Skip to content

docs: changelog for 0.5.1 and 0.5.2 - #11

Merged
glani merged 1 commit into
mainfrom
docs/changelog-0.5.2
Jul 30, 2026
Merged

glani merged 1 commit into
mainfrom
docs/changelog-0.5.2

Conversation

@glani

@glani glani commented Jul 30, 2026

Copy link
Copy Markdown
Collaborator

Records what shipped since the upstream merge, following the existing CHANGELOG.md / RELEASING.md convention rather than adding a competing release-notes file.

Entries

  • 0.5.1 — the NuQ prefetch livelock: async jobs never ran while sync scrape worked.
  • 0.5.2 — screenshot URL modes (public / signed / proxy), the DNA snip back in CI, and arm64 builds on native runners (68 min → ~8 min).
  • Unreleased — local-provider proxy coverage (test(storage): cover proxy mode over the local provider #10).

Two corrections, not additions

  • The 0.5.0 Deployment note is accurate about Redis/Postgres/RabbitMQ but names the wrong cause for what was actually broken. Redis was alive throughout and RabbitMQ was never required. Annotated in place rather than rewritten, so the record of what we believed at the time survives.
  • RELEASING.md still asked for 46/46 invariants. It is 53/53.

Local-run traps documented

Both produced wrong conclusions during this work, so they are now written down:

  • harness.ts starts its own RabbitMQ container when POSTGRES_HOST is localhost, which hides broker-less bugs. unset NUQ_RABBITMQ_URL does not help either — .env refills it.
  • playwright-service blocks private addresses without ALLOW_LOCAL_WEBHOOKS, so every local scrape returns empty. I first misattributed this to an unbuilt libhtml-to-markdown.so.
  • Screenshot snips need two MinIO buckets: public for the public-URL path, private for the presigned tests. Against a public bucket a broken signer still returns 200.

🤖 Generated with Claude Code

Records what shipped since the upstream merge: the NuQ prefetch livelock that
kept async jobs from running at all, the screenshot URL modes that make the
returned link fetchable, and the arm64 build going from 68 minutes to eight.

Two corrections rather than additions:

- The 0.5.0 "Deployment note" is accurate about Redis/Postgres/RabbitMQ but
  names the wrong cause for what was broken. Redis was alive; RabbitMQ was
  never required. Annotated in place instead of edited, so the record of what
  we believed at the time survives.
- RELEASING.md still asked for 46/46 invariants; it is 53/53.

Also documents the two local-run traps that cost real time and produced a wrong
diagnosis: the harness starting its own broker when POSTGRES_HOST is localhost,
and playwright-service blocking private addresses without ALLOW_LOCAL_WEBHOOKS
(which I first misread as an unbuilt go-html-to-md).

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
@glani
glani merged commit ce7303a into main Jul 30, 2026
9 checks passed
@glani
glani deleted the docs/changelog-0.5.2 branch July 30, 2026 17:04
@glani
glani restored the docs/changelog-0.5.2 branch October 25, 2026 21:06
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant