Ignis Router

Intelligent LLM routing library that selects the best model for every query using ML routers and rule-based intent detection.

Features

• ML-based routing — KNN, SVM, Graph, MF routers predict the best LLM (via LLMRouter)
• Rule-based fallback — intent detection with automatic fallback when ML confidence is low
• Provider fallback — auto-switches to available LLM when API key is missing
• Decorators — @route(), @chat() for easy integration into any app
• REST API — FastAPI endpoints for routing and chat
• PostgreSQL logging — automatic persistence of routing decisions
• Retrainable — retrain ML models when new LLMs are introduced

Quick Install

pip install git+https://github.com/Infogain-GenAI/ignis_router.git@sakshi_dev_1

Quick Usage

from ignis_router import chat

@chat(system_prompt="You are a helpful assistant")
def ask(query, response):
    print(response["routing_decision"]["ml_router_predicted"])
    print(response["routing_decision"]["final_model"])
    print(response["content"])
    return response

ask("Write a Python sorting function")

Configuration

Create a .env file:

OPENAI_API_KEY=sk-your-key-here
ML_ROUTER_TYPE=svm
ENABLE_ML_MODEL_HINT_ROUTING=true
ML_CONFIDENCE_THRESHOLD=0.50
ROUTER_YAML_CONFIG=configs/cost-first.yaml

Run the API

python -m ignis_router.run_api
# Server at http://127.0.0.1:8080
# Swagger docs at http://127.0.0.1:8080/docs

API Endpoints

Architecture

User Query
    → ML Router (KNN/SVM/Graph/MF) predicts best model
    → API key check (fallback if unavailable)
    → Call LLM (OpenAI/Anthropic/Gemini)
    → Return response + routing decision
    → Save to PostgreSQL

Documentation

See user_guide.md for complete documentation including:

• Installation steps
• Configuration reference
• Decorator usage
• ML router training
• Database setup
• Troubleshooting

License

MIT