Skip to content

Releases: AnkitParekh007/devdocs-forge-agent

v0.1.1 — Transcript Intake Agent + Video Intake Guard

13 May 16:40
@AnkitParekh007 AnkitParekh007
v0.1.1
33b380e

Choose a tag to compare

View all tags
v0.1.1 — Transcript Intake Agent + Video Intake Guard Latest
Latest

What's New in v0.1.1

Transcript Intake Agent

A complete safe transcript import system. devdocs-forge-agent never scrapes YouTube — you bring your own transcript via:

  • transcript clipboard — import directly from system clipboard (macOS/Windows/Linux)
  • transcript paste — paste multi-line stdin, end with :::end
  • transcript import-file — import .md, .txt, .vtt, .srt (timestamps stripped automatically)
  • transcript youtube-owner — stub: OAuth owner caption download (instructions shown)
  • transcript transcribe-file — stub: local media transcription (instructions shown)

When --url is provided without --file, generate and validate-source now show step-by-step intake instructions instead of a raw error.

Video Intake Guard improvements

  • Added AI coding tool keywords: cursor, coding agent, coding agents, ai coding, claude code, agentic coding, ide, developer tools, and more
  • New multi-keyword title bonus (+20 points) when 2+ tech keywords appear in the title — "Cursor: coding agents tutorial" now correctly scores ≥ 60 (high confidence)

Validate-source UX fix

  • FAIL Transcript: 0 words (too short) → FAIL Transcript: file not found when the file doesn't exist
  • Shows intake instructions after a missing-file error

Config

New transcript_intake section in config/devdocs-forge-agent.yml:

transcript_intake:
  enabled: true
  allow_clipboard: true
  allow_paste: true
  allow_file_import: true
  allow_public_youtube_scraping: false  # always false
  min_words: 150
  warn_words: 500

CI

  • Removed EOL Node 18.x and 20.x from CI matrix
  • Added Node 22.x (Active LTS) and 24.x (Current)

Fixes

  • README tables aligned with consistent column widths
  • .env.example comment typo fixed (Referhttps:// → See: https://)
  • Node badge updated to >=22

Tests

82 tests passing across 9 test suites, including new:

  • tests/transcript-normalizer.test.ts (13 tests)
  • tests/transcript-intake.test.ts (12 tests)
  • 3 new tech-video-classifier tests for AI coding keywords

Full changelog: v0.1.0...v0.1.1

Assets 2
Loading

v0.1.0 — CLI MVP

12 May 21:13
@AnkitParekh007 AnkitParekh007
v0.1.0
9285dbe

Choose a tag to compare

View all tags
v0.1.0 — CLI MVP

v0.1.0 — CLI MVP

First public release of devdocs-forge-agent.

Release date: 2026-05-12

Highlights

  • Local-first AI documentation CLI — runs entirely on your machine
  • Mock provider — generates real structured output with zero API keys
  • OpenAI, Anthropic, and Gemini provider support via native fetch (no SDK packages)
  • 11 documentation output modes — docusaurus, blog, docs, gitbook, readme, faq, troubleshooting, lesson, social, changelog, seo
  • Video Intake Guard — validates video URLs, classifies tech content, requires user-provided transcripts
  • Docusaurus/GitBook-ready output with correct frontmatter
  • Batch mode — process all transcripts in a directory
  • Markdown validator — verify generated output before publishing
  • Human review checklist — generated with every run
  • Example transcripts and pre-generated outputs — zero-to-docs in one command
  • 54 passing Vitest tests
  • GitHub Actions CI — Node 18/20 matrix on ubuntu/macos

Safety

  • No YouTube scraping — never included, never planned
  • No video downloading
  • Transcript must be provided by the user
  • Source attribution support built in
  • Human review checklist generated with every run

Added

CLI Commands

  • init — initialize config, input/, output/, profile template
  • doctor — validate setup: Node version, config, provider, API keys, directories, YOUTUBE_API_KEY
  • generate — generate documentation from a transcript file with optional --url, --type, --force
  • batch — process all .md and .txt files in a directory
  • verify — validate generated documentation for quality issues
  • providers — list available AI providers
  • examples — copy example transcripts to input/
  • inspect-url — inspect and classify a video URL without fetching transcripts
  • validate-source — validate a video URL + local transcript before generation

Provider Abstraction

  • Mock provider (default, no key needed)
  • OpenAI provider (chat completions via native fetch)
  • Anthropic provider (messages API via native fetch)
  • Gemini provider (generateContent via native fetch)

Documentation Modes

  • docusaurus — Docusaurus v3 page with YAML frontmatter
  • blog — developer blog post with SEO metadata
  • docs — general documentation page
  • gitbook — GitBook-formatted doc with hints and callouts
  • readme — GitHub README tutorial
  • faq — FAQ organized by category
  • troubleshooting — error/fix troubleshooting guide
  • lesson — course lesson with objectives and exercises
  • social — LinkedIn + X + dev.to social summaries
  • changelog — Keep A Changelog format release notes
  • seo — SEO metadata and keyword analysis

Video Intake Guard

  • URL parsing for YouTube (youtube.com/watch, youtu.be) and Vimeo
  • YouTube Data API v3 metadata fetch (title, description, tags, category, duration — never captions)
  • Tech video classification with keyword scoring (score 0–100)
  • Transcript validation (word count, VTT/SRT timestamp stripping)
  • inspect-url command for standalone URL classification
  • validate-source command for full pre-generation validation
  • Graceful degradation without YOUTUBE_API_KEY

Infrastructure

  • TypeScript 5.x with strict mode, ESM, NodeNext module resolution
  • Vitest test suite (54 tests across 7 test files)
  • GitHub Actions CI (Node 18/20 × ubuntu/macos)
  • Zod config schema with full defaults
  • AGENTS.md canonical AI agent instructions
  • CLAUDE.md behavioral guidelines
  • Full open-source file suite (CONTRIBUTING, SECURITY, SUPPORT, LEGAL_DISCLAIMER, GOVERNANCE, CODE_OF_CONDUCT, ROADMAP, DATA_CONTRACT)

Known Limitations

  • Mock provider generates placeholder documentation structure (not AI-quality prose) — switch to OpenAI/Anthropic/Gemini for production output
  • YouTube metadata requires YOUTUBE_API_KEY — without it, classification uses title/URL heuristics only
  • Vimeo metadata fetch not yet implemented (classification uses URL/filename only)
  • Transcript chunking for very long videos (>10k words) not yet implemented

Breaking Changes

None — this is the initial release.

How to Install

git clone https://github.com/AnkitParekh007/devdocs-forge-agent.git
cd devdocs-forge-agent
npm install
cp .env.example .env
npm run demo

How to Publish the GitHub Release

# Requires GitHub CLI authenticated with repo write access
gh release create v0.1.0 \
  --title "v0.1.0 — CLI MVP" \
  --notes-file docs/release-notes/v0.1.0.md \
  --latest

Or via GitHub UI:

  1. Go to https://github.com/AnkitParekh007/devdocs-forge-agent/releases
  2. Click Draft a new release
  3. Tag: v0.1.0
  4. Title: v0.1.0 — CLI MVP
  5. Paste these release notes
  6. Click Publish release
Loading