Upload 5 files

agent.py

@@ -1,23 +1,31 @@
 """LangGraph Agent"""
 import os
+import re
 from dotenv import load_dotenv
 from langgraph.graph import START, StateGraph, MessagesState
 from langgraph.prebuilt import tools_condition
 from langgraph.prebuilt import ToolNode
 from langchain_google_genai import ChatGoogleGenerativeAI
 from langchain_groq import ChatGroq
-from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint, HuggingFaceEmbeddings
-from langchain_community.tools.tavily_search import TavilySearchResults
+from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint
+from langchain_community.tools import DuckDuckGoSearchRun
 from langchain_community.document_loaders import WikipediaLoader
 from langchain_community.document_loaders import ArxivLoader
-from langchain_community.vectorstores import SupabaseVectorStore
-from langchain_core.messages import SystemMessage, HumanMessage
+from langchain_core.messages import SystemMessage
 from langchain_core.tools import tool
-from langchain.tools.retriever import create_retriever_tool
-from supabase.client import Client, create_client
 
 load_dotenv()
 
+_FINAL_ANSWER_RE = re.compile(r"final answer\s*:\s*", re.IGNORECASE)
+
+
+def extract_final_answer(text: str) -> str:
+    """Strip the `FINAL ANSWER:` prefix the system prompt asks the model for."""
+    match = _FINAL_ANSWER_RE.search(text)
+    answer = text[match.end():] if match else text
+    return answer.strip()
+
+
 @tool
 def multiply(a: int, b: int) -> int:
     """Multiply two numbers.
@@ -30,7 +38,7 @@ def multiply(a: int, b: int) -> int:
 @tool
 def add(a: int, b: int) -> int:
     """Add two numbers.
-    
+
     Args:
         a: first int
         b: second int
@@ -40,7 +48,7 @@ def add(a: int, b: int) -> int:
 @tool
 def subtract(a: int, b: int) -> int:
     """Subtract two numbers.
-    
+
     Args:
         a: first int
         b: second int
@@ -48,9 +56,9 @@ def subtract(a: int, b: int) -> int:
     return a - b
 
 @tool
-def divide(a: int, b: int) -> int:
+def divide(a: int, b: int) -> float:
     """Divide two numbers.
-    
+
     Args:
         a: first int
         b: second int
@@ -62,7 +70,7 @@ def divide(a: int, b: int) -> int:
 @tool
 def modulus(a: int, b: int) -> int:
     """Get the modulus of two numbers.
-    
+
     Args:
         a: first int
         b: second int
@@ -72,45 +80,34 @@ def modulus(a: int, b: int) -> int:
 @tool
 def wiki_search(query: str) -> str:
     """Search Wikipedia for a query and return maximum 2 results.
-    
+
     Args:
         query: The search query."""
     search_docs = WikipediaLoader(query=query, load_max_docs=2).load()
-    formatted_search_docs = "\n\n---\n\n".join(
-        [
-            f'<Document source="{doc.metadata["source"]}" page="{doc.metadata.get("page", "")}"/>\n{doc.page_content}\n</Document>'
-            for doc in search_docs
-        ])
-    return {"wiki_results": formatted_search_docs}
+    return "\n\n---\n\n".join(
+        f'<Document source="{doc.metadata.get("source", "")}"/>\n{doc.page_content}\n</Document>'
+        for doc in search_docs
+    )
 
 @tool
 def web_search(query: str) -> str:
-    """Search Tavily for a query and return maximum 3 results.
-    
+    """Search the web (DuckDuckGo) for a query and return a few results.
+
     Args:
         query: The search query."""
-    search_docs = TavilySearchResults(max_results=3).invoke(query=query)
-    formatted_search_docs = "\n\n---\n\n".join(
-        [
-            f'<Document source="{doc.metadata["source"]}" page="{doc.metadata.get("page", "")}"/>\n{doc.page_content}\n</Document>'
-            for doc in search_docs
-        ])
-    return {"web_results": formatted_search_docs}
+    return DuckDuckGoSearchRun().invoke(query)
 
 @tool
 def arvix_search(query: str) -> str:
     """Search Arxiv for a query and return maximum 3 result.
-    
+
     Args:
         query: The search query."""
     search_docs = ArxivLoader(query=query, load_max_docs=3).load()
-    formatted_search_docs = "\n\n---\n\n".join(
-        [
-            f'<Document source="{doc.metadata["source"]}" page="{doc.metadata.get("page", "")}"/>\n{doc.page_content[:1000]}\n</Document>'
-            for doc in search_docs
-        ])
-    return {"arvix_results": formatted_search_docs}
-
+    return "\n\n---\n\n".join(
+        f'<Document source="{doc.metadata.get("source", "")}"/>\n{doc.page_content[:1000]}\n</Document>'
+        for doc in search_docs
+    )
 
 
 # load the system prompt from the file
@@ -120,25 +117,6 @@ with open("system_prompt.txt", "r", encoding="utf-8") as f:
 # System message
 sys_msg = SystemMessage(content=system_prompt)
 
