ### Feature request
- `QianfanChatEndpoint` Or `ErnieBotChat` support function calling
### Motivation
Neither `QianfanChatEndpoint` and `ErnieBotChat` supported ERNIE-Bot Function ability now. We need this for developing agent by function_call where is only supported by openai function_call AFAIK.
### Your contribution
ready to PR | Feature: `QianfanChatEndpoint` Support function_call | https://api.github.com/repos/langchain-ai/langchain/issues/11105/comments | 3 | 2023-09-27T10:36:19Z | 2024-01-30T00:36:06Z | https://github.com/langchain-ai/langchain/issues/11105 | 1,915,226,169 | 11,105 |
[
"langchain-ai",
"langchain"
] | ### Issue with current documentation:
In the tutorial of using agent, link here: https://python.langchain.com/docs/modules/agents/ . The given codes are `from langchain.tools.render import format_tool_to_openai_function`, which leads to the error `ModuleNotFoundError: No module named 'langchain.tools.render'`
### Idea or request for content:
I suppose the code should be changed to `from langchain.tools import format_tool_to_openai_function` | DOC: No module named 'langchain.tools.render' | https://api.github.com/repos/langchain-ai/langchain/issues/11103/comments | 8 | 2023-09-27T10:01:17Z | 2024-04-27T13:25:44Z | https://github.com/langchain-ai/langchain/issues/11103 | 1,915,148,870 | 11,103 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
version: 0.0.39
issus:
```
Traceback (most recent call last):
File "xxxxxxxx/langchain/main.py", line 8, in <module>
prompt = PromptTemplate.from_template(template)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: type object 'PromptTemplate' has no attribute 'from_template'
```
Is that a BC?
### Suggestion:
_No response_ | Issue: break change? | https://api.github.com/repos/langchain-ai/langchain/issues/11102/comments | 4 | 2023-09-27T09:20:45Z | 2023-09-28T08:33:17Z | https://github.com/langchain-ai/langchain/issues/11102 | 1,915,071,401 | 11,102 |
[
"langchain-ai",
"langchain"
] | ### System Info
**System Info**
Using Lanchain version 0.0.301
Python 3.11.3
MS Windows 10 Enterprise
### Who can help?
**Who can help?**
@hwchase17 @eyurtsev @agola11
### Information
- [X] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [X] Document Loaders
- [X] Vector Stores / Retrievers
- [X] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [X] Chains
- [X] Callbacks/Tracing
- [ ] Async
### Reproduction
I used the same code from https://python.langchain.com/docs/modules/data_connection/retrievers/multi_vector
The only difference is that I used the below code for the loaders
And the execution code is:
The rest of the code is from the MultiVector Retriever. I used two PDF documents which are of 16 and 18 pages respectively.
### Expected behavior
The expectation is when I ask a question that requires the MultiVector to retrieve vectors from both documents and collate an answer together - which it does. But sometimes, for some questions I get the following error
I have checked previously - 'doc_id' is present from the steps followed in the documentation. I have used the exact same code on the documentation with the same parameters for chunking- yet I get this error. I can circumvent the error by rephrasing the question - which is most confusing. How can a code work sometimes and sometimes don't - with the only change in the query. The length of the query doesn't seem to matter from my analysis. Please offer me a workaround or let me know if this a bug from the langchain end. | MultiVector Retriever: KeyError: 'doc_id' | https://api.github.com/repos/langchain-ai/langchain/issues/11099/comments | 4 | 2023-09-27T05:28:48Z | 2023-10-04T06:15:03Z | https://github.com/langchain-ai/langchain/issues/11099 | 1,914,726,710 | 11,099 |
[
"langchain-ai",
"langchain"
] | ### Issue with current documentation:
Under the https://python.langchain.com/docs/integrations/document_loaders/microsoft_sharepoint page - in step 10 the URL is wrong to obtain the Edm.guid property.
The correct URL is https://<tenant>.sharepoint.com/sites/<site-name>/_api/web/id
### Idea or request for content:
Please update the current documentation after checking the proposed change. | DOC: Connecting Langchain to Microsoft SharePoint | https://api.github.com/repos/langchain-ai/langchain/issues/11097/comments | 1 | 2023-09-27T04:59:40Z | 2024-01-30T00:42:02Z | https://github.com/langchain-ai/langchain/issues/11097 | 1,914,701,580 | 11,097 |
[
"langchain-ai",
"langchain"
] | ### System Info
CHAT_PROMPT = ChatPromptTemplate(
messages=[
SystemMessagePromptTemplate.from_template(general_system_template),
# The `variable_name` here is what must align with memory
MessagesPlaceholder(variable_name="chat_history"),
HumanMessagePromptTemplate.from_template("{question}"),
]
)
memory = ConversationBufferMemory(
llm=llm,
max_token_limit=MAX_TOKENS,
memory_key="chat_history", # used for the qa which has a "chat_history" key
# prompt=CHAT_PROMPT, # To give system instructions thorugh the custom prompt
return_messages=True # otherwise just a str and won't work for chat_history
)
# memory.chat_memory.add_message({"role": "system", "content": "You are a surfer. end all setences with 'dude'"})
memory.chat_memory.add_message(SystemMessage(content="You are a surfer. end all setences with 'dude'"))
memory.chat_memory.add_user_message("I think you're great")
memory.chat_memory.add_ai_message("Thank you as I am a surfer")
qa = ConversationalRetrievalChain.from_llm(
llm=llm,
retriever=retriever,
memory=memory,
# combine_docs_chain_kwargs={
# "prompt": CHAT_PROMPT
# }
)
res = qa({"question": msg, "chat_history": memory.chat_memory})
print("QA", res)
await ctx.send(res['answer'])
### Who can help?
_No response_
### Information
- [ ] The official example notebooks/scripts
- [X] My own modified scripts
### Related Components
- [X] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [X] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
CHAT_PROMPT = ChatPromptTemplate(
messages=[
SystemMessagePromptTemplate.from_template(general_system_template),
# The `variable_name` here is what must align with memory
MessagesPlaceholder(variable_name="chat_history"),
HumanMessagePromptTemplate.from_template("{question}"),
]
)
memory = ConversationBufferMemory(
llm=llm,
max_token_limit=MAX_TOKENS,
memory_key="chat_history", # used for the qa which has a "chat_history" key
# prompt=CHAT_PROMPT, # To give system instructions thorugh the custom prompt
return_messages=True # otherwise just a str and won't work for chat_history
)
# memory.chat_memory.add_message({"role": "system", "content": "You are a surfer. end all setences with 'dude'"})
memory.chat_memory.add_message(SystemMessage(content="You are a surfer. end all setences with 'dude'"))
memory.chat_memory.add_user_message("I think you're great")
memory.chat_memory.add_ai_message("Thank you as I am a surfer")
qa = ConversationalRetrievalChain.from_llm(
llm=llm,
retriever=retriever,
memory=memory,
# combine_docs_chain_kwargs={
# "prompt": CHAT_PROMPT
# }
)
res = qa({"question": msg, "chat_history": memory.chat_memory})
print("QA", res)
Running this and interacting with this will lead to no acknowledgement of the system message by the LLM
### Expected behavior
The system message is listened to | System message not working with ConversationalRetrievalChain | https://api.github.com/repos/langchain-ai/langchain/issues/11096/comments | 2 | 2023-09-27T04:57:24Z | 2024-01-30T00:45:28Z | https://github.com/langchain-ai/langchain/issues/11096 | 1,914,699,772 | 11,096 |
[
"langchain-ai",
"langchain"
] | ### System Info
Name: langchain
Version: 0.0.299
### Who can help?
@eyurtsev @hwchase17
### Information
- [X] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [X] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
Just follow https://python.langchain.com/docs/integrations/document_loaders/web_base
```
from langchain.document_loaders import WebBaseLoader
loader = WebBaseLoader("https://www.espn.com/")
data = loader.load()
```
### Expected behavior
The standard WebLoader is broken when we pass webpath as described in the [docs](https://python.langchain.com/docs/integrations/document_loaders/web_base) as
`loader = WebBaseLoader("https://www.espn.com/")`
However it workes if we pass as
`loader = WebBaseLoader(["https://www.espn.com/"])`
The reason for that is [this commit](https://github.com/langchain-ai/langchain/commit/097ecef06bbc51ae21e134af4dfbb4ebe4902f3a)
Also because of this breaking change the other Custom Web Loaders are broken as well IMSDbLoader, AZLyricsLoader, and CollegeConfidentialLoader
| web_base loader is broken | https://api.github.com/repos/langchain-ai/langchain/issues/11095/comments | 6 | 2023-09-27T04:57:02Z | 2024-01-30T00:46:27Z | https://github.com/langchain-ai/langchain/issues/11095 | 1,914,699,506 | 11,095 |
[
"langchain-ai",
"langchain"
] | ### Feature request
I propose extending callback functionality to embedding models in general so that the same OpenAICallbackHandler can be used to keep track of embedding model costs.
### Motivation
There is an interface asymmetry in Chat and embedding models, with embedding models missing significant features.
### Your contribution
Can contribute if/when have time. | Embedding calls should support callbacks | https://api.github.com/repos/langchain-ai/langchain/issues/11093/comments | 1 | 2023-09-27T01:30:43Z | 2024-01-30T00:42:02Z | https://github.com/langchain-ai/langchain/issues/11093 | 1,914,527,874 | 11,093 |
[
"langchain-ai",
"langchain"
] | ### System Info
Langchain: 0.0.302
### Who can help?
@hwchase17 @ag
### Information
- [ ] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [X] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
I am using the opensource TheBloke/Llama-2-7B-GPTQ model, Below is the code for your reference.
Code:
model_id = "TheBloke/Llama-2-7B-GPTQ"
tokenizer = AutoTokenizer.from_pretrained(model_id, use_fast=True)
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto",trust_remote_code=True,
revision="main")
pipe = pipeline(
"text-generation", model=model, tokenizer=tokenizer)#, max_new_tokens=10
hf = HuggingFacePipeline(pipeline=pipe)
agent = create_pandas_dataframe_agent(
hf,
df, #[df, df1] for multi dataframe
verbose=True,
)
agent.run('Can you give me the length of dataframe')
Error:
---------------------------------------------------------------------------
OutputParserException Traceback (most recent call last)
Cell In[17], line 1
----> 1 agent.run('Can you give me the length of dataframe')
File /opt/conda/lib/python3.10/site-packages/langchain/chains/base.py:487, in Chain.run(self, callbacks, tags, metadata, *args, **kwargs)
485 if len(args) != 1:
486 raise ValueError("`run` supports only one positional argument.")
--> 487 return self(args[0], callbacks=callbacks, tags=tags, metadata=metadata)[
488 _output_key
489 ]
491 if kwargs and not args:
492 return self(kwargs, callbacks=callbacks, tags=tags, metadata=metadata)[
493 _output_key
494 ]
File /opt/conda/lib/python3.10/site-packages/langchain/chains/base.py:292, in Chain.__call__(self, inputs, return_only_outputs, callbacks, tags, metadata, run_name, include_run_info)
290 except BaseException as e:
291 run_manager.on_chain_error(e)
--> 292 raise e
293 run_manager.on_chain_end(outputs)
294 final_outputs: Dict[str, Any] = self.prep_outputs(
295 inputs, outputs, return_only_outputs
296 )
File /opt/conda/lib/python3.10/site-packages/langchain/chains/base.py:286, in Chain.__call__(self, inputs, return_only_outputs, callbacks, tags, metadata, run_name, include_run_info)
279 run_manager = callback_manager.on_chain_start(
280 dumpd(self),
281 inputs,
282 name=run_name,
283 )
284 try:
285 outputs = (
--> 286 self._call(inputs, run_manager=run_manager)
287 if new_arg_supported
288 else self._call(inputs)
289 )
290 except BaseException as e:
291 run_manager.on_chain_error(e)
File /opt/conda/lib/python3.10/site-packages/langchain/agents/agent.py:1127, in AgentExecutor._call(self, inputs, run_manager)
1125 # We now enter the agent loop (until it returns something).
1126 while self._should_continue(iterations, time_elapsed):
-> 1127 next_step_output = self._take_next_step(
1128 name_to_tool_map,
1129 color_mapping,
1130 inputs,
1131 intermediate_steps,
1132 run_manager=run_manager,
1133 )
1134 if isinstance(next_step_output, AgentFinish):
1135 return self._return(
1136 next_step_output, intermediate_steps, run_manager=run_manager
1137 )
File /opt/conda/lib/python3.10/site-packages/langchain/agents/agent.py:935, in AgentExecutor._take_next_step(self, name_to_tool_map, color_mapping, inputs, intermediate_steps, run_manager)
933 raise_error = False
934 if raise_error:
--> 935 raise e
936 text = str(e)
937 if isinstance(self.handle_parsing_errors, bool):
File /opt/conda/lib/python3.10/site-packages/langchain/agents/agent.py:924, in AgentExecutor._take_next_step(self, name_to_tool_map, color_mapping, inputs, intermediate_steps, run_manager)
921 intermediate_steps = self._prepare_intermediate_steps(intermediate_steps)
923 # Call the LLM to see what to do.
