agents/topic_analyzer.py

"""
LangChain LLMChain for analyzing and refining a research topic with yield support.
"""
from langchain.chains import LLMChain
from langchain.prompts import PromptTemplate
from langchain_openai import OpenAI
from langchain.callbacks.base import BaseCallbackHandler
import os
from typing import Dict, Any, Generator
from streaming_config import get_chunk_size, is_yield_enabled

# Prompt template for topic analysis
TOPIC_ANALYZER_PROMPT = """
You are an expert research assistant. Given a user-provided research topic, refine it for clarity, focus, and academic rigor. 

Original topic: {topic}

Instructions:
- Refine the topic into a clear, focused research question
- Keep the response concise and academic
- Make sure response should be generated in 500 characters including spaces.

Refined topic (one sentence):
"""

def get_topic_analyzer_chain():
    """
    Returns a LangChain LLMChain for topic analysis.
    """
    prompt = PromptTemplate(
        input_variables=["topic"],
        template=TOPIC_ANALYZER_PROMPT
    )
    
    # Always enable streaming in the LLM - callbacks will be passed during execution
    llm = OpenAI(
        temperature=0.3,
        openai_api_key=os.getenv("OPENAI_API_KEY"),
        streaming=True  # Always enable streaming
    )
    
    return LLMChain(llm=llm, prompt=prompt, output_key="refined_topic")

def yield_topic_analysis(topic: str, preset: str = None) -> Generator[str, None, None]:
    """
    Yield topic analysis results progressively
    
    Args:
        topic: The research topic to analyze
        preset: Optional streaming preset
        
    Yields:
        str: Progressive analysis results
    """
    if not is_yield_enabled(preset):
        # Fallback to non-yield processing
        chain = get_topic_analyzer_chain()
        result = chain.run(topic=topic)
        yield result
        return
    
    try:
        # Import required modules
        from langchain_openai import ChatOpenAI
        from langchain.prompts import PromptTemplate
        import os
        
        # Create prompt template
        prompt = PromptTemplate(
            input_variables=["topic"],
            template=TOPIC_ANALYZER_PROMPT
        )
        
        # Create LLM with streaming
        llm = ChatOpenAI(
            temperature=0.3,
            openai_api_key=os.getenv("OPENAI_API_KEY"),
            streaming=True
        )
        
        # Format the prompt
        formatted_prompt = prompt.format(topic=topic)
        
        # Get chunk size for this step
        chunk_size = get_chunk_size("topic_analyzer", preset)
        
        # Call LLM and yield results progressively
        result = llm.invoke(formatted_prompt)
        content = result.content
        
        # Yield content in chunks
        for i in range(0, len(content), chunk_size):
            chunk = content[i:i + chunk_size]
            yield chunk
            
    except Exception as e:
        yield f"Error in topic analysis: {str(e)}"

def process_topic_with_feedback(topic: str, feedback: str = None) -> Generator[str, None, None]:
    """
    Process topic analysis with optional feedback using yield generators
    
    Args:
        topic: The research topic
        feedback: Optional feedback for refinement
        
    Yields:
        str: Progressive processing results
    """
    if feedback:
        # Include feedback in processing
        enhanced_prompt = f"""
        Original topic: {topic}
        Feedback: {feedback}
        
        Please refine the topic considering the feedback provided.
        """
        yield "Processing topic with feedback..."
    else:
        enhanced_prompt = f"Original topic: {topic}"
        yield "Processing topic analysis..."
    
    # Yield the enhanced prompt for processing
    for chunk in yield_topic_analysis(enhanced_prompt):
        yield chunk