TinySearch v0.2.0发布：面向本地LLM的轻量级开源网络搜索工具，默认后端切换至SearXNG
英文摘要
TinySearch v0.2.0, the first stable beta, has been released. The tool switched from DuckDuckGo to SearXNG as the default search backend after DuckDuckGo started imposing frequent rate limits and CAPTCHAs. It is an MCP/FastAPI web-search utility that crawls, chunks, retrieves, and re-ranks content, delivering a context blob capped at 8k tokens optimized for smaller local LLMs. The update allows pointing to a self-hosted SearXNG instance for more reliable, provider-independent search. The developer reports using it successfully with qwen3.5-9B, with each call taking around 10-15 seconds.
中文摘要
TinySearch v0.2.0首个稳定测试版发布。因DuckDuckGo频繁触发速率限制和验证码，工具将默认搜索后端切换至SearXNG。这是一个MCP/FastAPI网络搜索工具，可爬取、分块、检索和重排内容，输出8k token上限的上下文片段，专为较小本地LLM优化。更新支持指向自托管SearXNG实例，提升搜索可靠性和独立性。开发者称已在qwen3.5-9B上正常使用，每次调用约10-15秒。
关键要点
TinySearch v0.2.0 stable beta released with SearXNG as the new default search backend.
TinySearch v0.2.0稳定测试版发布，默认搜索后端改为SearXNG。
The switch from DuckDuckGo was driven by increasing rate limits and CAPTCHAs, ensuring more reliable agent usage.
因DuckDuckGo频繁限流和验证码，切换后端以提升Agent使用可靠性。
The tool outputs a compact, source-grounded context blob capped at 8k tokens, designed for small local LLMs and MCP agents.
工具输出8k token上限的紧凑上下文片段，专为小型本地LLM和MCP代理设计。
Supports pointing to a self-hosted SearXNG instance for flexibility and independence from single search providers.
支持指向自托管SearXNG实例，实现供应商独立的灵活搜索。