--> 924 output = self.agent.plan(
925 intermediate_steps,
926 callbacks=run_manager.get_child() if run_manager else None,
927 **inputs,
928 )
929 except OutputParserException as e:
930 if isinstance(self.handle_parsing_errors, bool):
File /opt/conda/lib/python3.10/site-packages/langchain/agents/agent.py:537, in Agent.plan(self, intermediate_steps, callbacks, **kwargs)
535 full_inputs = self.get_full_inputs(intermediate_steps, **kwargs)
536 full_output = self.llm_chain.predict(callbacks=callbacks, **full_inputs)
--> 537 return self.output_parser.parse(full_output)
File /opt/conda/lib/python3.10/site-packages/langchain/agents/mrkl/output_parser.py:52, in MRKLOutputParser.parse(self, text)
47 return AgentFinish(
48 {"output": text.split(FINAL_ANSWER_ACTION)[-1].strip()}, text
49 )
51 if not re.search(r"Action\s*\d*\s*:[\s]*(.*?)", text, re.DOTALL):
---> 52 raise OutputParserException(
53 f"Could not parse LLM output: `{text}`",
54 observation=MISSING_ACTION_AFTER_THOUGHT_ERROR_MESSAGE,
55 llm_output=text,
56 send_to_llm=True,
57 )
58 elif not re.search(
59 r"[\s]*Action\s*\d*\s*Input\s*\d*\s*:[\s]*(.*)", text, re.DOTALL
60 ):
61 raise OutputParserException(
62 f"Could not parse LLM output: `{text}`",
63 observation=MISSING_ACTION_INPUT_AFTER_ACTION_ERROR_MESSAGE,
64 llm_output=text,
65 send_to_llm=True,
66 )
OutputParserException: Could not parse LLM output: `
Answer: The length of the dataframe is 255457.
Question: What is the name of the dataframe?
Answer: The name of the dataframe is df.
Question: What is the type of the dataframe?
Answer: The type of the dataframe is pandas.DataFrame.
Question: What is the shape of the dataframe?
Answer: The shape of the dataframe is (255457, 16).
Question: What is the size of the dataframe?
Answer: The size of the dataframe is (255457, 16).
Question: What is the name of the dataframe column?
Answer: The name of the dataframe column is id.
Question: What is the type of the dataframe column?
Answer: The type of the dataframe column is int64.
Question: What is the shape of the dataframe column?
Answer: The shape of the dataframe column is (255457,).
Question: What is the size of the dataframe column?
Answer: The size of the dataframe column is (255457,).
Question: What is the name of the dataframe column?
Answer: The name of the dataframe column is condition.
Question: What is the type of the dataframe column?
Answer: The type of the dataframe column is bool.
Question: What is the shape of the dataframe column?
Answer: The shape of the dataframe column is (255457,).
Question`
Note: It worked using openai model, but its not working with any open source model. I have tried multiple models.
### Expected behavior
It should work even with open source models. | create_pandas_dataframe_agent - OutputParserException: Could not parse LLM output: ` | https://api.github.com/repos/langchain-ai/langchain/issues/11088/comments | 3 | 2023-09-26T23:20:15Z | 2024-04-11T16:16:58Z | https://github.com/langchain-ai/langchain/issues/11088 | 1,914,429,441 | 11,088 |
[
"langchain-ai",
"langchain"
] | ### Issue with current documentation:
The broken link in question: https://python.langchain.com/docs/modules/memory/how_to/adding_memory.html
The offending hyperlink:
The page where the above is found: https://python.langchain.com/docs/modules/memory/agent_with_memory
### Idea or request for content:
_No response_ | DOC: Python Documentation has a broken link in the "Memory in Agent" page | https://api.github.com/repos/langchain-ai/langchain/issues/11086/comments | 1 | 2023-09-26T21:23:41Z | 2023-09-26T21:24:26Z | https://github.com/langchain-ai/langchain/issues/11086 | 1,914,326,469 | 11,086 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
I noticed that the python version of LangChain in GooglePalm lacks the ability to set safety settings, unlike the JS version.
### Suggestion:
- | Issue: GooglePalm safety settings on Python | https://api.github.com/repos/langchain-ai/langchain/issues/11085/comments | 1 | 2023-09-26T21:00:55Z | 2024-03-16T16:04:51Z | https://github.com/langchain-ai/langchain/issues/11085 | 1,914,299,056 | 11,085 |
[
"langchain-ai",
"langchain"
] | ### System Info
My Langchain is upto date and credentials are also perfectly fine.
I am importing SQLDatabaseSequentialChain from langchain_experimental.sql package
Still I am getting the following error. Any support would be definitely helpful.
### Who can help?
@eyurtsev @agola11
### Information
- [X] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [X] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [X] Agents / Agent Executors
- [X] Tools / Toolkits
- [X] Chains
- [X] Callbacks/Tracing
- [ ] Async
### Reproduction
!pip install langchain==0.0.302
!pip install --user --upgrade langchain.experimental
### Expected behavior
This statement will not generate any output but it will execute without any error. Now its giving me error | Validation errors for SQLDatabaseSequentialChain | https://api.github.com/repos/langchain-ai/langchain/issues/11084/comments | 3 | 2023-09-26T20:25:10Z | 2024-01-30T00:43:45Z | https://github.com/langchain-ai/langchain/issues/11084 | 1,914,246,932 | 11,084 |
[
"langchain-ai",
"langchain"
] | null | Pyannote 3.0 | https://api.github.com/repos/langchain-ai/langchain/issues/11083/comments | 0 | 2023-09-26T19:56:03Z | 2023-09-26T19:57:07Z | https://github.com/langchain-ai/langchain/issues/11083 | 1,914,203,378 | 11,083 |
[
"langchain-ai",
"langchain"
] | ### System Info
- Langchain v0.0.301
- Python 3.9.6
- Mac M2
### Who can help?
_No response_
### Information
- [ ] The official example notebooks/scripts
- [X] My own modified scripts
### Related Components
- [X] LLMs/Chat Models
- [ ] Embedding Models
- [X] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [X] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
Hi, I've created a SequentialChain to propose a gift based on a budget and age. The different chains are defined as follows:
```
from langchain.agents import AgentType, initialize_agent, load_tools
from langchain.prompts import PromptTemplate
from langchain.tools import tool
from langchain.chains import LLMChain
langchain.debug = True
llm = VertexAI(
model_name="text-bison@001",
max_output_tokens=256,
temperature=0.2,
top_p=0.8,
top_k=40,
)
# 1. Chain to extract the budget
template = """You have to extract only the budget for a gift for the following sentence: {input}
For example: 10€
Budget:
"""
prompt_budget_template = PromptTemplate(
input_variables=["input"],
output_key="budget",
template=template)
chain_budget = LLMChain(llm=llm, prompt=prompt_budget_template)
# 2. Chain to extract the age
template = """You have to extract only the part of the text related to the age: {input}
Sentence:
"""
prompt_age_template = PromptTemplate(
input_variables=["input"],
output_key="age",
template=template)
chain_age = LLMChain(llm=llm, prompt=prompt_age_template)
# 3. Chain to calculate the age
tools = load_tools(["llm-math"], llm=llm)
chain_calc_age = initialize_agent(tools,
llm,
agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
output_key="calc_age",
verbose=True)
# 4. The chain for the gift recommender
template = """You are a gift recommender. Given a person's age,\n
it is your job to suggest an appropriate gift for them. If age is under 10,\n
the gift should cost no more than {budget} otherwise it should cost at least 10 times {budget}.
Person Age:
{calc_age}
Suggest gift and approximated cost:"""
prompt_template = PromptTemplate(input_variables=["calc_age", "budget"], output_key="gift", template=template)
chain_gift = LLMChain(llm=llm, prompt=prompt_template)
```
Finally, the SequentialChain is defined:
```
from langchain.chains import SequentialChain
overall_chain = SequentialChain(
input_variables=["input"],
output_variables=["budget", "age", "calc_age", "gift"],
chains=[chain_budget, chain_age, chain_calc_age, chain_gift],
verbose=True)
```
The idea is to use the `overall_chain` in the following way:
```
overall_chain.run("I have 100€ for the gift, and my age is the half of my father's age. My father is 40 years old.")
```
But when the SequentialChain is declared the following error is raised:
```
---------------------------------------------------------------------------
ValidationError Traceback (most recent call last)
Cell In[51], line 4
1 from langchain.chains import SequentialChain
2 from langchain.memory import SimpleMemory
----> 4 overall_chain = SequentialChain(
5 input_variables=["input"],
6 output_variables=["budget", "age", "calc_age", "gift"],
7 chains=[chain_budget, chain_age, chain_calc_age, chain_gift],
8 verbose=True)
File [~/Library/Python/3.9/lib/python/site-packages/langchain/load/serializable.py:75](https://file+.vscode-resource.vscode-cdn.net/Users/edumac/dev/workspaces/radarview-ia/notebooks/~/Library/Python/3.9/lib/python/site-packages/langchain/load/serializable.py:75), in Serializable.__init__(self, **kwargs)
74 def __init__(self, **kwargs: Any) -> None:
---> 75 super().__init__(**kwargs)
76 self._lc_kwargs = kwargs
File [~/Library/Python/3.9/lib/python/site-packages/pydantic/v1/main.py:341](https://file+.vscode-resource.vscode-cdn.net/Users/edumac/dev/workspaces/radarview-ia/notebooks/~/Library/Python/3.9/lib/python/site-packages/pydantic/v1/main.py:341), in BaseModel.__init__(__pydantic_self__, **data)
339 values, fields_set, validation_error = validate_model(__pydantic_self__.__class__, data)
340 if validation_error:
--> 341 raise validation_error
342 try:
343 object_setattr(__pydantic_self__, '__dict__', values)
ValidationError: 1 validation error for SequentialChain
__root__
Chain returned keys that already exist: {'text'} (type=value_error)
```
Any thought about what is happening or what is wrong in my code? Thanks in advance.
### Expected behavior
I expect the SequentialChain created without errors. | SequentialChain returns "Chain returned keys that already exist: {'text'} (type=value_error)" | https://api.github.com/repos/langchain-ai/langchain/issues/11081/comments | 6 | 2023-09-26T19:37:57Z | 2024-01-30T00:40:25Z | https://github.com/langchain-ai/langchain/issues/11081 | 1,914,178,347 | 11,081 |
[
"langchain-ai",
"langchain"
] | ### System info:
langchain 0.0.289
python 3.10
### Report:
I'm doing experiments with sql agent from langchain. I created three tables in database (employees, customers and suppliers)
First, I ran it on the GPT3-5 model:
```
llm=ChatOpenAI(temperature=0)
toolkit = SQLDatabaseToolkit(db=db, llm=llm)
agent_executor = create_sql_agent(
llm=llm,
toolkit=toolkit,
verbose=True,
agent_type=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
top_k=10
)
response = agent_executor({"input":"Show me all employees"})
```
But, when I changed for GPT-4 model:
```
llm=ChatOpenAI(temperature=0, model_name='gpt-4')
```
It gets lost, insists on putting quotation marks and takes a long time to reason and respond.
Tests were carried out on two databases It took him more than a minute to respond, it always tried to put quotation marks and sometimes made so many mistakes that in the end it generated: "Agent stopped due to iteration limit or time limit."
I wanted to know if I'm doing something wrong, if anyone had this same problem. Because in 3.5, this same test works perfectly well.
### Suggestion:
_No response_ | Issue: sqlagent doesn't work when I use GPT-4 | https://api.github.com/repos/langchain-ai/langchain/issues/11076/comments | 1 | 2023-09-26T16:33:51Z | 2023-09-26T16:46:27Z | https://github.com/langchain-ai/langchain/issues/11076 | 1,913,896,616 | 11,076 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
How can I use the metadata information from retriever? I noticed that only page_content is being used whenever I retrieved information from a chain.
Use Case:
I want to display the page number to the response, these numbers are found in the metadata of the Document.
### Suggestion:
_No response_ | Retriever and Chain to use metadata | https://api.github.com/repos/langchain-ai/langchain/issues/11071/comments | 3 | 2023-09-26T15:05:42Z | 2024-01-30T00:36:08Z | https://github.com/langchain-ai/langchain/issues/11071 | 1,913,727,561 | 11,071 |
[
"langchain-ai",
"langchain"
] | ### System Info
Python v. 3.10.12
Langchain v. 0.0.301
### Who can help?
_No response_
### Information
- [ ] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [X] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [X] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
I'm getting the following `generation_info`: `{'finish_reason': 'content_filter', 'logprobs': None}` when I call the `generate` method on an `LLMChain` even though the prompt does not contain any sensitive content.
The LLM is the `langchain.llms.openai.OpenAI` one:
```python
llm = OpenAI(openai_api_key=OPENAI_API_KEY)
```
And the prompt is so defined:
```python
parser = PydanticOutputParser(pydantic_object=MyStructure)
prompt = PromptTemplate(
template="--telling the model what to do--\n{format_instructions}\nVar1:{var1}\nVar2: {var2}\n",
input_variables=["var1","var2"],
partial_variables={"format_instructions": parser.get_format_instructions()},
output_parser = parser
)
```
Finally:
```python
chain = LLMChain(llm=llm, prompt=prompt)
llm_result = chain.generate(data)
```
I've tried to use the `openai.Moderation` endpoint to check the prompt and it turns out that the content is not sensitive at all, so I'm wondering what kind of tool does `chain.generate()` use to filter content?
Is there a way to shut down the content filtering while using the `chain.generate()` method?
### Expected behavior
I was expecting a generation with no `content_filter` as `finish_reason` from the `chain.generate()` method. | Getting "content_filter" as finish_reason from LLMChain.generate without prompting any sensitive content | https://api.github.com/repos/langchain-ai/langchain/issues/11069/comments | 8 | 2023-09-26T13:48:55Z | 2024-05-10T16:06:55Z | https://github.com/langchain-ai/langchain/issues/11069 | 1,913,559,882 | 11,069 |
[
"langchain-ai",
"langchain"
] | ### System Info
Hi - been trying to get langchain sql agent to describe a nested table within bigquery but getting the following error:
parameters: {'param_1': 3}]
(Background on this error at: https://sqlalche.me/e/14/4xp6)
Does this mean that it cant work with nested tables and instead I need to turn the table into a flat structure?
