Update space

README.md

@@ -1,5 +1,5 @@
 ---
-title: Ppo Test Space
+title: Tuned Llama Ai Interviewer
 emoji: 💬
 colorFrom: yellow
 colorTo: purple

app.py

@@ -1,52 +1,102 @@
-import gradio as gr
-from huggingface_hub import InferenceClient
+import gradio as gr 
+from typing import List, Union, Dict, Tuple
+from transformers import pipeline 
+from os import getenv
+from huggingface_hub import login 
+
+from history import get_history, update_history
+
+# Login to Hugging Face
+login(getenv("Token"))
+ 
+#name of model on huggingFace
+model="ikenna1234/EleutherAI_pythia_1b_rlhf"
+  
+# Define generator pipeline
+generator = pipeline("text-generation", model=model)
+ 
+
+#Transform gradio history by breaking any tuple into 2 dicts
+def transform_gradio_history(history: List[Union[Dict[str, str], Tuple[str, str]]]) -> List[Dict[str, str]]:
+    transformed_history = []
+    
+    for entry in history:
+        if (isinstance(entry, list) or isinstance(entry, tuple)) and len(entry) == 2:
+            transformed_history.append({"role": "user", "content": entry[0]})
+            transformed_history.append({"role": "assistant", "content": entry[1]})
+        elif isinstance(entry, dict):
+            transformed_history.append(entry)
+    
+    return transformed_history
+
+
+#Does the actual inference and streams (yield) the response
+def chat(history:list[dict[str, str]]):
+    for msg in generator(
+        history, #message list
+        max_new_tokens=10048,
+        return_full_text=False
+    ): 
+        yield msg['generated_text']
 
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
 
 
 def respond(
     message,
-    history: list[tuple[str, str]],
-    system_message,
+    history: list[dict[str, str]],
+    system_message, #system prompt
     max_tokens,
     temperature,
     top_p,
+    group_name #user Id
 ):
-    messages = [{"role": "system", "content": system_message}]
+    if not group_name:
+        #user must pass user Id to the group_name.
+        #This is used to identify the user
+        yield "User ID required"
+    else:
+        messages=history
+           
+        #If no history, get history from database 
+        if not len(messages):
+            messages=get_history(group_name) 
+        
+        #Break any tuples into 2 dicts
+        messages=transform_gradio_history(messages)
+        
+        #Add prompt to list of messages
+        messages.append({"role": "user", "content": message}) 
 
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
+        response = ""
 
-    messages.append({"role": "user", "content": message})
+        #Create new list of all messages, starting with system prompt
+        mainMessage=[{"role": "system", "content": system_message}, *messages] 
 
-    response = ""
+        #calls the inference function and streams the response
+        for msg in chat(mainMessage):
+            token = msg
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
+            # This is a stream. Meaning response comes in bits of string.
+            # Add new response string bit to previous response 
+            # strings to form the whole string
+            response += token 
+            yield response
 
-        response += token
-        yield response
+        #update the history in database
+        if response:
+            messages.append({"role": "assistant", "content": response})
+            update_history(group_name,messages)
 
+def initialize():
+  messages=[] 
+  return messages
 
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
 demo = gr.ChatInterface(
     respond,
+    type="messages",
+    chatbot=gr.Chatbot(value=initialize(),type="messages"),
     additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
+        gr.Textbox(value="You are an AI assistant that conducts interview", label="System message"),
         gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
         gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
         gr.Slider(
@@ -55,10 +105,11 @@ demo = gr.ChatInterface(
             value=0.95,
             step=0.05,
             label="Top-p (nucleus sampling)",
-        ),
+        ), 
+        gr.Textbox( label="User ID"),
     ],
 )
 
