Spaces:
Running
Running
File size: 3,728 Bytes
b09d5e9 |
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 99 100 101 102 103 104 105 106 107 108 |
from vector_rag import query_vector_store, llm # <--- FIX: Import llm here!
import wikipedia
from typing import List, Dict
# REMOVED: All duplicate model/pipeline/tokenizer imports and initialization code
# The 'llm' instance is now imported from vector_rag.py and is ready to use.
wikipedia.set_lang("en")
def format_conversation_context(history: List[Dict], max_messages: int = 10) -> str:
"""
Formats conversation history into a context string for the LLM.
Keeps only the most recent messages to prevent token overflow.
Args:
history: List of message dicts with 'role' and 'content' keys
max_messages: Maximum number of messages to include (default: 10)
Returns:
Formatted conversation history string
"""
if not history:
return ""
# Keep only the last N messages
recent_history = history[-max_messages:]
formatted_lines = []
for msg in recent_history:
role = "User" if msg["role"] == "user" else "Assistant"
formatted_lines.append(f"{role}: {msg['content']}")
return "\n".join(formatted_lines)
async def get_smart_rag_response(query: str, conversation_history: List[Dict] = None) -> tuple[str, str]:
"""
Get a smart RAG response with conversation context.
Args:
query: The user's current question
conversation_history: List of previous messages (optional)
Returns:
Tuple of (response, source)
"""
print(" Received Query:", query)
if conversation_history is None:
conversation_history = []
# Format conversation history for context
context_str = format_conversation_context(conversation_history)
# First: Try Wikipedia
try:
summary = wikipedia.summary(query, sentences=5)
print("Wikipedia summary found.")
# Build prompt with conversation context
prompt = f"""You are a helpful assistant engaged in a conversation.
"""
if context_str:
prompt += f"""
Previous conversation:
{context_str}
"""
prompt += f"""Use the following Wikipedia information to answer the current question as clearly as possible.
Wikipedia Context:
{summary}
Current question: {query}
Answer:"""
result = llm.invoke(prompt)
answer = result.replace(prompt, "").strip()
return answer, "Wikipedia"
except wikipedia.exceptions.PageError:
print("Wikipedia page not found.")
except wikipedia.exceptions.DisambiguationError as e:
return f"The query is ambiguous. Did you mean: {', '.join(e.options[:5])}", "Wikipedia"
# Second: Fallback to LLM with conversation context
try:
print("Fallback: LLM with conversation context")
fallback_prompt = "You are a knowledgeable assistant engaged in a conversation.\n\n"
if context_str:
fallback_prompt += f"Previous conversation:\n{context_str}\n\n"
fallback_prompt += f"Current question: {query}\nAnswer:"
llm_answer = llm.invoke(fallback_prompt)
answer = llm_answer.replace(fallback_prompt, "").strip()
if answer and "not sure" not in answer.lower():
return answer.strip(), "LLM"
except Exception as e:
print("Error during LLM fallback:", e)
# Finally: Fallback to Local Documents
try:
print("Fallback: Local vector search")
vector_answer = query_vector_store(query, conversation_history)
if vector_answer:
return vector_answer, "Local Document"
except Exception as e:
print("Error during local vector search:", e)
return "Sorry, I couldn't find any information to answer your question.", "System"
|