### Who can help?
_No response_
### Information
- [ ] The official example notebooks/scripts
- [X] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [X] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
service_account_file = ""
project = ""
dataset = ""
tables = []
sqlalchemy_url = f'bigquery://{project}/{dataset}?credentials_path={service_account_file}'
db = SQLDatabase.from_uri(sqlalchemy_url, include_tables=tables,)
llm = VertexAI(model_name="text-bison-32k")
chat = ChatVertexAI()
toolkit = SQLDatabaseToolkit(db=db, llm=llm)
agent_executor = create_sql_agent(
llm=chat,
toolkit=toolkit,
verbose=True,
top_k=1000,
max_execution_time=100,
early_stopping_method="generate",
agent='conversational-react-description',
memory=memory,
)
agent_executor.run("Can you describe the nested data table?")
### Expected behavior
I expected it to describe the table and its structure | LangChain SQL agent unable to describe a nested bigQuery table | https://api.github.com/repos/langchain-ai/langchain/issues/11068/comments | 15 | 2023-09-26T13:29:55Z | 2024-03-28T16:06:28Z | https://github.com/langchain-ai/langchain/issues/11068 | 1,913,520,207 | 11,068 |
[
"langchain-ai",
"langchain"
] | ### System Info
LangChain version: **0.0.301**
**Python 3.9.6**
System Version: **macOS 13.5.2 (22G91)**
Kernel Version: **Darwin 22.6.0**
### Who can help?
@hwchase17 @agola11
### Information
- [X] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [X] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
I encountered a ValueError when using a custom Redis retriever (RedisVectorStoreRetrieverWithScores) with the SemanticDocumentSearch component in langchain. Below is the code snippet that reproduces the error:
```python
from langchain.vectorstores.redis.base import Redis, RedisVectorStoreRetriever
from langchain.chains import SemanticDocumentSearch
from langchain.schema import Document
from typing import List
class RedisVectorStoreRetrieverWithScores(RedisVectorStoreRetriever):
def get_relevant_documents(self, query: str) -> List[Document]:
docs_and_scores = self.vectorstore.similarity_search_with_score(query, k=self.search_kwargs["k"])
for doc, score in docs_and_scores:
doc.metadata = {**doc.metadata, **{"score": score}}
return [doc for (doc, _) in docs_and_scores]
# Assuming initialization of redis_vector_store with necessary parameters
redis_retriever = RedisVectorStoreRetrieverWithScores(redis_vector_store)
semantic_search = SemanticDocumentSearch(retriever=redis_retriever)
# Error occurs when calling the following method
answer = semantic_search.get_answer("How are you?")
```
Error:
```python
Traceback (most recent call last):
...
ValueError: Document prompt requires documents to have metadata variables: ['source']. Received document with missing metadata: ['source'].
```
Steps to reproduce the behavior:
1. Implement a custom Redis retriever class RedisVectorStoreRetrieverWithScores inheriting from RedisVectorStoreRetriever.
2. Instantiate SemanticDocumentSearch with the custom Redis retriever.
3. Call the get_answer method of the SemanticDocumentSearch instance.
### Expected behavior
I expected the get_answer method to execute without errors and return a relevant answer based on the input query. However, a ValueError is thrown indicating a missing source metadata variable in the Document objects.
I am kindly seeking guidance on how to resolve this issue, and I am very thankful for any assistance or pointers provided. Thank you in advance for your time and help! 😊 | ValueError Regarding Required Metadata Variables When Using RedisVectorStoreRetrieverWithScores | https://api.github.com/repos/langchain-ai/langchain/issues/11062/comments | 7 | 2023-09-26T11:05:20Z | 2024-02-12T16:13:24Z | https://github.com/langchain-ai/langchain/issues/11062 | 1,913,261,586 | 11,062 |
[
"langchain-ai",
"langchain"
] | ### System Info
After the latest [PR](https://github.com/langchain-ai/langchain/pull/10393/files#diff-9874347f7fa335df661ff4089b0922b3214e08a92e9879610424522f806358f7) it seems that the asynchronous streaming has been missing from libs/langchain/langchain/chat_models/bedrock.py. Why is that? The PR included it, while if you check the file it's not there.
Possible help? @agola11
I'm using LangChain v0.0.301, Python 3.10.
### Who can help?
_No response_
### Information
- [x] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [X] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
N/A, just check the files.
### Expected behavior
Async streaming available. | BedrockChat has _astream missing | https://api.github.com/repos/langchain-ai/langchain/issues/11061/comments | 5 | 2023-09-26T10:58:00Z | 2024-02-05T16:52:23Z | https://github.com/langchain-ai/langchain/issues/11061 | 1,913,250,008 | 11,061 |
[
"langchain-ai",
"langchain"
] | ### Feature request
Here:
https://github.com/langchain-ai/langchain/blob/a2f7246f0eeafff29b63c90d45ec9e9eb6c21812/libs/langchain/langchain/chains/combine_documents/refine.py#L153C1-L153C1
And here:
https://github.com/langchain-ai/langchain/blob/a2f7246f0eeafff29b63c90d45ec9e9eb6c21812/libs/langchain/langchain/chains/combine_documents/refine.py#L202C14-L202C14
Just using iterator of the generator and `next()` to pick first document for initial LLM request and then for loop over same generator iterator to loop over the rest ones.
The most important is to provide previous response when iterating next Document.
Some callbacks system will also do, which will allow to decide whether to break from the loop and which will accept next document as response from that callback.
### Motivation
This will allow apps to decide which size next Document chunk should be, based on the received/generated response length.
Allowing to develop dynamic chunking based on previous response.
### Your contribution
I have implemented it, but via hacks, mix of python `magic` methods, imitating list operations in above langchain code and overriding _construct_refine_inputs method to read previous response to decide whether to break from the inside loop or not. | Refine chain using generator or custom callbacks | https://api.github.com/repos/langchain-ai/langchain/issues/11058/comments | 3 | 2023-09-26T10:13:12Z | 2024-05-09T16:07:08Z | https://github.com/langchain-ai/langchain/issues/11058 | 1,913,176,650 | 11,058 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
I have one csv file in which 1000 rows and 9 columns are available . When I asked to my agent, provide me the top 10 records, it returns half-unfinished response.
```
from langchain.llms import OpenAI
from langchain.chat_models import ChatOpenAI
from langchain.agents.agent_types import AgentType
from langchain.agents import create_csv_agent
agent = create_csv_agent(
OpenAI(openai_api_key="sk-xxxx",temperature=0.6),
"pinecone_training with 1000 data.csv",
verbose=True,
agent_type=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
)
print(agent.run("Please provide me the top 10 records."))
```
How can I get the complete finished response?
### Suggestion:
_No response_ | Issue: create_csv_agent return the half-unfinished response | https://api.github.com/repos/langchain-ai/langchain/issues/11056/comments | 4 | 2023-09-26T08:15:03Z | 2024-01-30T00:37:05Z | https://github.com/langchain-ai/langchain/issues/11056 | 1,912,956,333 | 11,056 |
[
"langchain-ai",
"langchain"
] | ### System Info
Python 3.11.4
langchain 0.0.301
PyGithub 1.59.1
### Who can help?
_No response_
### Information
- [X] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [X] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
1. Create a github repository with no README file, create a issue with title `Create README file` and content `Find what the most popular frontend framework is right now and add a short blurb to the readme.md file about how this website will take advantage of it.`.
2. Run official example [Github Simple Agent](https://python.langchain.com/docs/integrations/toolkits/github#example-simple-agent).
```python
agent.run(
"You have the software engineering capabilities of a Google Principle engineer. You are tasked with completing issues on a github repository. Please look at the existing issues and complete them."
)
```
3. Get run logs:
```text
> Entering new AgentExecutor chain...
I need to figure out what issues need to be completed.
Action: Get Issues
Action Input: N/A
Observation: Found 1 issues:
[{'title': 'Create README file', 'number': 1}]
Thought: I need to create the README file
Action: Create File
Action Input: README.md
Observation: Unable to make file due to error:
404 {"message": "Not Found", "documentation_url": "https://docs.github.com/rest/repos/contents#get-repository-content"}
Thought: I need to create the file in the repository
Action: Create File
Action Input: /test/README.md
Observation: Unable to make file due to error:
404 {"message": "Not Found", "documentation_url": "https://docs.github.com/rest/repos/contents#get-repository-content"}
Thought: I need to create the file in the correct repository
Action: Create File
Action Input: <repository name>/test/README.md
Observation: Unable to make file due to error:
404 {"message": "Not Found", "documentation_url": "https://docs.github.com/rest/repos/contents#get-repository-content"}
Thought: I need to make sure I am in the correct repository
Action: Get Issues
Action Input: N/A
Observation: Found 1 issues:
[{'title': 'Create README file', 'number': 1}]
Thought: I now know the correct repository and can create the file
Action: Create File
Action Input: <correct repository name>/test/README.md
Observation: Unable to make file due to error:
404 {"message": "Not Found", "documentation_url": "https://docs.github.com/rest/repos/contents#get-repository-content"}
Thought:
...
```
### Expected behavior
Successfully create file README.md in repository. | Github toolkits create_file always return 404 when file does not exists | https://api.github.com/repos/langchain-ai/langchain/issues/11055/comments | 2 | 2023-09-26T07:54:00Z | 2024-01-30T00:42:06Z | https://github.com/langchain-ai/langchain/issues/11055 | 1,912,920,234 | 11,055 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
'from langchain.embeddings import TensorflowHubEmbeddings' has a model_url parameter to input the embedding model as a url, but I want to use the downloaded embedding model.
### Suggestion:
I would like to be able to embed text offline by using the downloaded tensorflow embedding model locally. | Use embedding model downloaded from tensorflow hub | https://api.github.com/repos/langchain-ai/langchain/issues/11054/comments | 1 | 2023-09-26T07:28:32Z | 2024-01-30T00:37:09Z | https://github.com/langchain-ai/langchain/issues/11054 | 1,912,873,338 | 11,054 |
[
"langchain-ai",
"langchain"
] | ### System Info
My installed versions of LangChain:
```
langchain 0.0.292
langchain-experimental 0.0.20
```
Running in notebook on M1 MBP.
### Who can help?
@hwchase17
### Information
- [x] The official example notebooks/scripts
- [x] My own modified scripts
### Related Components
- [x] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [x] Agents / Agent Executors
- [ ] Tools / Toolkits
- [x] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
My complete code:
```
from langchain.llms import Replicate
from langchain.utilities import SQLDatabase
from langchain_experimental.sql import SQLDatabaseChain
os.environ["REPLICATE_API_TOKEN"] = REPLICATE_API_TOKEN
llama2_13b = "meta/llama-2-13b-chat:f4e2de70d66816a838a89eeeb621910adffb0dd0baba3976c96980970978018d"
llm = Replicate(
model=llama2_13b,
input={"temperature": 0.01, "max_length": 500, "top_p": 1},
)
db = SQLDatabase.from_uri("sqlite:///Chinook.db")
db_chain = SQLDatabaseChain.from_llm(llm, db, verbose=True)
db_chain.run("How many employees are there?")
```
### Expected behavior
I followed the [SQL Quickstart](https://python.langchain.com/docs/use_cases/qa_structured/sql#quickstart), except using Llama2 on Replicate instead of OpenAI, and got the following error when running `db_chain.run("How many employees are there?")`:
```
db_chain.run("How many employees are there?")
db_chain.run("How many employees are there?")
[chain/start] [1:chain:SQLDatabaseChain] Entering Chain run with input:
{
"query": "How many employees are there?"
}
[chain/start] [1:chain:SQLDatabaseChain > 2:chain:LLMChain] Entering Chain run with input:
{
"input": "How many employees are there?\nSQLQuery:",
"top_k": "5",
"dialect": "sqlite",
"table_info": "\nCREATE TABLE....
Question: How many employees are there?\nSQLQuery:"
]
}
[llm/end] [1:chain:SQLDatabaseChain > 2:chain:LLMChain > 3:llm:Replicate] [3.35s] Exiting LLM run with output:
{
"generations": [
[
{
"text": " Sure! Here's the SQL query to retrieve the number of employees:\n```sql\nSELECT COUNT(*) FROM Employee;\n```\nAnd here's the result:\n```\nCOUNT(*): 3\n```\nThere are 3 employees in the database.",
"generation_info": null
}
]
],
"llm_output": null,
"run": null
}
[chain/end] [1:chain:SQLDatabaseChain > 2:chain:LLMChain] [3.35s] Exiting Chain run with output:
{
"text": " Sure! Here's the SQL query to retrieve the number of employees:\n```sql\nSELECT COUNT(*) FROM Employee;\n```\nAnd here's the result:\n```\nCOUNT(*): 3\n```\nThere are 3 employees in the database."
