Summary

CookHero is an intelligent dietary and cooking management platform integrating LLM, RAG, Agent, and multimodal capabilities. It offers personalized meal planning, AI diet logging, nutrition analysis, and smart recipe querying, empowering users, especially novices, to achieve health and culinary goals.

How It Works

The system combines LLM reasoning with a hybrid RAG pipeline (Vector, BM25, Reranker) for precise retrieval, enhanced by multi-level caching. A ReAct Agent framework orchestrates tool usage, featuring a dynamic Subagent expert system for modularity. Multimodal input supports image-based food logging and recognition, protected by robust security layers against prompt injection and rate limits.

Quick Start & Requirements

• Installation: Docker Compose is recommended for setting up PostgreSQL, Redis, Milvus, and MinIO. Manual Python setup with pip install -r requirements.txt is also supported.
• Prerequisites: Python >= 3.12, Node.js >= 18. Essential API keys are required for LLM providers, image generation (DALL-E 3), web search (Tavily), image storage (imgbb), and database/cache services.
• Resources: Requires running multiple backend services.
• Docs: API documentation available at /docs.

Highlighted Details

• Agent System: Features a ReAct engine with a flexible toolset for diet management, knowledge retrieval, web search, and AI image generation. Supports a Subagent architecture and MCP protocol integration.
• Hybrid RAG: Combines vector search (Milvus), BM25, metadata filtering, and a Qwen3-Reranker for relevant results, optimized with Redis and Milvus caching.
• Multimodal Capabilities: Integrates image recognition for food/ingredient identification, enhancing diet logging and user intent understanding.
• Security: Implements multi-layered defenses including prompt injection protection, rate limiting, input validation, and structured audit logging.
• Nutrition & Planning: Provides detailed daily/weekly nutrition analysis, plan vs. actual tracking, and goal management.

Maintenance & Community

The project includes a detailed roadmap for future features like voice interaction and community functionalities. Contributions are welcomed via standard GitHub pull requests.

Licensing & Compatibility

Licensed under the Apache License 2.0, permitting commercial use and integration into closed-source projects.

Limitations & Caveats

The platform relies heavily on external API keys, incurring potential costs. Several advanced features, such as voice interaction and AR guidance, are still in planning. Setup requires managing multiple services and API configurations.