switched to qwen

agent.py

@@ -18,7 +18,6 @@ from langchain.tools.retriever import create_retriever_tool
 from supabase.client import Client, create_client
 import re
 
-# === Load environment ===
 load_dotenv()
 
 # === Tools ===
@@ -72,7 +71,7 @@ with open("system_prompt.txt", "r", encoding="utf-8") as f:
     system_prompt = f.read()
 sys_msg = SystemMessage(content=system_prompt)
 
-# === Embedding & Vector DB ===
+# === Embeddings & Vector Store ===
 embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2")
 supabase: Client = create_client(os.getenv("SUPABASE_URL"), os.getenv("SUPABASE_SERVICE_KEY"))
 vector_store = SupabaseVectorStore(
@@ -85,31 +84,30 @@ vector_store = SupabaseVectorStore(
 # === Tools ===
 tools = [multiply, add, subtract, divide, modulus, wiki_search, web_search, arvix_search]
 
-# === LangGraph Agent Definition ===
-def build_graph(provider: str = "claude"):
-    if provider == "claude":
-        llm = ChatAnthropic(
-            model="claude-3-opus-20240229",
-            temperature=0,
-            anthropic_api_key=os.getenv("CLAUDE_API_KEY")
+# === LangGraph Builder ===
+def build_graph(provider: str = "huggingface"):
+    if provider == "huggingface":
+        llm = ChatHuggingFace(
+            llm=HuggingFaceEndpoint(
+                repo_id="Qwen/Qwen1.5-7B-Chat",
+                temperature=0,
+            )
         )
-    elif provider == "groq":
-        llm = ChatGroq(model="qwen-qwq-32b", temperature=0)
-    elif provider == "google":
-        llm = ChatGoogleGenerativeAI(model="gemini-2.0-flash", temperature=0)
-    elif provider == "huggingface":
-        llm = ChatHuggingFace(llm=HuggingFaceEndpoint(
-            url="https://api-inference.huggingface.co/models/Meta-DeepLearning/llama-2-7b-chat-hf",
-            temperature=0))
     else:
-        raise ValueError("Invalid provider")
+        raise ValueError("Only 'huggingface' (Qwen3) is supported in this build.")
 
     llm_with_tools = llm.bind_tools(tools)
 
     def retriever(state: MessagesState):
         query = state["messages"][-1].content
         similar = vector_store.similarity_search(query)
-        return {"messages": [sys_msg, state["messages"][-1], HumanMessage(content=f"Reference: {similar[0].page_content}")]}
+        return {
+            "messages": [
+                sys_msg,
+                state["messages"][-1],
+                HumanMessage(content=f"Reference: {similar[0].page_content}")
+            ]
+        }
 
     def assistant(state: MessagesState):
         response = llm_with_tools.invoke(state["messages"])
@@ -117,7 +115,6 @@ def build_graph(provider: str = "claude"):
 
     def formatter(state: MessagesState):
         last = state["messages"][-1].content.strip()
-
         cleaned = re.sub(r"<.*?>", "", last)
         cleaned = re.sub(r"(Final\s*Answer:|Answer:)", "", cleaned, flags=re.IGNORECASE)
         cleaned = cleaned.strip().split("\n")[0].strip()
@@ -128,6 +125,7 @@ def build_graph(provider: str = "claude"):
     builder.add_node("assistant", assistant)
     builder.add_node("tools", ToolNode(tools))
     builder.add_node("formatter", formatter)
+
     builder.add_edge(START, "retriever")
     builder.add_edge("retriever", "assistant")
     builder.add_conditional_edges("assistant", tools_condition)
@@ -136,9 +134,9 @@ def build_graph(provider: str = "claude"):
 
     return builder.compile()
 
-# === Test ===
+# === Run Test ===
 if __name__ == "__main__":
-    graph = build_graph("claude")
+    graph = build_graph()
     result = graph.invoke({"messages": [HumanMessage(content="What is the capital of France?")]})
     for m in result["messages"]:
         m.pretty_print()

app.py

@@ -13,8 +13,8 @@ cached_answers = []
 
 class ChatAgent:
     def __init__(self):
-        print("ChatAgent initialized with Claude LangGraph workflow.")
-        self.graph = build_graph("claude")
+        print("ChatAgent initialized with Qwen LangGraph workflow.")
+        self.graph = build_graph("huggingface")  # Uses Qwen endpoint
 
     def __call__(self, question: str) -> str:
         print(f"Processing question: {question[:60]}...")
@@ -59,7 +59,11 @@ def run_agent_only(profile: gr.OAuthProfile | None):
             cached_answers.append({"task_id": task_id, "submitted_answer": answer})
             results_log.append({"Task ID": task_id, "Question": question, "Submitted Answer": answer})
         except Exception as e:
-            results_log.append({"Task ID": task_id, "Question": question, "Submitted Answer": f"AGENT ERROR: {e}"})
+            results_log.append({
+                "Task ID": task_id,
+                "Question": question,
+                "Submitted Answer": f"AGENT ERROR: {e}"
+            })
 
     return "Agent finished. Now click 'Submit Cached Answers'", pd.DataFrame(results_log)
 
@@ -83,7 +87,8 @@ def submit_cached_answers(profile: gr.OAuthProfile | None):
         result = response.json()
         final_status = (
             f"Submission Successful!\nUser: {result.get('username')}\n"
-            f"Score: {result.get('score', 'N/A')}% ({result.get('correct_count', '?')}/{result.get('total_attempted', '?')})"
+            f"Score: {result.get('score', 'N/A')}% "
+            f"({result.get('correct_count', '?')}/{result.get('total_attempted', '?')})"
         )
         return final_status, None
     except Exception as e:

requirements.txt

@@ -11,9 +11,11 @@ langchain-tavily
 langchain-chroma
 langgraph
 huggingface_hub
+sentence-transformers
 supabase
 arxiv
 pymupdf
 wikipedia
 pgvector
 python-dotenv
+tqdm