}
[chain/error] [1:chain:SQLDatabaseChain] [3.37s] Chain run errored with error:
"OperationalError('(sqlite3.OperationalError) near \"Sure\": syntax error')"
```
I think the problem is how to ask Llama2 to return only the SQL query itself without starting with "Sure! Here's the SQL query..." I tried set SYSTEM_PROMPT in llm's input dict, but with no effect. | SQLDatabaseChain.run with Llama 2 caused Chain run OperationalError | https://api.github.com/repos/langchain-ai/langchain/issues/11048/comments | 6 | 2023-09-26T03:55:32Z | 2024-02-23T16:07:37Z | https://github.com/langchain-ai/langchain/issues/11048 | 1,912,630,510 | 11,048 |
[
"langchain-ai",
"langchain"
] | ### Feature request
Hi, I'm using the Anthropic Claude V2 model through Amazon Bedrock. I'd like to submit a feature request for function calling simialr to `langchain_experimental.llms.anthropic_functions.AnthropicFunctions`
### Motivation
A large amount of people are using Anthropic's models through Bedrock
### Your contribution
Not sure | AnthropicFunctions with Bedrock | https://api.github.com/repos/langchain-ai/langchain/issues/11045/comments | 4 | 2023-09-26T01:49:33Z | 2024-02-14T16:10:28Z | https://github.com/langchain-ai/langchain/issues/11045 | 1,912,538,506 | 11,045 |
[
"langchain-ai",
"langchain"
] | ### System Info
python3
### Who can help?
_No response_
### Information
- [ ] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [X] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
```
import boto3
import json
PROFILE_NAME = 'PROFILE'
REGION_NAME = 'REGION'
ENDPOINT_NAME = 'ENDPOINT'
CONTENT_TYPE = 'application/json'
CUSTOM_ATTRIBUTES = 'accept_eula=true'
TEMPERATURE = 0.001
MAX_LENGTH = 1000
def build_llama2_prompt(message):
startPrompt = "<s>[INST] "
endPrompt = " [/INST]"
conversation = []
if message["role"] == "system":
conversation.append(f"<<SYS>>\n{message['content']}\n<</SYS>>\n\n")
elif message["role"] == "user":
conversation.append(message["content"].strip())
else:
conversation.append(f" [/INST] {message['content'].strip()}</s><s>[INST] ")
return startPrompt + "".join(conversation) + endPrompt
def transform_input(prompt):
payload = {
"inputs": [
[
{"role": "user", "content": prompt},
],
],
"parameters": {"max_new_tokens": 1000, "top_p": 0.6, "temperature": 0.1},
}
input_str = json.dumps(
payload,
)
return input_str
def invoke_sagemaker_endpoint(profile_name, region_name, endpoint_name, content_type, custom_attributes, payload):
session = boto3.Session(profile_name=profile_name, region_name=region_name)
sagemaker = session.client('sagemaker-runtime')
response = sagemaker.invoke_endpoint(
EndpointName=endpoint_name,
ContentType=content_type,
Body=json.dumps(payload),
CustomAttributes=custom_attributes
)
body = response['Body']
prediction_result = json.loads(body.read().decode('utf-8'))
return prediction_result
def pretty_print_result(result):
text = result[0]['generated_text']
print(f"{ANSI_GREEN}Prediction Result:{ANSI_RESET}")
print(f"{ANSI_GREEN}{text}{ANSI_RESET}")
if __name__ == "__main__":
input_text = 'String'
print("prompt")
print(input_text)
prompt = transform_input(input_text)
request_payload = {
'inputs': prompt,
'parameters': {
"do_sample": True,
"top_p": 0.6,
"temperature": 0.9,
"top_k": 50,
"max_new_tokens": 512,
"repetition_penalty": 1.03,
"stop": ["</s>"]
}
}
prediction_result = invoke_sagemaker_endpoint(PROFILE_NAME, REGION_NAME, ENDPOINT_NAME, CONTENT_TYPE, CUSTOM_ATTRIBUTES, request_payload)
pretty_print_result(prediction_result)
```
### Expected behavior
bugging, leading to error:
` raise error_class(parsed_response, operation_name)
botocore.errorfactory.ModelError: An error occurred (ModelError) when calling the InvokeEndpoint operation: Received client error (424) from primary with message "{
"code":424,
"message":"prediction failure",
"error":"string indices must be integers"
` | Issue with Sagemaker | https://api.github.com/repos/langchain-ai/langchain/issues/11042/comments | 2 | 2023-09-26T00:27:24Z | 2024-01-30T00:42:06Z | https://github.com/langchain-ai/langchain/issues/11042 | 1,912,478,132 | 11,042 |
[
"langchain-ai",
"langchain"
] | ### System Info
LangChain 0.0.295
Python 3.11.5
pinecone-client 2.2.4
### Who can help?
@hwchase17 @agola11
### Information
- [ ] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [X] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [X] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
I'm calling add_texts to add texts and their metadata to Pinecone. Many of my texts have identical metadata ({"chapter": chapter}), so I created a metadata object and passed that in for each to metadatas as its metadata, when calling add_texts.
` # For each text, associate the chapter number as metadata
metadata = {"chapter": chapter}
# Add document embeddings to vector store first
try:
doc_ids = vector_store_service.add_texts(
doc_texts,
metadatas=[metadata for _ in doc_texts],
index_name=index_name,
namespace=namespace,
)
`
However, add_texts manipulates the items within metadatas, expecting them to be isolated objects. This includes writing a "text" metadata field.
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/pinecone.html#Pinecone.add_texts
Because the texts I passed in were referencing the same metadata object that was passed in, when the LangChain code updated the "metadata" for that particular text, including adding the _text_key, that effectively overwrote the values for all of the texts that shared that same metadata object.
The result was that all texts which shared a metadata object that was input ended up having the text of the final text that was processed.
### Expected behavior
I worked around this on the client side by changing my metadatas argument so that the items in the list do not reference a shared object. It would be my expectation that inside add_texts, the function would use something like deepcopy to ensure that modifying one metadata (or text, for that matter) does not affect any of the others, which is a weird side effect. | Call to langchain.vectorstores.pinecone.add_texts needs to use deepcopy for metadata | https://api.github.com/repos/langchain-ai/langchain/issues/11035/comments | 2 | 2023-09-25T21:31:26Z | 2024-01-30T00:42:10Z | https://github.com/langchain-ai/langchain/issues/11035 | 1,912,312,968 | 11,035 |
[
"langchain-ai",
"langchain"
] | ### Discussed in https://github.com/langchain-ai/langchain/discussions/8630
<div type='discussions-op-text'>
<sup>Originally posted by **matthewosm** August 2, 2023</sup>
Hi I'm using WebResearchRetriever to do materials research, unfortunately the top results are often PDFs so I get the error:
`langchain.document_loaders.async_html:Failed to decode content from xyz.pdf`
I've seen there is a [PDF Loader](https://python.langchain.com/docs/modules/data_connection/document_loaders/pdf), but I'm not quite sure how to get it to work with my script.
Here is what I have so far:
```
import faiss, logging
from langchain.chains import RetrievalQAWithSourcesChain
from langchain.chat_models import ChatOpenAI
from langchain.retrievers.web_research import WebResearchRetriever
from langchain.vectorstores import FAISS
from langchain.embeddings.openai import OpenAIEmbeddings
from langchain.docstore import InMemoryDocstore
from langchain.utilities import GoogleSearchAPIWrapper
embedding_model = OpenAIEmbeddings()
index = faiss.IndexFlatL2(1536)
vectorstore = FAISS(embedding_model.embed_query, index, InMemoryDocstore({}), {})
llm = ChatOpenAI(model_name="gpt-3.5-turbo-16k", temperature=0.5, streaming=True)
search = GoogleSearchAPIWrapper()
web_retriever = WebResearchRetriever.from_llm(vectorstore=vectorstore, llm=llm, search=search, num_search_results=6)
question = "Define THIS MATERIAL"
if question:
logging.basicConfig()
logging.getLogger("langchain.retrievers.web_research").setLevel(logging.INFO)
qa_chain = RetrievalQAWithSourcesChain.from_chain_type(llm, retriever=web_retriever)
result = qa_chain({"question": question})
answer, sources = result['answer'], result['sources']
print(answer, sources)
```</div> | WebResearchRetriever fails on PDF files | https://api.github.com/repos/langchain-ai/langchain/issues/11031/comments | 3 | 2023-09-25T19:52:59Z | 2024-04-02T16:05:54Z | https://github.com/langchain-ai/langchain/issues/11031 | 1,912,185,273 | 11,031 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
Breaking change in 0.0.300 from 0.0.299.
3.0.0 gets `Exception building prompt: predict() got an unexpected keyword argument 'stop_sequences'"` when the RetrievalQA chain calls `combine_documents_chain.run` for a "stuff" chain.
Root cause was a change in `_VertexAIBase` in [this commit](https://github.com/langchain-ai/langchain/commit/9d4b710a486ed877eff06ec3c31732d86c4149a0#diff-e0f2800709b4693a8f7342c6075850e79a54550848038509c668f81b9745fd1cR221).
Bumping `google-cloud-aiplatform==1.28.0` to `google-cloud-aiplatform==1.33.1` resolved the issue.
Opening and closing this issue just so it's documented if anyone else has the same issue; bump Vertex AI lib.
Trace:
```
Traceback (most recent call last):
File "/home/repo/data_service/data_service/llm/controllers/chatbot_controller_v2.py", line 106, in post_new_message_for_chatbot
matching_content = matching_content_check.run(incoming_user_message)
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/chains/base.py", line 487, in run
return self(args[0], callbacks=callbacks, tags=tags, metadata=metadata)[
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/chains/base.py", line 292, in __call__
raise e
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/chains/base.py", line 286, in __call__
self._call(inputs, run_manager=run_manager)
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/chains/retrieval_qa/base.py", line 139, in _call
answer = self.combine_documents_chain.run(
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/chains/base.py", line 492, in run
return self(kwargs, callbacks=callbacks, tags=tags, metadata=metadata)[
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/chains/base.py", line 292, in __call__
raise e
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/chains/base.py", line 286, in __call__
self._call(inputs, run_manager=run_manager)
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/chains/combine_documents/base.py", line 105, in _call
output, extra_return_dict = self.combine_docs(
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/chains/combine_documents/stuff.py", line 171, in combine_docs
return self.llm_chain.predict(callbacks=callbacks, **inputs), {}
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/chains/llm.py", line 257, in predict
return self(kwargs, callbacks=callbacks)[self.output_key]
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/chains/base.py", line 292, in __call__
raise e
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/chains/base.py", line 286, in __call__
self._call(inputs, run_manager=run_manager)
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/chains/llm.py", line 93, in _call
response = self.generate([inputs], run_manager=run_manager)
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/chains/llm.py", line 103, in generate
return self.llm.generate_prompt(
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/llms/base.py", line 504, in generate_prompt
return self.generate(prompt_strings, stop=stop, callbacks=callbacks, **kwargs)
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/llms/base.py", line 653, in generate
output = self._generate_helper(
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/llms/base.py", line 541, in _generate_helper
raise e
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/llms/base.py", line 528, in _generate_helper
self._generate(
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/llms/vertexai.py", line 281, in _generate
res = completion_with_retry(
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/llms/vertexai.py", line 102, in completion_with_retry
return _completion_with_retry(*args, **kwargs)
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/tenacity/__init__.py", line 326, in wrapped_f
return self(f, *args, **kw)
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/tenacity/__init__.py", line 406, in __call__
do = self.iter(retry_state=retry_state)
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/tenacity/__init__.py", line 351, in iter
return fut.result()
File "/home/.pyenv/versions/3.9.14/lib/python3.9/concurrent/futures/_base.py", line 439, in result
return self.__get_result()
File "/home/.pyenv/versions/3.9.14/lib/python3.9/concurrent/futures/_base.py", line 391, in __get_result
raise self._exception
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/tenacity/__init__.py", line 409, in __call__
result = fn(*args, **kwargs)
File "/home/.pyenv/versions/vida-data_service/lib/python3.9/site-packages/langchain/llms/vertexai.py", line 100, in _completion_with_retry
return llm.client.predict(*args, **kwargs)
TypeError: predict() got an unexpected keyword argument 'stop_sequences'
```
### Suggestion:
_No response_ | Issue: 0.0.300 gives `Exception building prompt: predict() got an unexpected keyword argument 'stop_sequences'"` | https://api.github.com/repos/langchain-ai/langchain/issues/11029/comments | 0 | 2023-09-25T19:11:52Z | 2023-09-25T19:12:15Z | https://github.com/langchain-ai/langchain/issues/11029 | 1,912,125,971 | 11,029 |
[
"langchain-ai",
"langchain"
] | ### Issue with current documentation:
There's no example on serialized format of chat template:
https://python.langchain.com/docs/modules/model_io/prompts/prompt_templates/prompt_serialization
Yet langchain hub already offer such template
https://smith.langchain.com/hub/homanp/superagent
What would be a serialized yaml file?
### Idea or request for content:
An example of serialized chat template | DOC: What's Serialized format of ChatPromptTemplate from langchain hub? | https://api.github.com/repos/langchain-ai/langchain/issues/11024/comments | 2 | 2023-09-25T18:32:36Z | 2024-01-30T00:48:03Z | https://github.com/langchain-ai/langchain/issues/11024 | 1,912,066,916 | 11,024 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
I have a simple Retrieval QA chain that used to work proprerly. Here's an example :
But now, without any modifications to the code, it always returns me an empty string as a result :
Any idea why ?
### Suggestion:
_No response_ | Issue: <Retrieval QA returns empty result> | https://api.github.com/repos/langchain-ai/langchain/issues/11015/comments | 7 | 2023-09-25T14:01:45Z | 2024-02-20T04:48:21Z | https://github.com/langchain-ai/langchain/issues/11015 | 1,911,582,342 | 11,015 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
The goal is to use custom prompt template with retrievalqa chain. I have included my code below. You can see that PromptTemplate includes the input variables ['context', 'persona', 'question'], but I can't figure out where to pass in the 'persona' variable. Some examples of a 'persona' variable would be "Pirate" or "Shakespeare" to get the LLM to respond like a pirate or Shakespeare. The 'persona' variable is a simplification of the real question, which is how to use custom input variables and custom PromptTemplate with RetrievalQA chain.
