yokingma/SearChat

What it solves

SearChat provides a conversational search experience that combines the capabilities of large language models (LLMs) with real-time web search. It solves the problem of LLM "hallucinations" and outdated knowledge by grounding AI responses in current web data and offers a specialized "Deep Research" mode for generating comprehensive, cited reports on complex topics.

How it works

The system uses a multi-model and multi-engine architecture. It integrates various AI providers (OpenAI, Anthropic, Gemini) and search engines (SearXNG, Bing, Google, Tavily, etc.) via an adapter pattern. For standard queries, it performs conversational search; for complex tasks, it utilizes a "Deep Research" workflow orchestrated by LangChain and LangGraph to iteratively identify knowledge gaps, perform follow-up searches, and synthesize findings into a structured report with citations.

Who it’s for

It is designed for users who need an AI-powered research tool that can browse the live web, as well as developers looking to integrate deep research capabilities into their own Node.js projects via the deepsearcher NPM package.

Highlights

• Multi-Model & Engine Support: Compatible with major LLM APIs and a wide array of search engines, including the open-source SearXNG.
• Deep Research Mode: An iterative AI-driven workflow that automatically explores topics and generates structured reports with source citations.
• Conversational Interface: Supports multi-turn dialogues, real-time streaming responses, and browser-based chat history.
• Developer Friendly: Built with a Turborepo monorepo (Node.js/Koa and Vue 3/TypeScript) and deployable via Docker.