File size: 4,468 Bytes
24ae72d
 
 
 
 
 
 
 
 
 
 
 
 
 
b9ccd0b
24ae72d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b9ccd0b
24ae72d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b9ccd0b
24ae72d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
# /home/bk_anupam/code/LLM_agents/RAG_BOT/agent/evaluation_nodes.py
import datetime
import os
import sys

from langchain_google_genai import ChatGoogleGenerativeAI
from langchain_core.messages import AIMessage

# Add the project root to the Python path
project_root = os.path.abspath(os.path.join(os.path.dirname(__file__), '..', '..'))
sys.path.insert(0, project_root)

from RAG_BOT.logger import logger
from RAG_BOT.agent.state import AgentState
from RAG_BOT.agent.prompts import get_evaluate_context_chat_prompt, get_reframe_question_chat_prompt


def evaluate_context_node(state: AgentState, llm: ChatGoogleGenerativeAI):
    """
    Evaluates the reranked context (read from state) based on the original query.
    """
    logger.info("--- Executing Evaluate Context Node ---")
    original_query = state.get('original_query')
    # Context is now expected to be populated by the rerank_context_node
    context_to_evaluate = state.get('context')

    # Check if context exists in the state
    if context_to_evaluate is None: # Check for None explicitly
         logger.warning("Evaluate context node reached but no context found in state (potentially due to retrieval/reranking failure). Treating as insufficient.")
         return {"evaluation_result": "insufficient"} # Context remains None

    if not original_query:
        logger.warning("Missing original query for evaluation.")
        # Context is already stored, just return insufficient
        return {"evaluation_result": "insufficient"}

    # Handle empty string context (e.g., if reranker returned no docs)
    if not context_to_evaluate:
        logger.info("Context in state is empty. Treating as insufficient.")
        return {"evaluation_result": "insufficient"}

    # Check for error messages passed as context (though reranker should return None on error)
    if isinstance(context_to_evaluate, str) and context_to_evaluate.startswith("Error:"):
        logger.warning(f"Context contains an error message: {context_to_evaluate}. Treating as insufficient.")
        return {"evaluation_result": "insufficient"}


    logger.info("Evaluating context from state...")
    eval_chain = get_evaluate_context_chat_prompt() | llm
    evaluation_result_str = eval_chain.invoke({
        "original_query": original_query,
        "context": context_to_evaluate # Use context from state
    }).content.strip().upper()

    logger.info(f"Context evaluation result: {evaluation_result_str}")
    evaluation = "sufficient" if evaluation_result_str == "YES" else "insufficient"

    # Context is already in state, just return the evaluation result
    return {"evaluation_result": evaluation}


def reframe_query_node(state: AgentState, llm: ChatGoogleGenerativeAI):
    """
    Reframes the query if the first retrieval attempt was insufficient.
    """
    logger.info("--- Executing Reframe Query Node ---")
    original_query = state['original_query']
    # The query used in the failed attempt
    failed_query = state['current_query']

    if not original_query or not failed_query:
        logger.error("Missing original or current query for reframing.")
        # Handle error state - perhaps skip reframing and go to final "cannot find"
        # For now, just update state to prevent infinite loops
        return {"retry_attempted": True, "current_query": original_query} # Fallback

    logger.info("Reframing question...")
    reframe_chain = get_reframe_question_chat_prompt() | llm
    reframed_question = reframe_chain.invoke({
        "original_query": original_query,
        "failed_query": failed_query
    }).content.strip()

    logger.info(f"Reframed question: {reframed_question}")

    # Create an AIMessage to trigger the tool call again
    tool_call_id = f"call_retry_{datetime.datetime.now().isoformat()}"
    retry_tool_call_message = AIMessage(
        content=f"Reframing query to: {reframed_question}", # Optional content for logging/tracing
        tool_calls=[{
            "name": "retrieve_context",
            "args": {"query": reframed_question},
            "id": tool_call_id
        }]
    )

    # Return state updates AND the message to trigger the tool
    return {
        "messages": [retry_tool_call_message], # Add this message to the state
        "current_query": reframed_question,
        "retry_attempted": True,
        "evaluation_result": None, # Reset evaluation for the next attempt
        "context": None # Clear context before retry
    }