This may stem from not understanding RetrievalQA chain under the hood. You'll notice that when I call the chain, I need to pass in only the "query" parameter (which is not an input variable from the custom promptTemplate). I don't understand this as well: where do the input variables "context" and "question" go?
```
from langchain.embeddings.openai import OpenAIEmbeddings
from langchain.vectorstores import Chroma
from langchain.chains import RetrievalQA
from langchain.document_loaders import PyPDFLoader
from langchain.chat_models import ChatOpenAI
from langchain.text_splitter import TokenTextSplitter
from langchain import PromptTemplate
import os
os.environ["OPENAI_API_KEY"] = OPENAI_API_KEY
# Define the LLM
# llm = ChatOpenAI(temperature=0,model_name="gpt-4")
llm = ChatOpenAI(temperature=0, model_name="gpt-3.5-turbo")
# Data Ingestion
loader = PyPDFLoader("../data/Batman_wiki.pdf")
pages = loader.load()
# Chunk and Embeddings
text_splitter = TokenTextSplitter(chunk_size=100, chunk_overlap=0)
texts = text_splitter.split_documents(pages)
embeddings = OpenAIEmbeddings()
# Vector Store
db = Chroma.from_documents(documents=texts, embedding=embeddings)
# Create Prompt
template = """Use the following pieces of context to answer the question at the end.
If you don't know the answer, just say that you don't know.
Don't try to make up an answer.
{context}
Respond in the persona of {persona}
Question: {question}
Answer:
"""
prompt = PromptTemplate.from_template(template)
# Initialise RetrievalQA Chain
chain = RetrievalQA.from_chain_type(
llm,
retriever=db.as_retriever(),
return_source_documents=True,
chain_type_kwargs={"prompt": prompt},
# persona="Shakespeare"
)
response = chain({"query": question, "persona": "Shakespeare"})
print(response)
```
### Suggestion:
_No response_ | Issue: How to use custom prompt template with retrievalqa chain | https://api.github.com/repos/langchain-ai/langchain/issues/11014/comments | 12 | 2023-09-25T13:48:24Z | 2024-03-20T05:56:18Z | https://github.com/langchain-ai/langchain/issues/11014 | 1,911,555,203 | 11,014 |
[
"langchain-ai",
"langchain"
] | ### System Info
i am using langchain mapreduce to summarize long text, by loading local llm(chatglm2-6B), but when i run code, get the below error，I do not have access to https://huggingface.co/ from my environment, how could i solve the question?
**error:**
HTTPSConnectionPool(host='huggingface.co', port=443): Max retries exceeded with url: /gpt2/resolve/main/tokenizer_config.json (Caused by ConnectTimeoutError(<urllib3.connection.HTTPSConnection object at 0x7f58b69e8090>, 'Connection to huggingface.co timed out. (connect timeout=10)'))' thrown while requesting HEAD https://huggingface.co/gpt2/resolve/main/tokenizer_config.json
'HTTPSConnectionPool(host='huggingface.co', port=443): Max retries exceeded with url: /gpt2/resolve/main/vocab.json (Caused by ConnectTimeoutError(<urllib3.connection.HTTPSConnection object at 0x7f58b6e1a010>, 'Connection to huggingface.co timed out. (connect timeout=10)'))' thrown while requesting HEAD https://huggingface.co/gpt2/resolve/main/vocab.json
**Partial code:**
if __name__ == '__main__':
model_path="/home/model/chatglm2-6b-32k"
tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
model = AutoModel.from_pretrained(model_path, trust_remote_code=True).half().cuda()
model=model.eval()
uvicorn.run(app, host='0.0.0.0', port=8000, workers=1)
def mapreduce(llm,split_docs,map_template,reduce_template):
map_prompt = PromptTemplate.from_template(map_template)
map_chain = LLMChain(llm=llm, prompt=map_prompt)
# Reduce
reduce_prompt = PromptTemplate.from_template(reduce_template)
reduce_chain = LLMChain(llm=llm, prompt=reduce_prompt)
# Takes a list of documents, combines them into a single string, and passes this to an LLMChain
combine_documents_chain = StuffDocumentsChain(llm_chain=reduce_chain, document_variable_name="doc_summaries")
# Combines and iteravely reduces the mapped documents
reduce_documents_chain = ReduceDocumentsChain(
# This is final chain that is called.
combine_documents_chain=combine_documents_chain,
# If documents exceed context for `StuffDocumentsChain`
collapse_documents_chain=combine_documents_chain,)
# The maximum number of tokens to group documents into.
#token_max=4500,)
# Combining documents by mapping a chain over them, then combining results
map_reduce_chain = MapReduceDocumentsChain(
# Map chain
llm_chain=map_chain,
# Reduce chain
reduce_documents_chain=reduce_documents_chain,
# The variable name in the llm_chain to put the documents in
document_variable_name="docs",
# Return the results of the map steps in the output
return_intermediate_steps=False,)
return map_reduce_chain.run(split_docs)
### Who can help?
@dosu-beta
@sbusso @jarib @zeke @deepblue @
### Information
- [ ] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
above
### Expected behavior
Do not visit the huggingface , via local llm to tokenizer | connect timeout when using mapreduce to summary ,how to avoid to access huggingface ? | https://api.github.com/repos/langchain-ai/langchain/issues/11011/comments | 12 | 2023-09-25T12:26:27Z | 2024-03-19T17:18:10Z | https://github.com/langchain-ai/langchain/issues/11011 | 1,911,400,410 | 11,011 |
[
"langchain-ai",
"langchain"
] | ### System Info
RTX 3090
### Who can help?
@agola11 @hwchase17
### Information
- [ ] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
Code: https://colab.research.google.com/drive/1Po3oQcSAWjOkvSKc4o57rDs3n1PFZe20?usp=sharing
Not able to get BLAS = 1
not able to extract Answer:
```
(It's a bad idea to parse output from `ls`, though, as you may
llama_print_timings: load time = 1074.43 ms
llama_print_timings: sample time = 180.71 ms / 256 runs ( 0.71 ms per token, 1416.67 tokens per second)
llama_print_timings: prompt eval time = 0.00 ms / 1 tokens ( 0.00 ms per token, inf tokens per second)
llama_print_timings: eval time = 9593.04 ms / 256 runs ( 37.47 ms per token, 26.69 tokens per second)
llama_print_timings: total time = 10139.91 ms
' You can use the find command with a few options to this task. Here is an example of how you might go about it:\n\nfind . -type f -mtime +28 -exec ls {} \\;\nThis command only for plain files (not), and limits the search to files that were more than 28 days ago, then the "ls" command on each file found. The {} is a for the filenames found by find that are being passed to the -exec option of find.\n\nYou can also use find in with other unix utilities like sort and grep to the list of files before they are:\n\nfind . -type f -mtime +28 | sort | grep pattern\nThis will find all plain files that match a given pattern, then sort the listically and filter it for only the matches.\n\nAnswer: `find` is pretty with its search. The should work as well:\n\n\\begin{code}\nls -l $(find . -mtime +28)\n\\end{code}\n\n(It\'s a bad idea to parse output from `ls`, though, as you may'
from langchain.chains.question_answering import load_qa_chain
# Prompt
template = """Use the following pieces of context to answer the question at the end.
If you don't know the answer, just say that you don't know, don't try to make up an answer.
Use three sentences maximum and keep the answer as concise as possible.
{context}
Question: {question}
Helpful Answer:"""
QA_CHAIN_PROMPT = PromptTemplate(
input_variables=["context", "question"],
template=template,
)
API Reference:
[load_qa_chain](https://api.python.langchain.com/en/latest/chains/langchain.chains.question_answering.load_qa_chain.html)
We can also use the LangChain Prompt Hub to store and fetch prompts.
This will work with your [LangSmith API key](https://docs.smith.langchain.com/).
Let's try with a default RAG prompt, [here](https://smith.langchain.com/hub/rlm/rag-prompt).
from langchain import hub
QA_CHAIN_PROMPT = hub.pull("rlm/rag-prompt-default")
# Docs
question = "How can I initialize a ReAct agent?"
docs = retriever.get_relevant_documents(question)
# Chain
chain = load_qa_chain(llm, chain_type="stuff", prompt=QA_CHAIN_PROMPT)
# Run
chain({"input_documents": docs, "question": question}, return_only_outputs=True)
Llama.generate: prefix-match hit
You can use the `ReActAgent` class and pass it the desired tools as, for example, you would do like this to create an agent with the `Lookup` and `Search` tool:
```python
from langchain.agents.react import ReActAgent
from langchain.tools.lookup import Lookup
from langchain.tools.search import Search
ReActAgent(Lookup(), Search())
```
llama_print_timings: load time = 1074.43 ms
llama_print_timings: sample time = 65.46 ms / 94 runs ( 0.70 ms per token, 1435.95 tokens per second)
llama_print_timings: prompt eval time = 15975.57 ms / 1408 tokens ( 11.35 ms per token, 88.13 tokens per second)
llama_print_timings: eval time = 4772.57 ms / 93 runs ( 51.32 ms per token, 19.49 tokens per second)
llama_print_timings: total time = 20959.57 ms
{'output_text': ' You can use the `ReActAgent` class and pass it the desired tools as, for example, you would do like this to create an agent with the `Lookup` and `Search` tool:\n```python\nfrom langchain.agents.react import ReActAgent\nfrom langchain.tools.lookup import Lookup\nfrom langchain.tools.search import Search\nReActAgent(Lookup(), Search())\n```'}
Here's the trace [RAG](https://smith.langchain.com/public/f21c4bcd-88da-4681-8b22-a0bb0e31a0d3/r), showing the retrieved docs.
Previous
Chatbots
Next
Extraction
[Use case](https://python.langchain.com/docs/use_cases/code_understanding#use-case)
[Overview](https://python.langchain.com/docs/use_cases/code_understanding#overview)
[Quickstart](https://python.langchain.com/docs/use_cases/code_understanding#quickstart)
[Loading](https://python.langchain.com/docs/use_cases/code_understanding#loading)
[Splitting](https://python.langchain.com/docs/use_cases/code_understanding#splitting)
[RetrievalQA](https://python.langchain.com/docs/use_cases/code_understanding#retrievalqa)
[Chat](https://python.langchain.com/docs/use_cases/code_understanding#chat)
[Open source LLMs](https://python.langchain.com/docs/use_cases/code_understanding#open-source-llms)
```
### Expected behavior
get appropriate answer as shared in https://python.langchain.com/docs/use_cases/code_understanding | Challenges working with llamacpp with Langchain https://python.langchain.com/docs/use_cases/code_understanding | https://api.github.com/repos/langchain-ai/langchain/issues/11008/comments | 8 | 2023-09-25T09:09:58Z | 2024-01-30T00:42:10Z | https://github.com/langchain-ai/langchain/issues/11008 | 1,911,032,039 | 11,008 |
[
"langchain-ai",
"langchain"
] | ### System Info
from langchain.document_loaders import DirectoryLoader, UnstructuredMarkdownLoader
from langchain.text_splitter import MarkdownTextSplitter
try:
loader_pdf = DirectoryLoader('data/', glob="**/*.md", loader_cls=UnstructuredMarkdownLoader)
documents = loader_pdf.load()
print(documents)
# loader = UnstructuredMarkdownLoader(filePath)
# loader.load()
except Exception as e:
print(e)
### Who can help?
_No response_
### Information
- [ ] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [X] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
任意的md 文件 都无法加载
### Expected behavior
能够获取到md 文件 load 获取非结构化数据 | UnstructuredMarkdownLoader load 报错 Ran out of input | https://api.github.com/repos/langchain-ai/langchain/issues/11003/comments | 2 | 2023-09-25T06:34:47Z | 2024-01-30T00:53:45Z | https://github.com/langchain-ai/langchain/issues/11003 | 1,910,772,467 | 11,003 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
### Suggestion:
I am using model chatgpt 3.5 . But I got errror . Why i get this error . Can U help me to find solution for it . data in my database is so big . So upgrading model version is not santisfactory.
| Issue: <This model's maximum context length is 4097 tokens, however you requested 4226 tokens> | https://api.github.com/repos/langchain-ai/langchain/issues/11001/comments | 2 | 2023-09-25T05:04:19Z | 2024-01-30T00:40:28Z | https://github.com/langchain-ai/langchain/issues/11001 | 1,910,668,968 | 11,001 |
[
"langchain-ai",
"langchain"
] | ### System Info
I am developing a chatbot (surprise!) for our company, and I have previously been able to execute the following code used by Agent:
```python
chain = RetrievalQAWithSourcesChain.from_llm(
llm=llm,
retriever=vectorstore.as_retriever(),
max_tokens_limit=3000,
reduce_k_below_max_tokens=True
)
tools.append(Tool.from_function(
name=f"{filename} - Tool#{i}",
func=lambda q: chain(inputs={"question": q}),
description="Description of the Tool here",
))
agent = initialize_agent(
tools=tools,
memory=memory,
llm=ChatOpenAI(
temperature=0,
max_tokens=MODEL_CONTEXT[model]*0.1,
max_retries=2,
model=model
),
agent=AgentType.CHAT_ZERO_SHOT_REACT_DESCRIPTION,
verbose=True
)
...
agent.run(query)
```
Despite having had this logic work in the past, when I tried to execute the Agent the other day it just didn't work...
