OfferRadar

# 🍔 OfferRadar — AI-Powered Swiggy Deal Finder > Chat with an AI agent that finds the best Swiggy food offers, ranks them by actual savings, and auto-applies the top coupon. Works with any MCP-compatible LLM — Claude, GPT-4o, Gemini, Llama, or local models via Ollama. --- ## What it does Most people miss out on Swiggy coupons because finding and comparing them across restaurants is tedious. OfferRadar automates the whole thing: 1. You say what you're craving (e.g. *"I want biryani"*) 2. The agent searches nearby restaurants 3. Fetches available coupons **in parallel** across all results 4. Ranks them by actual ₹ savings for your order 5. Builds your cart and auto-applies the best coupon 6. Shows you exactly how much you saved --- ## Supported LLMs OfferRadar uses **LiteLLM** as a universal LLM gateway. Switch models with one env var — no code changes: | Provider | Example Model | Needs API Key | |---|---|---| | Anthropic | `claude-opus-4-5` | Yes | | OpenAI | `gpt-4o` | Yes | | Google | `gemini/gemini-1.5-pro` | Yes | | Mistral | `mistral/mistral-large` | Yes | | Groq | `groq/llama-3-70b` | Yes (free tier) | | Ollama (local) | `ollama/llama3` | No | | Any OpenAI-compatible API | custom base URL | Optional | --- ## Demo ``` You: I want to order biryani, find me the best deal Agent: Searching restaurants near you... Found 8 places serving biryani. Fetching offers... 🏆 Best Deal: Behrouz Biryani Coupon: FLAT60 → Save ₹120 on ₹299+ order Runner-up: Paradise Biryani Coupon: SAVE30 → Save ₹90 on ₹349+ order Add Behrouz to cart and apply FLAT60? (yes/no): yes ✅ Coupon applied. Final total: ₹179 + delivery ``` --- ## Architecture ``` User (CLI) │ ▼ LLM Agent (agent.py) │ ├── LiteLLM Gateway (llm_client.py) │ ├── Anthropic Claude │ ├── OpenAI GPT-4o │ ├── Google Gemini │ ├── Groq / Llama 3 │ └── Ollama (local) │ └── Swiggy MCP Tools (swiggy_mcp.py) ├── search_restaurants() ├── fetch_food_coupons() ← parallelized with asyncio.gather ├── update_food_cart() ├── apply_food_coupon() └── place_food_order() │ ▼ mcp.swiggy.com/food ``` Tool definitions are written once in `tools.py` using OpenAI-style JSON schema. LiteLLM translates them to each provider's native format automatically. --- ## Setup ### Install ```bash git clone https://github.com/yourusername/offerradar cd offerradar pip install -r requirements.txt ``` ### Configure ```bash cp .env.example .env ``` ```env # Swiggy (required) SWIGGY_MCP_TOKEN=your_swiggy_mcp_token SWIGGY_ADDRESS_ID=your_address_id # Pick your LLM LLM_MODEL=claude-opus-4-5 # Anthropic # LLM_MODEL=gpt-4o # OpenAI # LLM_MODEL=gemini/gemini-1.5-pro # Google # LLM_MODEL=groq/llama-3-70b # Groq # LLM_MODEL=ollama/llama3 # Local — no API key needed # API key for your provider (skip for Ollama) ANTHROPIC_API_KEY=sk-ant-... # OPENAI_API_KEY=sk-... # GROQ_API_KEY=... ``` ### Run ```bash python main.py ``` ### Run fully local (free) ```bash # Install Ollama: https://ollama.com ollama pull llama3 LLM_MODEL=ollama/llama3 python main.py ``` --- ## Project Structure ``` offerradar/ ├── main.py # CLI entry point ├── agent.py # LLM agent + tool orchestration ├── llm_client.py # LiteLLM wrapper (swappable LLM layer) ├── swiggy_mcp.py # Swiggy MCP client ├── tools.py # Tool definitions (JSON schema for LLM) ├── coupon_ranker.py # Savings ranking logic ├── config.py # Env config ├── requirements.txt ├── .env.example ├── CLAUDE.md └── README.md ``` --- ## Roadmap - [ ] Instamart grocery deal finder - [ ] Dineout table booking - [ ] Streamlit web UI - [ ] WhatsApp bot interface - [ ] Streaming responses --- ## License MIT