Skip to content

Releases: dondai44423/donsetch

v0.5.0-beta.1

v0.5.0-beta.1 Pre-release
Pre-release

Choose a tag to compare

@github-actions github-actions released this 07 Aug 15:07

v0.5.0 Beta — First Public Release

Pre-release

Choose a tag to compare

@dondai44423 dondai44423 released this 07 Aug 10:13

DonSeTch v0.5.0 Beta

Web fetch, search, and crawl for AI agents. Built from scratch in Rust. No keys, no accounts.

What's included

3 MCP tools:

  • fetch — Fetch any URL as clean markdown. Two-tier escalation: fast stealth HTTP (custom BoringSSL TLS, real Chrome ClientHello) → headless browser on bot wall. PDF parsing with OCR. BM25 focus extraction. Pagination.
  • search — Keyless multi-engine web search. 10+ backends in parallel, fused by cross-engine consensus + semantic reranking (local ONNX cross-encoder). Intent detection. Adaptive egress governor.
  • crawl — Best-first same-domain crawl. Sitemap discovery + frontier walk. Adaptive pacing. Resume tokens. Focus-ranked budget management.

Built from scratch

Component What it is
Custom BoringSSL TLS Real Chrome TLS fingerprint, mlkem post-quantum
Own HTTP/1.1 + HTTP/2 HPACK, flow control, connection pooling
DonSift extraction Block model, BM25 focus, token-war policies
DonSeek search Weighted RRF + BM25 + consensus + semantic rerank
DonTread crawl Sitemap, frontier, Governor pacing, resume
DonSheet PDF PDFium FFI, OCR arbitration, three-engine fusion

No Python, no Playwright, no Selenium, no reqwest, no hyper. Every layer built in Rust.

Stats

  • 249 tests, all passing
  • Zero clippy warnings (-Dwarnings)
  • 3-platform CI (Linux, macOS, Windows)
  • AGPL v3 licensed

Install

git clone https://github.com/dondai44423/donsetch.git
cd donsetch
cargo build --release

See the README for prerequisites and MCP client configuration.

Known limitations

  • Interactive captchas (hCaptcha, reCAPTCHA) are not solved — no solving service by design
  • ML-DSA post-quantum signatures pending BoringSSL support
  • Windows/macOS PDF subsystem CI verification pending

License

Copyright (c) 2026 Bishesh Bhandari. AGPL v3.