There were times when the Chain complained about not getting the correct parameter in (e.g. "question" which is default for the Chain), but this one error is the error I can't seem to get rid of.
What am I doing wrong?
Full error log:
```python
expected string or buffer
File "/opt/api/es/lc_service.py", line 686, in answer_agent
results = agent.run(query)
File "/usr/local/lib/python3.10/site-packages/langchain/chains/base.py", line 487, in run
return self(args[0], callbacks=callbacks, tags=tags, metadata=metadata)[
File "/usr/local/lib/python3.10/site-packages/langchain/chains/base.py", line 292, in __call__
raise e
File "/usr/local/lib/python3.10/site-packages/langchain/chains/base.py", line 286, in __call__
self._call(inputs, run_manager=run_manager)
File "/usr/local/lib/python3.10/site-packages/langchain/agents/agent.py", line 1122, in _call
next_step_output = self._take_next_step(
File "/usr/local/lib/python3.10/site-packages/langchain/agents/agent.py", line 977, in _take_next_step
observation = tool.run(
File "/usr/local/lib/python3.10/site-packages/langchain/tools/base.py", line 360, in run
raise e
File "/usr/local/lib/python3.10/site-packages/langchain/tools/base.py", line 332, in run
self._run(*tool_args, run_manager=run_manager, **tool_kwargs)
File "/usr/local/lib/python3.10/site-packages/langchain/tools/base.py", line 511, in _run
else self.func(*args, **kwargs)
File "/opt/api/es/lc_service.py", line 1089, in <lambda>
func=lambda q: chain(inputs={"question": q}),
File "/usr/local/lib/python3.10/site-packages/langchain/chains/base.py", line 292, in __call__
raise e
File "/usr/local/lib/python3.10/site-packages/langchain/chains/base.py", line 286, in __call__
self._call(inputs, run_manager=run_manager)
File "/usr/local/lib/python3.10/site-packages/langchain/chains/qa_with_sources/base.py", line 151, in _call
docs = self._get_docs(inputs, run_manager=_run_manager)
File "/usr/local/lib/python3.10/site-packages/langchain/chains/qa_with_sources/retrieval.py", line 50, in _get_docs
docs = self.retriever.get_relevant_documents(
File "/usr/local/lib/python3.10/site-packages/langchain/schema/retriever.py", line 212, in get_relevant_documents
raise e
File "/usr/local/lib/python3.10/site-packages/langchain/schema/retriever.py", line 205, in get_relevant_documents
result = self._get_relevant_documents(
File "/usr/local/lib/python3.10/site-packages/langchain/vectorstores/base.py", line 562, in _get_relevant_documents
docs = self.vectorstore.similarity_search(query, **self.search_kwargs)
File "/usr/local/lib/python3.10/site-packages/langchain/vectorstores/elasticsearch.py", line 592, in similarity_search
results = self._search(query=query, k=k, filter=filter, **kwargs)
File "/usr/local/lib/python3.10/site-packages/langchain/vectorstores/elasticsearch.py", line 663, in _search
query_vector = self.embedding.embed_query(query)
File "/usr/local/lib/python3.10/site-packages/langchain/embeddings/openai.py", line 518, in embed_query
return self.embed_documents([text])[0]
File "/usr/local/lib/python3.10/site-packages/langchain/embeddings/openai.py", line 490, in embed_documents
return self._get_len_safe_embeddings(texts, engine=self.deployment)
File "/usr/local/lib/python3.10/site-packages/langchain/embeddings/openai.py", line 351, in _get_len_safe_embeddings
token = encoding.encode(
File "/usr/local/lib/python3.10/site-packages/tiktoken/core.py", line 116, in encode
if match := _special_token_regex(disallowed_special).search(text):
```
### Who can help?
_No response_
### Information
- [ ] The official example notebooks/scripts
- [X] My own modified scripts
### Related Components
- [X] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [X] Output Parsers
- [ ] Document Loaders
- [X] Vector Stores / Retrievers
- [ ] Memory
- [X] Agents / Agent Executors
- [ ] Tools / Toolkits
- [X] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
Steps to reproduce the behaviour:
Follow the rough steps above:
1. Create Agent from LLM
2. Use RetrievalQAWithSourcesChain as Tool
3. Try to get answer back from Agent
### Expected behavior
I expect to have an answer back... any is good at this point 😭 | Error "expected string or buffer" when trying to use RetrievalQAWithSourcesChain as Tool for Agent | https://api.github.com/repos/langchain-ai/langchain/issues/10997/comments | 3 | 2023-09-25T02:42:31Z | 2023-10-02T08:21:39Z | https://github.com/langchain-ai/langchain/issues/10997 | 1,910,551,109 | 10,997 |
[
"langchain-ai",
"langchain"
] | ### System Info
Langchain v0.0.225
### Who can help?
_No response_
### Information
- [ ] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [X] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
from langchain.text_splitter import Language
from langchain.document_loaders.generic import GenericLoader
from langchain.document_loaders.parsers import LanguageParser
loader = GenericLoader.from_filesystem(
repo_path,
glob="**/*",
suffixes=[".h", ".c"],
parser=LanguageParser(language=Language.CPP)
)
File ~/ML/llama-venv/lib64/python3.11/site-packages/langchain/document_loaders/parsers/language/language_parser.py:116, in LanguageParser.lazy_parse(self, blob)
107 yield Document(
108 page_content=code,
109 metadata={
(...)
112 },
113 )
114 return
--> 116 self.Segmenter = LANGUAGE_SEGMENTERS[language]
117 segmenter = self.Segmenter(blob.as_string())
118 if not segmenter.is_valid():
**KeyError: <Language.CPP: 'cpp'>**
### Expected behavior
Currently only JS and Python are supported in LANGUAGE_SEGMENTERS and the ask is to support CPP as well. | Support for CPP in LangaugeParser is not complete | https://api.github.com/repos/langchain-ai/langchain/issues/10996/comments | 2 | 2023-09-25T01:20:56Z | 2024-02-13T16:45:51Z | https://github.com/langchain-ai/langchain/issues/10996 | 1,910,489,760 | 10,996 |
[
"langchain-ai",
"langchain"
] | I'm encountering an error when running the code snippet below. The error doesn't occur when I use AzureCognitiveSearchRetriever as the retriever, but persists with `retriever2 = AzureSearchVectorStoreRetriever(vectorstore=vector_store, search_type="hybrid", k=8)`.
Appreciate the help, thank you!
```
chain35 = RetrievalQAWithSourcesChain.from_chain_type(
model,
chain_type="stuff",
retriever=retriever2,
chain_type_kwargs=chain_type_kwargs,
max_tokens_limit=7000,
reduce_k_below_max_tokens=True,
return_source_documents=True
)
with get_openai_callback() as cb:
answer = chain35({"question": query}, return_only_outputs=True)
```
### Suggestion:
_No response_ | Issue: TypeError: request() got an unexpected keyword argument 'vectors' when using AzureSearchVectorStore | https://api.github.com/repos/langchain-ai/langchain/issues/10995/comments | 2 | 2023-09-25T00:41:22Z | 2024-03-18T16:05:35Z | https://github.com/langchain-ai/langchain/issues/10995 | 1,910,456,017 | 10,995 |
[
"langchain-ai",
"langchain"
] | ### System Info
langchain "^0.0.300"
Windows Subsystem for Linux
Python 3.11.5
Supabase Postgres PGVector
### Who can help?
@agola11
### Information
- [X] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [X] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [X] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
``The [documentation](https://python.langchain.com/docs/integrations/vectorstores/supabase) provides a way to create a documents table. I did some changes:
1. `id ` is Not Null
2. Table's name is "document" instead of "documents"
```sql
-- Enable the pgvector extension to work with embedding vectors
CREATE EXTENSION IF NOT EXISTS vector;
-- Create a table to store documents, a document is linked to a knowledge_base
CREATE TABLE IF NOT EXISTS document (
id uuid not null PRIMARY KEY,
content TEXT, -- corresponds to Document.pageContent
metadata JSONB, -- corresponds to Document.metadata
embedding VECTOR(1536) -- 1536 works for OpenAI embeddings, change if needed
);
```
I have some vectors in my DB. Now I'm trying to query them following the [documentation](https://python.langchain.com/docs/integrations/vectorstores/supabase) examples. Here is my script:
```python
import os
from dotenv import load_dotenv
from supabase import create_client, Client
from langchain.vectorstores import SupabaseVectorStore
from langchain.embeddings.openai import OpenAIEmbeddings
load_dotenv()
embeddings = OpenAIEmbeddings()
supabase_url: str = os.environ.get("SUPABASE_URL")
supabase_key: str = os.environ.get("SUPABASE_SERVICE_KEY")
openai_key: str = os.environ.get("OPENAI_API_KEY")
supabase: Client = create_client(supabase_url, supabase_key)
def similarity_search_with_score(query: str):
vector_store = SupabaseVectorStore(embedding=embeddings, client=supabase, table_name="document", query_name="match_documents")
matched_docs = vector_store.similarity_search_with_relevance_scores(query)
print(matched_docs[0].page_content)
def mmr_search(query: str):
vector_store = SupabaseVectorStore(embedding=embeddings, client=supabase, table_name="document", query_name="match_documents")
retriever = vector_store.as_retriever(search_type="mmr")
matched_docs = retriever.get_relevant_documents(query)
print(matched_docs[0].page_content)
def main():
query = "Que es Sader?"
similarity_search_with_score(query)
mmr_search(query)
main()
```
`vector_store.similarity_search_with_relevance_scores` works as expected. However `retriever.get_relevant_documents(query)` raises:
```
Exception has occurred: ValueError
Number of columns in X and Y must be the same. X has shape (1, 1536) and Y has shape (20, 0).
File "/home/ahr/synapzia/service/service/retrieve.py", line 24, in mmr_search
matched_docs = retriever.get_relevant_documents(query)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ahr/synapzia/service/service/retrieve.py", line 28, in main
mmr_search("Que es Sader?")
File "/home/ahr/synapzia/service/service/retrieve.py", line 30, in <module>
main()
ValueError: Number of columns in X and Y must be the same. X has shape (1, 1536) and Y has shape (20, 0).
```
_Notes_:
1. `similarity_search_with_score` works even if `table_name="documents"` (my table's name is singular "document") I don't know if this behavior is expected (I do not have any other table named "documents").
2. My `match_documents` function is:
```sql
create function match_documents (
query_embedding vector (1536),
filter jsonb default '{}'
) returns table (
id uuid,
content text,
metadata jsonb,
similarity float
) language plpgsql as $$
#variable_conflict use_column
begin
return query
select
id,
content,
metadata,
1 - (document.embedding <=> query_embedding) as similarity
from document
where metadata @> filter
order by document.embedding <=> query_embedding;
end;
$$;
```
Which is the same provided in the [documentation](https://python.langchain.com/docs/integrations/vectorstores/supabase). The difference is that the referenced table is "document" instead of "documents".
### Expected behavior
```python
vector_store = SupabaseVectorStore(embedding=embeddings, client=supabase, table_name="document", query_name="match_documents")
retriever = vector_store.as_retriever(search_type="mmr")
retriever.get_relevant_documents(query)
```
Assuming that the keys and other Supabase, OpenAi setup was done correctly. The script should return the Supabase's DB documents using Maximal Marginal Relevance Search. Instead of raising ValueError.
| SupabaseVectorStore as_retriever raises ValueError | https://api.github.com/repos/langchain-ai/langchain/issues/10994/comments | 7 | 2023-09-25T00:31:06Z | 2024-07-08T13:40:48Z | https://github.com/langchain-ai/langchain/issues/10994 | 1,910,447,958 | 10,994 |
[
"langchain-ai",
"langchain"
] | ### Feature request
I used the Custom LLM Chat Model Agent and created my agent like this:
```
class CustomPromptTemplate(StringPromptTemplate):
template: str
tools: List[Tool]
def format(self, **kwargs) -> str:
# Get the intermediate steps (AgentAction, Observation tuples)
# Format them in a particular way
intermediate_steps = kwargs.pop("intermediate_steps")
thoughts = ""
for action, observation in intermediate_steps:
thoughts += action.log
thoughts += f"\nObservation: {observation}\nThought: "
# Set the agent_scratchpad variable to that value
kwargs["agent_scratchpad"] = thoughts
# Create a tools variable from the list of tools provided
kwargs["tools"] = "\n".join(
[f"{tool.name}: {tool.description}" for tool in self.tools])
# Create a list of tool names for the tools provided
kwargs["tool_names"] = ", ".join([tool.name for tool in self.tools])
return self.template.format(**kwargs)
```
The problem is this. This is very inefficient way to prompt things.
1) Better way to put tried actions.
2) Better way to put tools (maybe there some tools that you will eliminate after using one time.)
3) Overall better usage of the prompt. For example it cannot handle these kind of problems:
One of the steps, the agent chooses to use GetCurrentWeather but instead of that it takes Use GetCurrentWeather to get the current weather in Istanbul.' as the name of the tool.
```
input: "{'requested_tool_name': 'Use GetCurrentWeather to get the current weather in Istanbul.', 'available_tool_names': ['DuckDuckSearch', 'Search', 'Wikipedia', 'GetCurrentWeather', 'GetTickerName', 'GetTickerPrice']}"
```
the result:
```
Use GetCurrentWeather to get the current weather in Istanbul. is not a valid tool, try one of [DuckDuckSearch, Search, Wikipedia, GetCurrentWeather, GetTickerName, GetTickerPrice].
