DocHarvestv10.0.1
Universal Doc Harvester & RAG Compiler
// platforms// output// studio// matrix// mcp// releases
Star2
DocHarvest v10.0.1 — AST Compiler & FastMCP

Turn any documentation into LLM-ready context & offline books.

Crawl GitBook, Mintlify, Docusaurus, Nextra, and ReadMe with AST precision. Compile clean Markdown, vector RAG JSONL, llms.txt, and printable PDFs with zero HTML noise.

Download for Windows (.exe)
$pip install gitbook-downloader
Speed

20.0 pgs/sec

Noise Redux

89% Token Cut

Local AI

FastMCP Ready

docharvest@terminal — v10.0.1
$ docharvest crawl https://docs.openalgo.in/v/v2.0 --rag --pdf --fast-ast
⚡ [Heuristic] Detected GitBook Space engine (version selector: v2.0)
🔍 Discovering documentation tree via sitemap and AST BFS...
├── Discovered: /api-reference/oauth [3.2 KB raw .md]
├── Discovered: /api-reference/orders [14.8 KB raw .md]
├── Discovered: /api-reference/positions [6.1 KB raw .md]
└── Discovered: /algorithms/quickstart [8.4 KB raw .md]
📥 Parallel AST crawl: 364 pages harvested in 18.2s (20.0 pages/sec)
📦 Compiling outputs:
├── book.md (consolidated single handbook with TOC)
├── llms.txt (standardized AI context manifest)
├── openalgo_rag.jsonl (vector chunks + SHA-256 metadata)
└── openalgo_handbook.pdf (publication-grade printable PDF)
✨ FastMCP server listening on stdio. Ready for Cursor & Claude Code!
STATUS: 364/364 HARVESTED
TIME: 18.2s (20.0 pgs/sec)
// 01 / FRAMEWORK INTELLIGENCE

Engineered For Every Major Documentation Platform

DocHarvest uses dynamic heuristics to detect documentation frameworks, bypassing DOM noise to extract author-original markdown directly.

Native Space IndexerPRIORITY: 100/100

GitBook Harvesting Heuristics

Deep GitBook spaces integration. Traverses multi-version dropdowns, parses space manifest JSON, and probes raw .md endpoints to fetch author-original markdown.

Target Sample:
https://docs.openalgo.in/v/v2.0/api-reference
Parsing Capabilities:
  • Direct .md raw endpoint probing bypassing HTML conversion
  • Multi-version space selector traversal (/v/v2.0/, /v/latest)
  • Preserves embedded code block tabs and parameter tables
  • Automatic TOC generation from summary tree structure
❌ Raw Scraper / Curl (HTML Noise)~42.8 KB
<!-- Raw Scraper Output (42.8 KB HTML Soup) -->
<div class="gitbook-root-container" data-space="sp_987x">
  <nav class="sidebar-nav-sticky-top"><div class="cookie-banner-wrap">...</div>
  <ul class="nav-tree-level-1"><li class="active"><a href="/v/v2.0/auth">OAuth</a></li>...</ul>
  <main class="page-content-wrapper">
    <div class="header-anchor-wrap"><h1 id="oauth2">OAuth 2.0 Auth<a class="anchor" href="#oauth2">¶</a></h1></div>
    <div class="alert alert-warning"><svg class="icon">...</svg><span>Token expires in 3600s</span></div>
    <div class="code-block-container" data-lang="python">
      <div class="code-header"><span class="lang-label">Python</span><button class="copy-btn">Copy</button></div>
      <pre><code><span class="token-keyword">import</span> <span class="token-variable">requests</span>...</code></pre>
    </div>
  </main>
</div>
✓ DocHarvest Output (LLM Context)~3.2 KB Clean
---
source_url: https://docs.openalgo.in/v/v2.0/api-reference/oauth
title: "OAuth 2.0 Authentication"
crawl_date: "2026-08-23T16:42:19Z"
content_hash: "sha256-e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
site_version: "v2.0"
---

# OAuth 2.0 Authentication

> ⚠️ **Warning:** Access tokens expire after 3600 seconds. Refresh via `/oauth/v2/token`.

