Is TutuFind semantic search?

Today it is primarily BM25 keyword retrieval + jieba tokenization — explainable and controllable — with phrase queries matched via the field tokenizer.

Tuned specifically: a Chinese phrase is not collapsed into one token, so recall works as expected.

How does pricing compare to Exa?

TutuFind offers a free allowance; the public per-IP tier lets you try without paying first.

Can I get sources for citation?

Yes, every result carries a source (local/brave) so you can attribute in AI answers.

Exa alternative — the TutuFind API

BM25 keyword retrieval + jieba for Chinese, own index, structured results.

Read the search API docs

Exa is known for English-oriented vector / semantic retrieval. When you need reliable keyword recall — especially for Chinese — the BM25 + jieba approach of TutuFind is usually more direct and controllable.

TutuFind returns structured JSON with source attribution, falls back to the web when the local index is thin, and supports llms.txt — a good fit for RAG, agents and AI apps.

TutuFind vs Exa

比較項目	TutuFind	Exa
Retrieval	BM25 keyword + jieba CJK	Mostly vector / semantic
Explainability	✅ Term matches explainable	Black-box similarity
CJK tokenizer	✅ jieba	English-leaning
Source attribution	✅ local / brave	Yes
Free trial	✅ Public per-IP tier	By quota

主な機能

BM25 keyword retrieval

Explainable, controllable results — not a black-box vector; exact term / phrase recall is stable.

Chinese-first

jieba tokenization makes Chinese recall accurate; English gets a LowerCaser so case is not sensitive.

Source attribution

Every result is tagged local / brave for trust and citation.

Web fallback

When the local index lacks results, external search fills in automatically.

使い方

1
Get an API key

Request a key on the developer page, or try the public per-IP tier for free.
2
GET /api/v1/search

One request with your query and Authorization header.
3
Feed your model

Pass result titles / snippets / links as context to the LLM and keep source for citation.

よくある質問

Is TutuFind semantic search?: Today it is primarily BM25 keyword retrieval + jieba tokenization — explainable and controllable — with phrase queries matched via the field tokenizer.
How good is Chinese?: Tuned specifically: a Chinese phrase is not collapsed into one token, so recall works as expected.
How does pricing compare to Exa?: TutuFind offers a free allowance; the public per-IP tier lets you try without paying first.
Can I get sources for citation?: Yes, every result carries a source (local/brave) so you can attribute in AI answers.

今すぐ TutuFind API を組み込む

無料枠 · リクエスト 1 回で構造化された結果 · llms.txt 対応

Read the search API docs