Update app.py

app.py

@@ -123,30 +123,39 @@ class BasicAgent:
         # initialize HF inference pipeline once
         if HF_TOKEN is None:
             raise ValueError("HF_TOKEN not set in environment")
-        #pipe = pipeline("text-generation", model="EleutherAI/gpt-neo-125M", max_new_tokens=16)
-        #self.llm = HuggingFacePipeline(pipeline=pipe).bind_tools(TOOLS)
-        hf_pipe = HuggingFacePipeline.from_model_id(
-            model_id="EleutherAI/gpt-neo-125M",
-            task="text-generation",
-            pipeline_kwargs={"max_new_tokens":16},
+        pipe = pipeline("text-generation", model="EleutherAI/gpt-neo-125M", max_new_tokens=16)
+        self.llm = HuggingFacePipeline(pipeline=pipe) #.bind_tools(TOOLS)
+        #hf_pipe = HuggingFacePipeline.from_model_id(
+        #   model_id="EleutherAI/gpt-neo-125M",
+        #   task="text-generation",
+        #   pipeline_kwargs={"max_new_tokens":16},
+        #)
+        #chat = ChatHuggingFace(llm=hf_pipe)           # wrap in chat‐model
+        #self.llm = chat.bind_tools(TOOLS)                 # now this works :contentReference[oaicite:0]{index=0}
+
+        self.agent = initialize_agent(
+            tools=TOOLS,
+            llm=self.llm,
+            agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
+            system_message=SYSTEM_MESSAGE,
+            verbose=True,
         )
-        chat = ChatHuggingFace(llm=hf_pipe)           # wrap in chat‐model
-        self.llm = chat.bind_tools(TOOLS)                 # now this works :contentReference[oaicite:0]{index=0}
         
         # The GAIA system prompt (no "FINAL ANSWER:" at the end)
-        self.system_prompt = SYSTEM_MESSAGE
+        #self.system_prompt = SYSTEM_MESSAGE
         print("BasicAgent initialized with LLM.")
 
     # --- Core dispatcher/fallback ---
     def __call__(self, question: str) -> str:
-        prompt = f"{self.system_prompt}Q: {question}\nA:"
+        #prompt = f"{self.system_prompt}Q: {question}\nA:"
         #out = self.generator(prompt, max_new_tokens=16, return_full_text=False)
         #return out[0]["generated_text"].strip()
         # build a zero-shot-react-description agent for LLM+tools
-        agent_executor = initialize_agent(tools=TOOLS, llm=self.llm, agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION, verbose=True)
+        #agent_executor = initialize_agent(tools=TOOLS, llm=self.llm, agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION, verbose=True)
         # simply run the agent on the user’s question
-        answer = agent_executor.run(question)
-        return answer.strip()
+        #answer = agent_executor.run(question)
+        #return answer.strip()
+        return self.agent.run(question).strip()
         #agent = create_react_agent(llm=self.llm, tools=TOOLS, prompt=prompt)
         #agent = AgentExecutor(agent=agent, tools=TOOLS, verbose=True, return_intermediate_steps=False)
         #agent = AgentExecutor(agent=self.llm, tools=TOOLS, prompt=prompt, verbose=False, return_intermediate_steps=False)