"""
Reasoner Agent (Rows -> Friendly Explanation)
Purpose
-------
Take the raw database rows returned by the Executor/Tools layer and generate a
short, human-friendly explanation.
Important note
--------------
This agent does NOT query the database. It only summarizes what the database
already returned.
Example
-------
Input: user_query="Fetch employees in AI"
execution_result=[(1, "Alice", ...), (2, "Bob", ...)]
Output: "I found 2 employees in the AI department: Alice and Bob..."
"""
import json
from agents.llm_provider import LLMProvider
from utils.logger import get_logger
from utils.serializer import json_dumps
logger = get_logger(__name__)
class ReasonerAgent:
"""Summarizes raw results into a short explanation."""
def __init__(self, llm_provider="ollama", model="mistral"):
self.llm = LLMProvider(provider=llm_provider, model=model)
def reason(self, user_query, execution_result):
"""
Use the LLM to produce a short explanation (3-4 sentences) based on:
- the original user query
- the raw data returned from the database
"""
logger.info(f"Reasoning about results...")
# Convert result to string if it's a dict/list
if isinstance(execution_result, (dict, list)):
result_str = json_dumps(execution_result, indent=2)
else:
result_str = str(execution_result)
prompt = f"""User asked: "{user_query}"
Database returned:
{result_str}
Provide a clear, concise explanation of:
1. What data was retrieved
2. Key insights or patterns
3. Any recommendations
Be conversational and helpful. Keep response to 3-4 sentences."""
try:
explanation = self.llm.generate(prompt)
logger.info(f"Reasoning complete")
return explanation
except Exception as e:
logger.error(f"Reasoning failed: {e}")
return f"Database returned {len(execution_result)} records" if isinstance(execution_result, list) else str(execution_result)