## Request Signature

```python
import requests

response = requests.post(
    "https://api.openalgo.in/oauth/token",
    json={"client_id": "pk_live_...", "grant_type": "client_credentials"}
)
```
// 02 / THE FOUR-PART OUTPUT CONTRACT

One Crawl. Four Production-Ready Formats.

Every crawl emits a predictable four-part output matrix ready for Claude, Cursor, vector databases, and offline reading.

Universal ContextFILE: book.md

1. Consolidated Markdown — Unified book.md for LLM System Prompts

Merges the entire documentation portal into a single coherent, top-to-bottom Markdown file with an automatic table of contents and internal anchor links.

Contract Guarantees:
  • Automated hierarchical Table of Contents
  • Relative link remapping to internal document anchors
  • Zero redundant headers, sidebars, or footers
  • Perfect for pasting directly into Claude Project Knowledge or ChatGPT Custom GPTs
TARGET: book.mdREADY
$ docharvest export --format markdown
✓ Synthesized AST nodes into uniform format.
✓ Injected SHA-256 cryptographic hashes.
✓ Output written to disk in 0.4s.
// 03 / EXPORT STUDIO PREVIEW

Inspect the Generated Compilation Artifacts

Switch between raw generated artifacts to see how DocHarvest formats Markdown, vector JSONL, llms.txt, and SQLite FTS5 indices.

docs.openalgo.in/book.md
---
title: "OpenAlgo v2.0 Complete Developer Handbook"
source_domain: "docs.openalgo.in"
harvest_timestamp: "2026-08-23T16:42:19Z"
total_pages: 364
generator: "DocHarvest v10.0.1"
---

# Table of Contents
1. [Architecture Overview](#1-architecture-overview)
2. [OAuth 2.0 Authentication](#2-oauth-20-authentication)
3. [Order Execution API](#3-order-execution-api)
4. [WebSocket Market Data Feed](#4-websocket-market-data-feed)

---

# 1. Architecture Overview
OpenAlgo operates as an ultra-low latency execution broker wrapper...
SYNTAX: MARKDOWN
ZERO CLIENT OVERHEAD
// 04 / CAPABILITY COMPARISON

Why DocHarvest Outperforms Raw Scrapers & Cloud APIs

Compare DocHarvest against raw scrapers and proprietary cloud services.

Engine CapabilityDocHarvest v10.0.1Raw Scrapers (curl/Scrapy)Cloud Reader APIs
Native AST Heuristic Framework Detection (GitBook, Mintlify, Docusaurus)
Automatically isolates article DOMs and probes raw markdown endpoints directly.
Partial
Zero HTML/JSX Soup in Markdown Output
Strips cookie banners, navbars, sidebars, and interactive widget code.
Standard llms.txt & Vector RAG JSONL Compilation
Builds unified RAG chunk files with token counts and SHA-256 content hashes.
Pure-Python PDF Handbook Generation with TOC (fpdf2)
Zero external C-library dependencies (no WeasyPrint or wkhtmltopdf).
Built-in FastMCP Server for Cursor, Claude Desktop & Windsurf
8 native MCP tools running over stdio directly inside your IDE agent.
API Key Req
Embedded SQLite FTS5 BM25 Full-Text Search Database
Instant sub-15ms keyword search queries across thousands of harvested pages.
100% Free, Open Source (MIT) & Local Privacy
No subscription fees, no credit limits, and zero data leaves your local machine.
// 05 / MODEL CONTEXT PROTOCOL (FASTMCP)

Give Cursor & Claude Native Documentation Powers

Connect DocHarvest to your AI agent via FastMCP stdio in 30 seconds. Your agent gains tools to crawl, index, and query external documentation on demand.

.cursor/mcp.json
{
  "mcpServers": {
    "docharvest": {
      "command": "python",
      "args": ["-m", "gitbook_downloader.mcp_server"]
    }
  }
}
💡 Zero configuration needed. Once registered, ask your agent: "Search DocHarvest for OpenAlgo order payload schema."
8 Available Agent Tool Endpoints:
@docharvest_search()
SQLite FTS5 BM25 keyword query across all harvested documentation.
@docharvest_read_page()
Retrieves clean markdown article for a given URL or title anchor.
@docharvest_crawl()
Initiates background AST crawl on any external doc portal.
@docharvest_export_rag()
Generates chunked JSONL vector payload for vector embeddings.
// 06 / BUILT FOR YOUR WORKFLOW

Three Specialized Workflows. Zero Friction.