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch(share=True,ssr_mode=False)

history.py

@@ -0,0 +1,26 @@
+from mongo_client import get_client, get_collection
+
+client=get_client()
+collection=get_collection("history",client)
+
+def get_history(group_name:str): 
+    history= collection.find_one({"group_name":group_name})
+    
+    value=[]
+
+    if history and 'value' in history:
+        value=history['value']
+     
+
+    return value
+
+
+def update_history(group_name:str,value): 
+    query_filter = {"group_name":group_name}
+
+    update_operation = { "$set" :{ "value" : value }}
+
+    collection.update_one(query_filter, update_operation, upsert=True)
+ 
+
+    return True

mongo_client.py

@@ -0,0 +1,15 @@
+from pymongo import MongoClient, server_api
+
+from os import getenv
+ 
+database_name=getenv("DATABASE_NAME")
+
+def get_client():
+    client = MongoClient(getenv("MONGODB_CONNECTION_STRING"), server_api=server_api.ServerApi(
+        version="1", strict=True, deprecation_errors=True))
+    
+    return client
+
+def get_collection(collection_name:str,client:MongoClient):
+    database=client[database_name]
+    return database[collection_name]

requirements.txt

@@ -1 +1,5 @@
-huggingface_hub==0.25.2
+huggingface_hub 
+transformers>=4.48
+pymongo 
+torch  
+accelerate  

works.py

@@ -0,0 +1,108 @@
+import gradio as gr
+from huggingface_hub import InferenceClient 
+from typing import List, Union, Dict, Tuple
+from transformers import pipeline
+
+from os import getenv
+from huggingface_hub import login 
+
+from history import get_history, update_history
+
+# Login to Hugging Face
+login(getenv("Token"))
+
+#client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+
+client = InferenceClient("meta-llama/Llama-3.2-3B-Instruct")
+
+generator = pipeline("text-generation", model="ikenna1234/ai_interviewer_2")
+ 
+
+def transform_gradio_history(history: List[Union[Dict[str, str], Tuple[str, str]]]) -> List[Dict[str, str]]:
+    transformed_history = []
+    
+    for entry in history:
+        if (isinstance(entry, list) or isinstance(entry, tuple)) and len(entry) == 2:
+            transformed_history.append({"role": "user", "content": entry[0]})
+            transformed_history.append({"role": "assistant", "content": entry[1]})
+        elif isinstance(entry, dict):
+            transformed_history.append(entry)
+    
+    return transformed_history
+
+
+def chat(history:list[dict[str, str]]):
+    for msg in generator(
+        history,
+        max_new_tokens=10048,
+        return_full_text=False
+        ):
+        yield msg[0] 
+
+
+
+def respond(
+    message,
+    history: list[dict[str, str]],
+    system_message,
+    max_tokens,
+    temperature,
+    top_p,
+    group_name
+):
+    if not group_name:
+        yield "User ID required"
+    else:
+        messages=history
+           
+        #If no history, get history from database 
+        if not len(messages):
+            messages=get_history(group_name)
+            #messages=old_history
+        
+        #Break any tuples into 2 dicts
+        messages=transform_gradio_history(messages)
+        
+        messages.append({"role": "user", "content": message}) 
+
+        response = ""
+
+        mainMessage=[{"role": "system", "content": system_message}, *messages] 
+
+        for msg in chat(mainMessage):
+            token = msg
+
+            response += token
+            yield response
+
+        #update the history in database
+        if response:
+            messages.append({"role": "assistant", "content": response})
+            update_history(group_name,messages)
+
+def initialize():
+  messages=[] 
+  return messages
+
+demo = gr.ChatInterface(
+    respond,
+    type="messages",
+    chatbot=gr.Chatbot(value=initialize(),type="messages"),
+    additional_inputs=[
+        gr.Textbox(value="You are an AI assistant that conducts interview", label="System message"),
+        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
+        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
+        gr.Slider(
+            minimum=0.1,
+            maximum=1.0,
+            value=0.95,
+            step=0.05,
+            label="Top-p (nucleus sampling)",
+        ), 
+        gr.Textbox( label="User ID"),
+    ],
+)
+
+
+if __name__ == "__main__":
+    demo.launch(share=True,ssr_mode=False)