Spaces:
Sleeping
Sleeping
| """Intent classification node for DocIntel.""" | |
| from app.agents.state import AgentState | |
| from app.config import settings | |
| import json | |
| try: | |
| from groq import Groq | |
| client = Groq(api_key=settings.GROQ_API_KEY) if settings.GROQ_API_KEY else None | |
| except ImportError: | |
| client = None | |
| def classify_intent(state: AgentState) -> AgentState: | |
| """Classify user intent""" | |
| print(f"📋 Classifying intent for: {state['query']}") | |
| # Simple keyword-based classification first (faster, no API call) | |
| query = state["query"].lower() | |
| # Check for summary intent | |
| summary_keywords = ["summarize", "summary", "brief", "overview", "summarise"] | |
| if any(word in query for word in summary_keywords): | |
| state["intent"] = "summary" | |
| print(f"✅ Intent: summary (keyword match)") | |
| return state | |
| # Check for email intent | |
| email_keywords = ["email", "draft", "send", "compose", "write an email"] | |
| if any(word in query for word in email_keywords): | |
| state["intent"] = "email" | |
| print(f"✅ Intent: email (keyword match)") | |
| return state | |
| # Check for extraction intent | |
| extract_keywords = [ | |
| "extract", "list", "find", "show me", | |
| "what are", "dates", "amounts", "numbers", | |
| "entities", "names", "people", "organizations" | |
| ] | |
| if any(word in query for word in extract_keywords): | |
| state["intent"] = "extract" | |
| print(f"✅ Intent: extract (keyword match)") | |
| return state | |
| # Use LLM fallback for better classification | |
| if client is not None: | |
| try: | |
| prompt = f""" | |
| Analyze this user query and classify it into one of these intents: | |
| 1. "qa" - General question about document content | |
| 2. "summary" - Request for document summary | |
| 3. "email" - Need to draft an email based on documents | |
| 4. "extract" - Need specific data extraction (dates, amounts, names, etc.) | |
| Query: {state["query"]} | |
| Return ONLY a JSON object with: | |
| {{ | |
| "intent": "category", | |
| "reasoning": "brief explanation" | |
| }} | |
| """ | |
| response = client.chat.completions.create( | |
| model=settings.GROQ_MODEL, | |
| messages=[{"role": "user", "content": prompt}], | |
| temperature=0.1, | |
| max_tokens=200 | |
| ) | |
| result = json.loads(response.choices[0].message.content) | |
| state["intent"] = result.get("intent", "qa") | |
| print(f"✅ Intent: {state['intent']} (LLM)") | |
| except Exception as e: | |
| print(f"⚠️ Classification error: {e}") | |
| state["intent"] = "qa" | |
| state["error"] = str(e) | |
| else: | |
| # Default to Q&A if no client | |
| state["intent"] = "qa" | |
| print(f"✅ Intent: qa (no LLM client)") | |
| return state |