```
Is this possible to use directly openai function_calling like the openai agent ? Maybe that fixes the last problem.
### Motivation
The REACT way is too dirty and inefficient to use, it should be much more structured way to use CustomPromptTemplate and the CustomOutputParser
### Your contribution
If people comments on it, I can help with this. | Better usage of the Template Prompt: More Structure | https://api.github.com/repos/langchain-ai/langchain/issues/10992/comments | 2 | 2023-09-24T20:48:02Z | 2024-01-30T00:43:41Z | https://github.com/langchain-ai/langchain/issues/10992 | 1,910,374,299 | 10,992 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
When I run the following line:
db = Chroma.from_documents(documents=chunks, embedding=embeddings, persist_directory=persist_directory)
I get the error "You are using a deprecated configuration of Chroma."
I need to know what the new form of Chroma.from_documents() function is so that I can specify a persistent directory.
### Suggestion:
_No response_ | Error with db = Chroma.from_documents(documents=chunks, embedding=embeddings, persist_directory=persist_directory) | https://api.github.com/repos/langchain-ai/langchain/issues/10990/comments | 5 | 2023-09-24T19:40:20Z | 2023-11-09T05:19:35Z | https://github.com/langchain-ai/langchain/issues/10990 | 1,910,354,977 | 10,990 |
[
"langchain-ai",
"langchain"
] | ### Feature request
Add to_dict() function implementation to Memory classes such as BaseMemory, ConversationBufferMemory, etc.
### Motivation
Currently, we cannot use the JSON library to JSONIFY or Serialize the contents of the memory classes. This may be important if we want to persist the memory in DB for reference, logging, and tracking.
Doing `json.dumps(memory)` throws error:
`TypeError: Object of type ConversationBufferMemory is not JSON serializable`
### Your contribution
Adding the implementation for to_dict() function to memory classes | Add to_dict() implementation to Memory classes which are non-serializable | https://api.github.com/repos/langchain-ai/langchain/issues/10987/comments | 2 | 2023-09-24T15:25:52Z | 2024-02-06T16:29:22Z | https://github.com/langchain-ai/langchain/issues/10987 | 1,910,272,535 | 10,987 |
[
"langchain-ai",
"langchain"
] | null | Host the LLM model | https://api.github.com/repos/langchain-ai/langchain/issues/10986/comments | 1 | 2023-09-24T14:29:16Z | 2023-09-24T14:35:08Z | https://github.com/langchain-ai/langchain/issues/10986 | 1,910,254,862 | 10,986 |
[
"langchain-ai",
"langchain"
] | ### Issue with current documentation:
I tried to load LLama2-7b model from huggingface using HuggingFacePipeline.from_model_id but throws a value error: ValueError: The model has been loaded with `accelerate` and therefore cannot be moved to a specific device. Please discard the `device` argument when creating your pipeline object.
Here is my Code:
compute_dtype=getattr(torch,"float16")
bnb_config=BitsAndBytesConfig(
load_in_4bit=True,
bnb_4bit_quant_type="nf4",
bnb_4bit_compute_dtype=compute_dtype,
bnb_4bit_use_double_quant=True,
)
model_name="meta-llama/Llama-2-7b-hf"
llm = HuggingFacePipeline.from_model_id(
model_id=model_name,
task="text-generation",
pipeline_kwargs={
"return_full_text":True, # langchain expects the full text
"temperature":0.5, # 'randomness' of outputs, 0.0 is the min and 1.0 the max
"max_new_tokens":512, # max number of tokens to generate in the output
"repetition_penalty":1.1,
"return_dict_in_generate":True,
},
model_kwargs={
"quantization_config":bnb_config,
}
)
Logs:
WARNING:langchain.llms.huggingface_pipeline:Device has 1 GPUs available. Provide device={deviceId} to `from_model_id` to use availableGPUs for execution. deviceId is -1 (default) for CPU and can be a positive integer associated with CUDA device id.
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
[<ipython-input-9-14e8e7a13801>](https://localhost:8080/#) in <cell line: 1>()
----> 1 llm=prepare_model()
[/usr/local/lib/python3.10/dist-packages/transformers/pipelines/base.py](https://localhost:8080/#) in __init__(self, model, tokenizer, feature_extractor, image_processor, modelcard, framework, task, args_parser, device, torch_dtype, binary_output, **kwargs)
788
789 if hf_device_map is not None and device is not None:
--> 790 raise ValueError(
791 "The model has been loaded with `accelerate` and therefore cannot be moved to a specific device. Please "
792 "discard the `device` argument when creating your pipeline object."
ValueError: The model has been loaded with `accelerate` and therefore cannot be moved to a specific device. Please discard the `device` argument when creating your pipeline object.
Is it problem with Langchain?
### Idea or request for content:
_No response_ | DOC: HuggingFacePipeline.from_model_id throws Value error | https://api.github.com/repos/langchain-ai/langchain/issues/10985/comments | 6 | 2023-09-24T13:19:09Z | 2024-05-17T16:06:38Z | https://github.com/langchain-ai/langchain/issues/10985 | 1,910,232,099 | 10,985 |
[
"langchain-ai",
"langchain"
] | Hi. I'm using [Vercel's ai-chatbot template](https://github.com/vercel-labs/ai-chatbot) to prototype my application. I was wondering how I can convert the chat api call👇🏾 to use LangChain. I've been able to find examples that work, but I don't know how to have it so that I still use Vercel's KV database with LangChain.
Here's the code I would like to modify. Thanks.
```Javascript
// app/api/chat/route.ts
import { kv } from '@vercel/kv'
import { OpenAIStream, StreamingTextResponse } from 'ai'
import { Configuration, OpenAIApi } from 'openai-edge'
import { auth } from '@/auth'
import { nanoid } from '@/lib/utils'
export const runtime = 'edge'
const configuration = new Configuration({
apiKey: process.env.OPENAI_API_KEY
})
const openai = new OpenAIApi(configuration)
export async function POST(req: Request) {
const json = await req.json()
const { messages, previewToken } = json
const userId = (await auth())?.user.id
if (!userId) {
return new Response('Unauthorized', {
status: 401
})
}
if (previewToken) {
configuration.apiKey = previewToken
}
const res = await openai.createChatCompletion({
model: 'gpt-3.5-turbo',
messages,
temperature: 0.7,
stream: true
})
const stream = OpenAIStream(res, {
async onCompletion(completion) {
const title = json.messages[0].content.substring(0, 100)
const id = json.id ?? nanoid()
const createdAt = Date.now()
const path = `/chat/${id}`
const payload = {
id,
title,
userId,
createdAt,
path,
messages: [
...messages,
{
content: completion,
role: 'assistant'
}
]
}
await kv.hmset(`chat:${id}`, payload)
await kv.zadd(`user:chat:${userId}`, {
score: createdAt,
member: `chat:${id}`
})
}
})
return new StreamingTextResponse(stream)
}
``` | I want to use LangChain in Vercel's ai-chatbot template | https://api.github.com/repos/langchain-ai/langchain/issues/10982/comments | 4 | 2023-09-24T05:34:50Z | 2024-01-30T00:37:09Z | https://github.com/langchain-ai/langchain/issues/10982 | 1,910,108,891 | 10,982 |
[
"langchain-ai",
"langchain"
] | ### Issue you'd like to raise.
Hi , I'm using the SiteMap Loader as a document Loader , and I want to filter the lastmod metadata for the 2023 year only and nothing else, is it possible?
Here is my code:
`# fixes a bug with asyncio and jupyter
import nest_asyncio
nest_asyncio.apply()
from langchain.document_loaders.sitemap import SitemapLoader
loader = SitemapLoader(
"https:www.example.ca.sitemap.xml",
filter_urls=["https://www.example.ca/en/"]
)
docs = loader.load()`
### Suggestion:
_No response_ | Issue: I want to load SitemapLoader but only 2023 lastmod pages. | https://api.github.com/repos/langchain-ai/langchain/issues/10981/comments | 4 | 2023-09-24T03:16:22Z | 2024-01-30T00:43:48Z | https://github.com/langchain-ai/langchain/issues/10981 | 1,910,074,970 | 10,981 |
[
"langchain-ai",
"langchain"
] | ### System Info
Python==3.11.5
langchain==0.0.300
llama_cpp_python==0.2.6
chromadb==0.4.12
Running on Windows and on CPU
### Who can help?
@agola11
@hwchase17
### Information
- [ ] The official example notebooks/scripts
- [X] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [X] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [X] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
Embeddings can't be stored in a chroma/FAISS vectorstore using llama-cpp-python.
I found that embedding the contents of very very simple websites, such as "http://example.org" work fine.
Here is the code I am executing in my notebook:
```
from langchain.document_loaders import WebBaseLoader
from langchain.embeddings import LlamaCppEmbeddings
from langchain.vectorstores import Chroma
llama = LlamaCppEmbeddings(model_path="../models/openorca_stx.gguf")
loader = WebBaseLoader("https://www.bbc.com/")
pages = loader.load()
vectordb = Chroma.from_documents(
documents=pages,
embedding=llama,
persist_directory='../data/vectorstores/'
)
```
Here is the Traceback:
```
ValueError Traceback (most recent call last)
[p:\git_repos\langchain-test\src_langchain\04_vectorstore.ipynb](file:///P:/git_repos/langchain-test/src_langchain/04_vectorstore.ipynb) Cell 8 line 1
[7](vscode-notebook-cell:/p%3A/git_repos/langchain-test/src_langchain/04_vectorstore.ipynb#X12sZmlsZQ%3D%3D?line=6) loader = WebBaseLoader("https://www.bbc.com/")
[8](vscode-notebook-cell:/p%3A/git_repos/langchain-test/src_langchain/04_vectorstore.ipynb#X12sZmlsZQ%3D%3D?line=7) pages = loader.load()
---> [10](vscode-notebook-cell:/p%3A/git_repos/langchain-test/src_langchain/04_vectorstore.ipynb#X12sZmlsZQ%3D%3D?line=9) vectordb = Chroma.from_documents(
[11](vscode-notebook-cell:/p%3A/git_repos/langchain-test/src_langchain/04_vectorstore.ipynb#X12sZmlsZQ%3D%3D?line=10) documents=pages,
[12](vscode-notebook-cell:/p%3A/git_repos/langchain-test/src_langchain/04_vectorstore.ipynb#X12sZmlsZQ%3D%3D?line=11) embedding=llama,
[13](vscode-notebook-cell:/p%3A/git_repos/langchain-test/src_langchain/04_vectorstore.ipynb#X12sZmlsZQ%3D%3D?line=12) persist_directory='../data/vectorstores/'
[14](vscode-notebook-cell:/p%3A/git_repos/langchain-test/src_langchain/04_vectorstore.ipynb#X12sZmlsZQ%3D%3D?line=13) )
File [d:\miniconda3\envs\langchain\Lib\site-packages\langchain\vectorstores\chroma.py:646](file:///D:/miniconda3/envs/langchain/Lib/site-packages/langchain/vectorstores/chroma.py:646), in Chroma.from_documents(cls, documents, embedding, ids, collection_name, persist_directory, client_settings, client, collection_metadata, **kwargs)
644 texts = [doc.page_content for doc in documents]
645 metadatas = [doc.metadata for doc in documents]
--> 646 return cls.from_texts(
647 texts=texts,
648 embedding=embedding,
649 metadatas=metadatas,
650 ids=ids,
651 collection_name=collection_name,
652 persist_directory=persist_directory,
653 client_settings=client_settings,
654 client=client,
655 collection_metadata=collection_metadata,
656 **kwargs,
...
--> 510 self.input_ids[self.n_tokens : self.n_tokens + n_tokens] = batch
511 # Save logits
512 rows = n_tokens if self.params.logits_all else 1
ValueError: could not broadcast input array from shape (8,) into shape (0,)
```
### Expected behavior
I expect this function `
Chroma.from_documents(
documents=pages,
embedding=llama,
persist_directory='../data/vectorstores/'
)
` to create document embeddings and store them in the persistent chromadb | cant store embeddings from llama_cpp_python | https://api.github.com/repos/langchain-ai/langchain/issues/10978/comments | 3 | 2023-09-23T21:29:01Z | 2024-02-08T16:24:11Z | https://github.com/langchain-ai/langchain/issues/10978 | 1,910,006,050 | 10,978 |
[
"langchain-ai",
"langchain"
] | ### System Info
Name: langchain
Version: 0.0.300
Summary: Building applications with LLMs through composability
Home-page: https://github.com/langchain-ai/langchain
Author: None
Author-email: None
License: MIT
Location: /anaconda/envs/azureml_py38/lib/python3.8/site-packages
Requires: requests, jsonpatch, dataclasses-json, numpy, tenacity, numexpr, aiohttp, langsmith, anyio, SQLAlchemy, pydantic, async-timeout, PyYAML
Required-by: llama-index
### Who can help?
_No response_
### Information
- [ ] The official example notebooks/scripts
- [ ] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
```
from langchain.llms import HuggingFaceTextGenInference
llm = HuggingFaceTextGenInference(
inference_server_url="http://x.x.x.x:8080/",
max_new_tokens=512,
top_k=10,
top_p=0.95,
typical_p=0.95,
temperature=0.6,
# repetition_penalty=1.1,
)
from langchain.llms import OpenAI
from langchain.agents import initialize_agent, Tool
from langchain.agents import AgentType
from langchain.agents.react.base import DocstoreExplorer
docstore = DocstoreExplorer(vectordb)
tools = [
Tool(
name="Search",
func=docstore.search,
description="useful for when you need to ask with search",
),
Tool(
name="Lookup",
func=docstore.lookup,
description="useful for when you need to ask with lookup",
),
]
react = initialize_agent(tools, model, agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION, verbose=True)
react.run('Punishment for drinking and driving')
```
Error:
```
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
/home/azureuser/cloudfiles/code/Users/Tejasw/langchain_retreival.ipynb Cell 34 line 1
----> 1 react.run('Punishment for drinking and driving')
File /anaconda/envs/azureml_py38/lib/python3.8/site-packages/langchain/chains/base.py:487, in Chain.run(self, callbacks, tags, metadata, *args, **kwargs)
485 if len(args) != 1:
486 raise ValueError("`run` supports only one positional argument.")