-# build a retriever
-embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2") #  dim=768
-supabase: Client = create_client(
-    os.environ.get("SUPABASE_URL"), 
-    os.environ.get("SUPABASE_SERVICE_KEY"))
-vector_store = SupabaseVectorStore(
-    client=supabase,
-    embedding= embeddings,
-    table_name="documents",
-    query_name="match_documents_langchain",
-)
-create_retriever_tool = create_retriever_tool(
-    retriever=vector_store.as_retriever(),
-    name="Question Search",
-    description="A tool to retrieve similar questions from a vector store.",
-)
-
-
-
 tools = [
     multiply,
     add,
@@ -153,19 +131,23 @@ tools = [
 # Build graph function
 def build_graph(provider: str = "google"):
     """Build the graph"""
-    # Load environment variables from .env file
     if provider == "google":
-        # Google Gemini
+        # Google Gemini - needs GOOGLE_API_KEY
         llm = ChatGoogleGenerativeAI(model="gemini-2.0-flash", temperature=0)
     elif provider == "groq":
-        # Groq https://console.groq.com/docs/models
-        llm = ChatGroq(model="qwen-qwq-32b", temperature=0) # optional : qwen-qwq-32b gemma2-9b-it
+        # Groq https://console.groq.com/docs/models - needs GROQ_API_KEY
+        llm = ChatGroq(model="llama-3.3-70b-versatile", temperature=0)
     elif provider == "huggingface":
-        # TODO: Add huggingface endpoint
+        # HF Inference Providers - needs HF_TOKEN. "auto" provider selection can land on a
+        # backend that doesn't support tool calling, so pin one that does (override via
+        # HF_AGENT_PROVIDER if it doesn't happen to host HF_AGENT_MODEL).
         llm = ChatHuggingFace(
             llm=HuggingFaceEndpoint(
-                url="https://api-inference.huggingface.co/models/Meta-DeepLearning/llama-2-7b-chat-hf",
+                repo_id=os.getenv("HF_AGENT_MODEL", "meta-llama/Llama-3.1-8B-Instruct"),
+                provider=os.getenv("HF_AGENT_PROVIDER", "deepinfra"),
+                huggingfacehub_api_token=os.getenv("HF_TOKEN"),
                 temperature=0,
+                max_new_tokens=1024,
             ),
         )
     else:
@@ -177,47 +159,13 @@ def build_graph(provider: str = "google"):
     def assistant(state: MessagesState):
         """Assistant node"""
         return {"messages": [llm_with_tools.invoke(state["messages"])]}
-    
-    # def retriever(state: MessagesState):
-      #  """Retriever node"""
-       # similar_question = vector_store.similarity_search(state["messages"][0].content)
-        #example_msg = HumanMessage(
-         #   content=f"Here I provide a similar question and answer for reference: \n\n{similar_question[0].page_content}",
-       # )
-       # return {"messages": [sys_msg] + state["messages"] + [example_msg]}
-
-    from langchain_core.messages import AIMessage
-
-    def retriever(state: MessagesState):
-        query = state["messages"][-1].content
-        similar_doc = vector_store.similarity_search(query, k=1)[0]
-
-        content = similar_doc.page_content
-        if "Final answer :" in content:
-            answer = content.split("Final answer :")[-1].strip()
-        else:
-            answer = content.strip()
-
-        return {"messages": [AIMessage(content=answer)]}
-
-   # builder = StateGraph(MessagesState)
-    #builder.add_node("retriever", retriever)
-    #builder.add_node("assistant", assistant)
-    #builder.add_node("tools", ToolNode(tools))
-    #builder.add_edge(START, "retriever")
-    #builder.add_edge("retriever", "assistant")
-    #builder.add_conditional_edges(
-     #   "assistant",
-      #  tools_condition,
-    #)
-    #builder.add_edge("tools", "assistant")
 
     builder = StateGraph(MessagesState)
-    builder.add_node("retriever", retriever)
-
-    # Retriever ist Start und Endpunkt
-    builder.set_entry_point("retriever")
-    builder.set_finish_point("retriever")
+    builder.add_node("assistant", assistant)
+    builder.add_node("tools", ToolNode(tools))
+    builder.add_edge(START, "assistant")
+    builder.add_conditional_edges("assistant", tools_condition)
+    builder.add_edge("tools", "assistant")
 
     # Compile graph
     return builder.compile()

app.py

@@ -5,7 +5,7 @@ import gradio as gr
 import requests
 import pandas as pd
 from langchain_core.messages import HumanMessage
-from agent import build_graph
+from agent import build_graph, sys_msg, extract_final_answer
 
 
 
@@ -25,10 +25,10 @@ class BasicAgent:
 
     def __call__(self, question: str) -> str:
         print(f"Agent received question (first 50 chars): {question[:50]}...")
-        messages = [HumanMessage(content=question)]
+        messages = [sys_msg, HumanMessage(content=question)]
         result = self.graph.invoke({"messages": messages})
-        answer = result['messages'][-1].content
-        return answer  # kein [14:] mehr nötig!
+        answer = extract_final_answer(result['messages'][-1].content)
+        return answer
 
 
 

requirements.txt

@@ -1,18 +1,14 @@
 gradio
 requests
-langchain
+pandas
 langchain-community
 langchain-core
 langchain-google-genai
 langchain-huggingface
 langchain-groq
-langchain-tavily
-langchain-chroma
 langgraph
 huggingface_hub
-supabase
+duckduckgo-search
 arxiv
-pymupdf
 wikipedia
-pgvector
 python-dotenv

system_prompt.txt

@@ -5,9 +5,7 @@ FINAL ANSWER: [ANSWER]
 
 Only write the answer in that exact format. Do not explain anything. Do not include any other text.
 
-If you are provided with a similar question and its final answer, and the current question is **exactly the same**, then simply return the same final answer without using any tools.
-
-Only use tools if the current question is different from the similar one.
+Use the available tools whenever you need external information or computation to answer.
 
 Examples:
 - FINAL ANSWER: FunkMonk