Revert "conditional edge that checks for hallucinations (#401)"

backend/retrieval_graph/graph.py

@@ -11,7 +11,6 @@
 from langchain_core.messages import BaseMessage
 from langchain_core.runnables import RunnableConfig
 from langgraph.graph import END, START, StateGraph
-from pydantic import BaseModel, Field
 
 from backend.retrieval_graph.configuration import AgentConfiguration
 from backend.retrieval_graph.researcher_graph.graph import graph as researcher_graph
@@ -150,7 +149,6 @@ class Plan(TypedDict):
         "steps": response["steps"],
         "documents": "delete",
         "query": state.messages[-1].content,
-        "num_response_attempts": 0,
     }
 
 
@@ -209,57 +207,18 @@ async def respond(
     """
     configuration = AgentConfiguration.from_runnable_config(config)
     model = load_chat_model(configuration.response_model)
-    num_response_attempts = state.num_response_attempts
     # TODO: add a re-ranker here
     top_k = 20
     context = format_docs(state.documents[:top_k])
     prompt = configuration.response_system_prompt.format(context=context)
     messages = [{"role": "system", "content": prompt}] + state.messages
     response = await model.ainvoke(messages)
-    return {
-        "messages": [response],
-        "answer": response.content,
-        "num_response_attempts": num_response_attempts + 1,
-    }
-
-
-def check_hallucination(state: AgentState) -> Literal["respond", "end"]:
-    """Check if the answer is hallucinated."""
-    model = load_chat_model("openai/gpt-4o-mini")
-    top_k = 20
-    answer = state.answer
-    num_response_attempts = state.num_response_attempts
-    context = format_docs(state.documents[:top_k])
+    return {"messages": [response], "answer": response.content}
 
-    class GradeHallucinations(BaseModel):
-        """Binary score for hallucination present in generation answer."""
 
-        binary_score: str = Field(
-            description="Answer is grounded in the facts, 'yes' or 'no'"
-        )
-
-    grade_hallucinations_llm = model.with_structured_output(GradeHallucinations)
-    grade_hallucinations_system_prompt = """You are a grader assessing whether an LLM generation is grounded in / supported by a set of retrieved facts. \n 
-    Give a binary score 'yes' or 'no'. 'Yes' means that the answer is grounded in / supported by the set of facts."""
-    grade_hallucinations_prompt = (
-        "Set of facts: \n\n {context} \n\n LLM generation: {answer}"
-    )
-    grade_hallucinations_prompt_formatted = grade_hallucinations_prompt.format(
-        context=context, answer=answer
-    )
-    result = grade_hallucinations_llm.invoke(
-        [
-            {"role": "system", "content": grade_hallucinations_system_prompt},
-            {"role": "human", "content": grade_hallucinations_prompt_formatted},
-        ]
-    )
-    if result.binary_score == "yes" or num_response_attempts >= 2:
-        return "end"
-    else:
-        return "respond"
+# Define the graph
 
 
-# Define the graph
 builder = StateGraph(AgentState, input=InputState, config_schema=AgentConfiguration)
 builder.add_node(create_research_plan)
 builder.add_node(conduct_research)
@@ -268,9 +227,8 @@ class GradeHallucinations(BaseModel):
 builder.add_edge(START, "create_research_plan")
 builder.add_edge("create_research_plan", "conduct_research")
 builder.add_conditional_edges("conduct_research", check_finished)
-builder.add_conditional_edges(
-    "respond", check_hallucination, {"end": END, "respond": "respond"}
-)
+builder.add_edge("respond", END)
+
 # Compile into a graph object that you can invoke and deploy.
 graph = builder.compile()
 graph.name = "RetrievalGraph"

backend/retrieval_graph/state.py

@@ -80,8 +80,5 @@ class AgentState(InputState):
     documents: Annotated[list[Document], reduce_docs] = field(default_factory=list)
     """Populated by the retriever. This is a list of documents that the agent can reference."""
     answer: str = field(default="")
-    """Final answer. Useful for evaluations."""
+    """Final answer. Useful for evaluations"""
     query: str = field(default="")
-    """The user's query."""
-    num_response_attempts: int = field(default=0)
-    """The number of times the agent has tried to respond."""