--> 487 return self(args[0], callbacks=callbacks, tags=tags, metadata=metadata)[
488 _output_key
489 ]
491 if kwargs and not args:
492 return self(kwargs, callbacks=callbacks, tags=tags, metadata=metadata)[
493 _output_key
494 ]
File /anaconda/envs/azureml_py38/lib/python3.8/site-packages/langchain/chains/base.py:292, in Chain.__call__(self, inputs, return_only_outputs, callbacks, tags, metadata, run_name, include_run_info)
290 except BaseException as e:
291 run_manager.on_chain_error(e)
--> 292 raise e
293 run_manager.on_chain_end(outputs)
294 final_outputs: Dict[str, Any] = self.prep_outputs(
295 inputs, outputs, return_only_outputs
296 )
File /anaconda/envs/azureml_py38/lib/python3.8/site-packages/langchain/chains/base.py:286, in Chain.__call__(self, inputs, return_only_outputs, callbacks, tags, metadata, run_name, include_run_info)
279 run_manager = callback_manager.on_chain_start(
280 dumpd(self),
281 inputs,
282 name=run_name,
283 )
284 try:
285 outputs = (
--> 286 self._call(inputs, run_manager=run_manager)
287 if new_arg_supported
288 else self._call(inputs)
289 )
290 except BaseException as e:
291 run_manager.on_chain_error(e)
File /anaconda/envs/azureml_py38/lib/python3.8/site-packages/langchain/agents/agent.py:1122, in AgentExecutor._call(self, inputs, run_manager)
1120 # We now enter the agent loop (until it returns something).
1121 while self._should_continue(iterations, time_elapsed):
-> 1122 next_step_output = self._take_next_step(
1123 name_to_tool_map,
1124 color_mapping,
1125 inputs,
1126 intermediate_steps,
1127 run_manager=run_manager,
1128 )
1129 if isinstance(next_step_output, AgentFinish):
1130 return self._return(
1131 next_step_output, intermediate_steps, run_manager=run_manager
1132 )
File /anaconda/envs/azureml_py38/lib/python3.8/site-packages/langchain/agents/agent.py:977, in AgentExecutor._take_next_step(self, name_to_tool_map, color_mapping, inputs, intermediate_steps, run_manager)
975 tool_run_kwargs["llm_prefix"] = ""
976 # We then call the tool on the tool input to get an observation
--> 977 observation = tool.run(
978 agent_action.tool_input,
979 verbose=self.verbose,
980 color=color,
981 callbacks=run_manager.get_child() if run_manager else None,
982 **tool_run_kwargs,
983 )
984 else:
985 tool_run_kwargs = self.agent.tool_run_logging_kwargs()
File /anaconda/envs/azureml_py38/lib/python3.8/site-packages/langchain/tools/base.py:360, in BaseTool.run(self, tool_input, verbose, start_color, color, callbacks, tags, metadata, run_name, **kwargs)
358 except (Exception, KeyboardInterrupt) as e:
359 run_manager.on_tool_error(e)
--> 360 raise e
361 else:
362 run_manager.on_tool_end(
363 str(observation), color=color, name=self.name, **kwargs
364 )
File /anaconda/envs/azureml_py38/lib/python3.8/site-packages/langchain/tools/base.py:332, in BaseTool.run(self, tool_input, verbose, start_color, color, callbacks, tags, metadata, run_name, **kwargs)
329 try:
330 tool_args, tool_kwargs = self._to_args_and_kwargs(parsed_input)
331 observation = (
--> 332 self._run(*tool_args, run_manager=run_manager, **tool_kwargs)
333 if new_arg_supported
334 else self._run(*tool_args, **tool_kwargs)
335 )
336 except ToolException as e:
337 if not self.handle_tool_error:
File /anaconda/envs/azureml_py38/lib/python3.8/site-packages/langchain/tools/base.py:511, in Tool._run(self, run_manager, *args, **kwargs)
502 if self.func:
503 new_argument_supported = signature(self.func).parameters.get("callbacks")
504 return (
505 self.func(
506 *args,
507 callbacks=run_manager.get_child() if run_manager else None,
508 **kwargs,
509 )
510 if new_argument_supported
--> 511 else self.func(*args, **kwargs)
512 )
513 raise NotImplementedError("Tool does not support sync")
File /anaconda/envs/azureml_py38/lib/python3.8/site-packages/langchain/agents/react/base.py:77, in DocstoreExplorer.search(self, term)
75 def search(self, term: str) -> str:
76 """Search for a term in the docstore, and if found save."""
---> 77 result = self.docstore.search(term)
78 if isinstance(result, Document):
79 self.document = result
TypeError: search() missing 1 required positional argument: 'search_type'
```
### Expected behavior
Should successfully execute. | ReAct Agent doesn't execute | https://api.github.com/repos/langchain-ai/langchain/issues/10976/comments | 4 | 2023-09-23T17:12:11Z | 2023-12-30T16:04:58Z | https://github.com/langchain-ai/langchain/issues/10976 | 1,909,939,040 | 10,976 |
[
"langchain-ai",
"langchain"
] | ### Feature request
The current HuggingFaceEndpoint class supports several tasks like text generation and summarization but lacks support for text classification. This feature request aims to extend the class to handle text classification tasks.
### Motivation
I am working on a project and I find myself having to trying to use langchain but it seems to be missing a text-classifcation is missing. I think it will be a great add as it allows us to chain text-classification tasks as well.
### Your contribution
I will make changes to the HuggingFaceEndpoint to allow anyone else to make a chain with text-classification task involved. At the moment this will not be for batch classification. | Extend HuggingFaceEndpoint Class to Support Text Classification Endpoints | https://api.github.com/repos/langchain-ai/langchain/issues/10972/comments | 3 | 2023-09-23T11:59:43Z | 2023-12-30T16:05:03Z | https://github.com/langchain-ai/langchain/issues/10972 | 1,909,843,858 | 10,972 |
[
"langchain-ai",
"langchain"
] | ### System Info
Doing experiment on google collab
using llam2 7b quantized model
below is my code
agent with memory & i tried A gentExecutor
```
source_text_tool = Tool(
name="The Prophet Source Text QA System",
func=prophet_qa_chain.run,
description="Useful when asked question related to philosophy or The Prophet."
)
analysis_text_tool = Tool(
name="Other philosophy QA System",
func=analysis_qa_chain.run,
description="Useful when asked questions related to philosophy or Stoicism or Sikhs"
)
prefix = """
You're having a conversation with a human. You're helpful and answering
questions to your maximum ability.You have access to the following tools:
"""
suffix = """Let's Go!"
{chat_history}
Question: {input}
{agent_scratchpad}
"""
tools = [source_text_tool, analysis_text_tool]
prompt = ZeroShotAgent.create_prompt(tools,
prefix=prefix,
suffix=suffix,
input_variables=["input", "chat_history","agent_scratchpad"]) #{chat_history}
memory = ConversationBufferMemory(memory_key="chat_history")
llm_chain = LLMChain(llm=load_llm(),
prompt=prompt)
agent = ZeroShotAgent(llm_chain=llm_chain,
tools=tools,
verbose=True)
agent_chain_memory = AgentExecutor.from_agent_and_tools(
agent=agent,
tools=tools,
memory=memory, #its without memo
verbose=True
)
```
error i im getting
```
> Entering new AgentExecutor chain...
---------------------------------------------------------------------------
OutputParserException Traceback (most recent call last)
[<ipython-input-31-a44cd2d0a1a6>](https://localhost:8080/#) in <cell line: 1>()
----> 1 agent_chain_memory.run(input="How should I handle my fear of death ?")
7 frames
[/usr/local/lib/python3.10/dist-packages/langchain/agents/mrkl/output_parser.py](https://localhost:8080/#) in parse(self, text)
50
51 if not re.search(r"Action\s*\d*\s*:[\s]*(.*?)", text, re.DOTALL):
---> 52 raise OutputParserException(
53 f"Could not parse LLM output: `{text}`",
54 observation=MISSING_ACTION_AFTER_THOUGHT_ERROR_MESSAGE,
OutputParserException: Could not parse LLM output: `*Thought:* "I will take a deep breath and reflect on the nature of mortality, considering both the transience of life and the certainty of death."
*Action Input:* Choose one of the following options:
A) Seek guidance from The Prophet Muhammad's teachings on the afterlife.
B) Explore Stoicism or Sikh philosophy for coping mechanisms.
C) Utilize a QA system specifically designed to address questions related to philosophy and death.
*Observation:* The action taken will determine the final answer provided.`
```
### Who can help?
@hwchase17 @agola11
### Information
- [ ] The official example notebooks/scripts
- [X] My own modified scripts
### Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [X] Agents / Agent Executors
- [ ] Tools / Toolkits
- [ ] Chains
- [ ] Callbacks/Tracing
- [ ] Async
### Reproduction
```
# i used documents to store embedding in chromdadb & using below data
analysis_qa_chain = RetrievalQA.from_chain_type(
llm=load_llm(),
chain_type="map_reduce",
retriever=web_db.as_retriever()
)
prophet_qa_chain = RetrievalQA.from_chain_type(
llm=load_llm(),
chain_type="map_reduce",
retriever=prophet_db.as_retriever()
)
source_text_tool = Tool(
name="The Prophet Source Text QA System",
func=prophet_qa_chain.run,
description="Useful when asked question related to philosophy or The Prophet."
)
analysis_text_tool = Tool(
name="Other philosophy QA System",
func=analysis_qa_chain.run,
description="Useful when asked questions related to philosophy or Stoicism or Sikhs"
)
prefix = """
You're having a conversation with a human. You're helpful and answering
questions to your maximum ability.You have access to the following tools:
"""
suffix = """Let's Go!"
{chat_history}
Question: {input}
{agent_scratchpad}
"""
tools = [source_text_tool, analysis_text_tool]
prompt = ZeroShotAgent.create_prompt(tools,
prefix=prefix,
suffix=suffix,
input_variables=["input", "chat_history","agent_scratchpad"]) #{chat_history}
memory = ConversationBufferMemory(memory_key="chat_history")
llm_chain = LLMChain(llm=load_llm(),
prompt=prompt)
agent = ZeroShotAgent(llm_chain=llm_chain,
tools=tools,
verbose=True)
agent_chain_memory = AgentExecutor.from_agent_and_tools(
agent=agent,
tools=tools,
memory=memory, #its without memo
verbose=True
)
```
```
agent_chain_memory.run(input="How should I handle my fear of death ?")
```
### Expected behavior
im getting op but error also coming with this
```
> Entering new AgentExecutor chain...
---------------------------------------------------------------------------
OutputParserException Traceback (most recent call last)
[<ipython-input-31-a44cd2d0a1a6>](https://localhost:8080/#) in <cell line: 1>()
----> 1 agent_chain_memory.run(input="How should I handle my fear of death ?")
7 frames
[/usr/local/lib/python3.10/dist-packages/langchain/agents/mrkl/output_parser.py](https://localhost:8080/#) in parse(self, text)
50
51 if not re.search(r"Action\s*\d*\s*:[\s]*(.*?)", text, re.DOTALL):
---> 52 raise OutputParserException(
53 f"Could not parse LLM output: `{text}`",
54 observation=MISSING_ACTION_AFTER_THOUGHT_ERROR_MESSAGE,
OutputParserException: Could not parse LLM output: `*Thought:* "I will take a deep breath and reflect on the nature of mortality, considering both the transience of life and the certainty of death."
*Action Input:* Choose one of the following options:
A) Seek guidance from The Prophet Muhammad's teachings on the afterlife.
B) Explore Stoicism or Sikh philosophy for coping mechanisms.
C) Utilize a QA system specifically designed to address questions related to philosophy and death.
*Observation:* The action taken will determine the final answer provided.`
``` | OutputParserException: Could not parse LLM output | https://api.github.com/repos/langchain-ai/langchain/issues/10970/comments | 2 | 2023-09-23T05:24:56Z | 2023-12-30T16:05:09Z | https://github.com/langchain-ai/langchain/issues/10970 | 1,909,720,016 | 10,970 |
[
"langchain-ai",
"langchain"
] | ### Feature request
Currently in OpenSearch vector store while using the function `from_texts` it doesn't take `ids` as a parameter, so if multiple times the code is run for same documents the data gets duplicated. Which results in duplicated response being returned.
To fix this we just need to pass `ids` as optional parameter in the function. which should then be passed to add_embeddings function.
This is similar to what we are doing in add_texts function.
### Motivation
if multiple times the code is run for same documents the data gets duplicated. Which results in duplicated response being returned.
### Your contribution
@naveentatikonda who work on the integration can do this. | add support for passing of ids in from_texts in OpenSearch vector store | https://api.github.com/repos/langchain-ai/langchain/issues/10967/comments | 0 | 2023-09-22T22:44:13Z | 2023-10-10T16:29:37Z | https://github.com/langchain-ai/langchain/issues/10967 | 1,909,594,912 | 10,967 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.