Give LLMs and agents web search

Make TutuSoo the model’s search tool so agents reach the whole web live.

Read the search API docs

A model relying on training data alone cannot answer the latest questions. Wire TutuSoo as a function-calling tool and the model can search on its own when needed and get real-time results.

One GET request returns structured results with sources for judgement and citation; llms.txt is supported and Chinese is jieba-tokenized — a stable web entry point for agents.

主要功能

Drop-in tool

Define a search tool pointing at GET /api/v1/search and the model retrieves on its own.

Live web

Access the web on demand to answer questions past the training cutoff.

Citable

Results carry source / url so agents give verifiable answers.

Stable & compliant

Crawling respects robots.txt, outbound SSRF guard; the docs stay readable even if the backend hiccups.

使用步驟

  1. 1
    Define the search tool
    Register a tool in your agent framework that calls /api/v1/search.
  2. 2
    Model calls it
    When the model decides it needs the web, it issues a search request.
  3. 3
    Answer with results
    Return structured results and sources to the model for a cited answer.

常見問題

How do I wire it as an agent tool?
Wrap GET /api/v1/search as a function/tool that takes a query and returns the results array.
Can the model see sources?
Yes, each result carries source and url for verifiable answers.
Does it support Chinese queries?
Yes and tuned: jieba tokenization improves Chinese recall.
Is there rate limiting?
With a key it is per-key with X-RateLimit-* headers; the public tier is per IP.

現在就接入圖圖搜 API

免費額度 · 一行請求即得結構化結果 · 支援 llms.txt

Read the search API docs