Whether you are training embeddings, coding on a plane, or managing documentation backups in CI/CD.

Vector & Agent Ready

AI & RAG Engineers

Pain Point:LLMs hallucinate on outdated training data. Scraping docs with curl dumps 40KB+ HTML noise that exhausts context windows.
DocHarvest Solution:DocHarvest compiles clean vector JSONL datasets with token counts, SHA-256 hashes, and standard llms.txt manifests for instant agent indexing.
Core Advantages:
  • Zero noise (89% prompt token reduction)
  • Direct drop-in for LangChain & ChromaDB
  • FastMCP server for Cursor & Claude Desktop
Air-Gapped Portability

Offline Developers & Researchers

Pain Point:Reading documentation on flights, during outages, or in secure air-gapped enterprise environments is painful with fragmented web pages.
DocHarvest Solution:Merges 500+ documentation pages into a single consolidated book.md and publication-grade PDF handbook with automated table of contents.
Core Advantages:
  • Pure-Python PDF generation (zero C-deps)
  • Embedded SQLite FTS5 search (<15ms queries)
  • Standalone desktop GUI (docharvest.exe)
Automated Synchronization

DevOps & Archival Teams

Pain Point:Docs change across releases without changelog notices. Mirroring docs locally often hits rate limits or gets blocked by Cloudflare.
DocHarvest Solution:Heuristic crawlers with cross-platform lock recovery, retry adapters, exponential backoffs, and automated Git diff tracking.
Core Advantages:
  • Semver snapshot diffs across crawls
  • Auto-recovering PID domain locks
  • Headless CLI automation in CI/CD
// 07 / RELEASES & TELEMETRY

Latest Release & Live Build Telemetry

Standalone binaries, release notes, and real-time commit logs from the active repository pipeline.

RohannShetty/gitbook-downloader★ 2
v10.0.1Published: 8/23/2026

v10.0.1

# DocHarvest v10.0.1 Version 10.0.1 resolves key UI state discrepancies, completes the standalone executable renaming (`docharvest.exe`), adds project renaming inside the Document Library, and introduces an interactive About section with full creator attribution. ### ✨ Features & Capabilities - ✏️ **Project Rename Feature in Document Library**: - Added `StorageManager.rename_domain` and `SearchIndex.rename_domain` to safely update directory paths, metadata JSON, and SQLite FTS5 search index entries. - Added a **Rename (✏️)** action button on library cards and an interactive rename modal dialog in `frontend/src/views/LibraryView.tsx`. - ❤️ **Interactive About Modal & Creator Attribution**: - Added `frontend/src/components/AboutModal.tsx` displaying engine specifications (AST + FastMCP + fpdf2), runtime telemetry, GitHub/Showcase links, and **"Made with ❤️ by Rohan Shetty"**. - Accessible via the sidebar footer heart trigger and the `Ctrl+K` command menu. ### 🐛 Bug Fixes & Hardening - 🏷️ **Universal Brand & Binary Propagation**: - Standalone build pipeline (`build_exe.py`) updated to produce `docharvest.exe` (primary) alongside `gitbook-dl.exe` for backwards compatibility. - Added `docharvest` CLI command entry point in `pyproject.toml` (`[project.scripts]`). - Desktop GUI window title, sidebar header, and Capture Studio badges updated to `DocHarvest v10.0.1` and `v10.0 Engine`. - `.github/workflows/build-release.yml` now stages and publishes both `docharvest-*` and `gitbook-dl-*` cross-platform binaries. ### ⚡ Performance & Architecture - ⚡ **Document Library Badge Race Condition Fixed**: - Eliminated the tick-0 mock fallback in `frontend/src/lib/bridge.ts` that caused the count badge to flash `(1)` before the WebView2 Python bridge attached. - Cleaned the sidebar navigation item badge for a distraction-free library view. ### 📦 Downloadable Binaries & Verification Checksums | Asset / Executable | Target Platform | File Size | SHA-256 Checksum | | :--- | :--- | :--- | :--- | | `docharvest-macos-latest` | macOS (Apple Silicon / Intel) | 29.4 MB | `9744f68b951cfe8b6d4e68e6038d7c1464c7b8eb97754b8ba73f0f5d38071c26` | | `docharvest-ubuntu-latest` | Linux x64 | 47.4 MB | `f9467b446b5528370bae581a41d1b170a731eccbbaffd4e3a1bf21c406b2bb63` | | `docharvest-windows-latest.exe` | Windows x64 | 33.5 MB | `03c6b6e06a6476000e15edc0ffc421c599b99bd8cdb1ff2517245edfbdf48135` | | `gitbook-dl-macos-latest` | macOS (Apple Silicon / Intel) | 29.4 MB | `9744f68b951cfe8b6d4e68e6038d7c1464c7b8eb97754b8ba73f0f5d38071c26` | | `gitbook-dl-ubuntu-latest` | Linux x64 | 47.4 MB | `f9467b446b5528370bae581a41d1b170a731eccbbaffd4e3a1bf21c406b2bb63` | | `gitbook-dl-windows-latest.exe` | Windows x64 | 33.5 MB | `03c6b6e06a6476000e15edc0ffc421c599b99bd8cdb1ff2517245edfbdf48135` | | `SHA256SUMS.txt` | All Platforms | < 1 KB | *(Verification Manifest)* | #### Verification Instructions ```bash # Linux / macOS integrity verification: sha256sum -c SHA256SUMS.txt # Windows PowerShell integrity verification: Get-FileHash -Algorithm SHA256 gitbook-dl-windows-latest.exe ``` --- **Full Changelog**: https://github.com/RohannShetty/gitbook-downloader/compare/v10.0.0...v10.0.1

