Spaces:
Running
Running
| # Enterprise Agent Router v3.0 - Simple & Clean | |
| """ | |
| Simple routing system for business analytics queries. | |
| Supports 5 modes: analyst, deep, vision, predict, agent | |
| NO DEPENDENCIES on deleted files! | |
| """ | |
| from typing import Tuple | |
| # ============================================================================ | |
| # KEYWORD-BASED ROUTING (Simple & Reliable) | |
| # ============================================================================ | |
| RAG_KEYWORDS = [ | |
| "what is", "what are", "show me", "tell me", | |
| "list all", "display", "from file", "from document", | |
| "total", "number of", "how many", "summarize" | |
| ] | |
| GRAPH_KEYWORDS = [ | |
| "why", "how did", "what caused", "trend", "pattern", | |
| "correlation", "relationship", "connection", "insight", | |
| "best", "worst", "top", "bottom", "customer", "product" | |
| ] | |
| PREDICT_KEYWORDS = [ | |
| "predict", "forecast", "future", "next month", "next year", | |
| "projection", "estimate", "what if", "scenario", "expect" | |
| ] | |
| VISION_KEYWORDS = [ | |
| "image", "picture", "photo", "chart", "graph", | |
| "diagram", "invoice", "receipt", "screenshot", "ocr", "scan" | |
| ] | |
| AGENT_KEYWORDS = [ | |
| "search", "find online", "web", "latest", "news", | |
| "industry", "benchmark", "competitor", "market", "external" | |
| ] | |
| def route_question( | |
| question: str, | |
| has_image: bool = False, | |
| mode: str = "auto" | |
| ) -> str: | |
| """ | |
| Simple routing for business analytics queries. | |
| Args: | |
| question: User's business question | |
| has_image: Whether query includes image/file | |
| mode: Force specific mode ("auto" or mode name) | |
| Returns: | |
| Route name: "analyst", "deep", "vision", "predict", "agent", | |
| or legacy modes: "rag", "graph", "hybrid" | |
| """ | |
| # New 5 modes - pass through directly | |
| NEW_MODES = ["analyst", "deep", "vision", "predict", "agent"] | |
| if mode in NEW_MODES: | |
| print(f"🚀 New mode selected: {mode.upper()}") | |
| return mode | |
| # Legacy modes - pass through | |
| LEGACY_MODES = ["rag", "graph", "hybrid", "graphrag", "prediction", "agentic", "multirag"] | |
| if mode in LEGACY_MODES: | |
| print(f"🔷 Legacy mode selected: {mode.upper()}") | |
| return mode | |
| # AI Model modes - pass through directly | |
| AI_MODEL_MODES = ["deepseek-chat", "mistral-7b", "llama-70b"] | |
| if mode in AI_MODEL_MODES: | |
| print(f"🤖 AI Model mode selected: {mode}") | |
| return mode | |
| # Vision mode priority | |
| if has_image: | |
| print(f"🟩 Routed to VISION mode (Image detected)") | |
| return "vision" | |
| # Auto-routing based on keywords | |
| q_lower = question.lower() | |
| # Check each category | |
| if any(kw in q_lower for kw in VISION_KEYWORDS): | |
| print(f"🟩 Routed to VISION mode (Image keywords)") | |
| return "vision" | |
| if any(kw in q_lower for kw in PREDICT_KEYWORDS): | |
| print(f"🔮 Routed to PREDICT mode (Forecast keywords)") | |
| return "predict" | |
| if any(kw in q_lower for kw in AGENT_KEYWORDS): | |
| print(f"🤖 Routed to AGENT mode (Web/external keywords)") | |
| return "agent" | |
| # Score graph vs rag | |
| graph_score = sum(1 for kw in GRAPH_KEYWORDS if kw in q_lower) | |
| rag_score = sum(1 for kw in RAG_KEYWORDS if kw in q_lower) | |
| if graph_score > rag_score: | |
| print(f"🟧 Routed to GRAPH mode (Relationship analysis)") | |
| return "graph" | |
| elif rag_score > 0: | |
| print(f"🟦 Routed to RAG mode (Document search)") | |
| return "rag" | |
| else: | |
| # Default to analyst mode for new 5-mode system | |
| print(f"📊 Routed to ANALYST mode (Default)") | |
| return "analyst" | |
| def is_business_query(question: str) -> bool: | |
| """Check if query is about business data (not personal/greeting)""" | |
| business_indicators = [ | |
| 'product', 'customer', 'revenue', 'sales', 'invoice', 'amount', | |
| 'total', 'lowest', 'highest', 'top', 'bottom', 'best', 'worst', | |
| 'performance', 'trend', 'analysis', 'data', 'report', 'show', | |
| 'list', 'which', 'what is the', 'how much', 'how many', 'compare' | |
| ] | |
| q_lower = question.lower() | |
| return any(kw in q_lower for kw in business_indicators) | |
| def is_greeting(question: str) -> bool: | |
| """Check if query is a greeting/personal message""" | |
| greetings = [ | |
| 'hello', 'hi', 'hey', 'howdy', 'how are you', | |
| 'thank you', 'thanks', 'bye', 'goodbye', | |
| 'good morning', 'good evening', 'good night', | |
| 'who are you', 'what are you', 'your name' | |
| ] | |
| q_lower = question.lower().strip() | |
| words = q_lower.split() | |
| # Short messages with greeting words | |
| if len(words) <= 5: | |
| return any(g in q_lower for g in greetings) | |
| return False | |