docs: bring help page, README and skill up to 3.7 - #48
Merged
Conversation
The help page had drifted to a v3.0 view of the service and described an extraction stack that no longer exists. It now covers what actually ships. Corrections: - "Readability + Turndown" described a dependency that was replaced by node-html-markdown; the pipeline list also omitted Trafilatura and the headless-Chromium stage entirely. - comments/comment_depth were documented as Reddit-only; they apply to Hacker News too since 3.1. - The share_id example was 6 hex, ids are 8. New help sections: endpoints (all 13, including /api/stream, archive, storage, stats, config, recipes status), response headers (incl. X-Transcript-Status and the five X-Extract-* headers), query extraction, site recipes, and authentication - the last one mattered most, since an auth-enabled instance previously served a help page that never mentioned login, API keys or OAuth. "Local HTML files" grew into "converting files" (/api/html, /api/file, document URLs, images/audio/YouTube), the parameter table gained render/extractor/pdf/yt_timecodes/yt_chunk/max_tokens, and the drop-in agent prompt was synced with the README version plus query. README gained the SSRF protection section (3.3.0 shipped the guard but the README never documented it), PULLMD_ALLOWED_HOSTS and PULLMD_SITE_RECIPES in the configuration table, /api/config and /api/recipes/status in the endpoint table, and a short list of what the 3.x line added after 3.0. The bundled skill learned Hacker News (pipeline + X-Source value), query / max_tokens with an example and a tip, X-Transcript-Status and X-Extract-*. Verified: 1038 tests pass; help renders with no leftover placeholders, 146/146 DE/EN element parity, TOC matching section order, every copy button resolving, no dangling anchors in either document, and /pullmd.zip building from the updated skill. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012EAefeA5Y3JuqZ5oAimbNw
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The in-app help page had drifted to a v3.0 view of the service, and described an extraction stack that no longer exists. This brings all three user-facing documents in line with what 3.7 actually ships.
Corrections
node-html-markdown. The same list also omitted Trafilatura and the headless-Chromium stage entirely.comments/comment_depthwere documented as Reddit-only; they apply to Hacker News as well since 3.1.share_idexample was 6 hex characters; ids are 8.Help page
Grew from 9 to 14 sections. New: endpoints (all 13, including
/api/stream, archive, storage, stats, config, recipes status), response headers (includingX-Transcript-Statusand the fiveX-Extract-*headers), query extraction, site recipes, and authentication.The authentication gap was the notable one: on an instance with
PULLMD_AUTH_MODEset,/helpstays public but never mentioned login, API keys or OAuth, so there was no documented way to find out how to authenticate against it."Local HTML files" became converting files, covering
/api/html,/api/file, document URLs and images/audio/YouTube. The parameter table gainedrender,extractor,pdf,yt_timecodes,yt_chunkandmax_tokens. The drop-in agent prompt is synced with the README version and addsquery.README
Adds the SSRF protection section: 3.3.0 shipped the guard, but the README never documented it, so self-hosters had no way to learn about the default-deny behavior or the
PULLMD_ALLOWED_HOSTSopt-out from the README alone. Also addsPULLMD_ALLOWED_HOSTSandPULLMD_SITE_RECIPESto the configuration table,/api/configand/api/recipes/statusto the endpoint table,query/max_tokens/pdf=ocrto the universal prompt, and a short list of what the 3.x line added after 3.0.Skill bundle
Hacker News (pipeline entry +
X-Sourcevalue),query/max_tokenswith an example and a usage tip,X-Transcript-Status, and theX-Extract-*headers.Verification
node --test: 1038 pass, 0 fail__PULLMD_*placeholders, version kicker renders, 146/146 DE/EN element parity with no per-section mismatch, TOC order matches section order in both languages, every copy button resolves its target/pullmd.zipbuilds from the updated skill (HTTP 200, instance URL substituted)Documentation only; no code paths touched.
🤖 Generated with Claude Code
https://claude.ai/code/session_012EAefeA5Y3JuqZ5oAimbNw