Official Cross-Platform Binaries:
Windowsdocharvest.exe (32.9MB)Linuxx86_64 Binary (28.4MB)macOSUniversal Binary (31.2MB)
All releases signed and verified with SHA-256 checksums in GitHub CI.
Recent Commit Stream (master)LIVE PIPELINE
[b26d355]

docs: overhaul README with screenshots, add GitHub Profile README, and standardize social links

RohannShetty•8/25/2026
[e943727]

feat: embed StartupBar widget loader script

RohannShetty•8/25/2026
[2a93586]

feat(showcase): revamp DocHarvest website to Next.js 16 with Terminal-Engine theme, FastMCP showcase & live telemetry

RohannShetty•8/25/2026
[091ad1c]

fix(pages): enable trailingSlash for static GitHub Pages subpage routing

RohannShetty•8/25/2026
[ce09b35]

feat(pages): deploy Next.js portfolio website to GitHub Pages

RohannShetty•8/25/2026
[d7e0f0c]

feat: add premium personal portfolio Next.js website under portfolio/

RohannShetty•8/25/2026
[8c61e9e]

chore(release): v10.0.1 - DocHarvest hotfix release, binary naming, library rename & about section

RohannShetty•8/23/2026
[d9b10dd]

fix(docs): track docs/src/data and fix root data ignore rule

RohannShetty•8/23/2026
// FREQUENTLY ASKED QUESTIONS

Frequently Answered Questions

Everything you need to know about AST extraction, SPA documentation support, and FastMCP integration.

Basic scrapers dump messy HTML soup loaded with 40KB+ of cookie banners, navigation menus, and fragmented code blocks with broken indentation. DocHarvest is an engineered documentation compiler: it automatically detects frameworks (GitBook, Mintlify, Docusaurus), probes native .md raw endpoints directly, locks crawls strictly to doc subpaths, injects cryptographic SHA-256 YAML frontmatter, compiles unified book.md handbooks, exports pure-Python PDFs, and indexes everything into an embedded SQLite FTS5 BM25 search database.
DocHarvestv10.0.1

Universal documentation harvester, RAG vector dataset compiler, and publication-grade offline PDF generator for AI agents.

Made withbyRohan Shetty
// Navigation
  • // platforms
  • // output-contract
  • // export-studio
  • // capability-matrix
  • // fastmcp-server
  • // live-releases
// Community & Connect

DocHarvest is open-source under the MIT License. Contributions and PRs are always welcome.

© 2026 DocHarvest (GitBook Downloader). MIT Licensed.•Created by Rohan Shetty