Owner decision 2026-11-16: the extracted passages live in
references/book_corpus/texts/ instead of a host-provisioned file. The
loader and extractor default to that path; the full transcriptions stay
local-only.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXp2Vp62Ur3C7yRWQdSq
Owner decision 2026-11-16: deliver B. V. Raman's "How to Judge a Horoscope"
passages. Whether the verbatim text may live in git is not decided, so the text
stays in a local data file and its location is a deployment choice.
- scripts/research/extract_raman_htjh_house_texts.py reads the two book
transcriptions from a given directory and writes one data file (default
references/book_corpus/local/, gitignored): per house, the passage for its
lord in each of the twelve houses and for each of the nine planets occupying
it, with volume, heading, page, source line range and the text as transcribed.
What cannot be located is left out and listed in the coverage table.
- scripts/book_citations/raman_house_texts.py loads that file from
JYOTISH_RAMAN_HOUSE_TEXTS_PATH (same default) and builds the native layer
`classical_texts`: rows only for this chart's own placements, each capped at
600 characters at a sentence end. Without the file the layer reports
not_available_yet with the reason and nothing else changes.
- The card lists, per domain section, the rows of that domain's houses whose
placement equals what the card already shows; a passage shared by two domains
is written once. Each row carries its source and a fixed reference-only status.
- P-10 gains one sentence on how to use the passages; the facts catalog lists
them with the book catalog as source; product_runtime_contract.md section R
records the decision and the manual provisioning step (also deploy/README.md).
No book text is committed; tests use invented passages.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXp2Vp62Ur3C7yRWQdSq
- Split `validate` into `validate-python` and `validate-frontend`; publish
needs both. Runner capacity is >= 2 (publish and migrate already overlap).
- deploy/ci-python-venv.sh: venv kept under /root/.cache keyed by Python
version, ISO week, pins and requirements; persistent pip cache. A fresh
venv per run spent ~3 min re-downloading the same wheels.
- npm ci gets a persistent download cache and --prefer-offline (every one
of 1,010 packages was a cache miss).
- publish builds and pushes the API and web images concurrently; the web
Dockerfile mounts BuildKit caches for npm and .next/cache.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
The manual release quality gate now runs the legal-consent test with
JYOTISHA_REQUIRE_LEGAL_ENTITY=1 right after its frontend install, so a
production release fails while legal-entity.ts holds 「待填写」 placeholders
or the terms / privacy text is still marked draft. It was a manual step in
deploy/README.md until now. Approved by the product owner 2026-11-01.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyeZrYn6ryNcSe5Y2Y
A shared (security_headers) snippet, imported first in every site block of the
public, staging and production Caddyfiles: HSTS (one year, no
includeSubDomains), nosniff, strict-origin-when-cross-origin, framing refused
(X-Frame-Options DENY + frame-ancestors 'none'), camera/microphone/
geolocation/payment off, and the Server / X-Powered-By headers dropped.
Approved by the product owner 2026-11-01.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyeZrYn6ryNcSe5Y2Y
Publish job 11273 and its rerun 11284 timed out fetching from
registry.npmjs.org (zwitch, then zustand) while every validate check passed.
The gate's own install already uses registry.npmmirror.com; the image's npm ci
now does too. The lock file pins versions and integrity hashes, so the
installed packages are unchanged. Approved by the product owner 2026-11-01.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyeZrYn6ryNcSe5Y2Y
Public, statically generated knowledge pages from the approved prototype:
/learn lists 全部 plus seven columns with question titles and one-line
summaries (a sideways row on phones); /learn/<column>/<slug> renders the
question, the one-line answer, reviewer and date (or 「初稿 · 待占星师审校」),
the Markdown body, 「我们采用的口径」, 「在 Jyotisha 里看」, 「接着读」 and
参考出处.
Articles are frontend/content/learn/*.md with the frontmatter template from
the issue, read by a small parser and validated with zod; a bad file fails
the tests and next build with the file name and a reason. Drafts are
readable but noindex and stay out of sitemap.xml. robots.txt opens only
production. The web image copies frontend/content before the build.
Ships the one example article 「生时校正能做到多准?」 as an unsigned draft.
The chart page's ascendant line links it while the ascendant can still
change, through a 盘面要素 → 文章 table that only links existing articles.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyeZrYn6ryNcSe5Y2Y
Expired daily files are pruned before a new backup, a verify older than 36 hours raises an alert, and the host monitors call sudo with the production env file. The Linux 0600 check is on the staging quick list.
Co-Authored-By: Grok (agent loop)
Safety-net records move to BUG-1358 through BUG-1361 because staging already uses BUG-1355 through BUG-1357 for capacity.
Co-Authored-By: Grok (agent loop)
Product chose to keep the scheduled backup, but the current recovery point must stay the one created by hand. Daily runs still encrypt and restore-verify, then record a separate latest-backup note.
Co-Authored-By: Grok (agent loop)
Consult, rectification turns, and report generation stop before a credit reservation when the product caps are hit. A full heavy-compute gate waits up to 8 seconds, and one consultation retries that 429 inside the same request.
Co-Authored-By: Grok (agent loop)
Host rehearsal no longer needs Node on the production machine, and a failed cleanup now stops with the rehearsal database name. Docs-only staging heads no longer block rollback. Tests no longer hardcode Windows paths, and a trailing slash on the production site still refuses the test code.
pytest 12 passed; tsc exit 0; lint 0 errors; identity tests 9 passed, 0 cancelled. test:db was not run.
Co-Authored-By: Grok (agent loop)
Gitea run 2828 validated, then publish hit the 60-minute deadline while
Aliyun apt/pip crawled at about 22 kB/s after a 72h cache prune on a
runner with 677 GiB free. Skip BuildKit prune when disk is above the
minimum, split API apt/pip layers, and point apt/pypi at Huawei Cloud.
Image build patches vedastro.check_for_update to a no-op after pip install so import cannot hit pypi or upgrade the pin. Free-tier queue waits are capped to VEDASTRO_TIMEOUT_SECONDS and fail fast as free_tier_rate_limited. Gateway status exposes runtime mode as booleans without secrets. BUG-719 resolved; BUG-720 stays investigating until production env is filled in.
Quality gate now dispatches Migrate Staging Database, waits for success, then dispatches Deploy staging. Automatic migrate attests the in-progress gate run so the two jobs cannot deadlock. Manual migrate is unchanged. Staging schema changes must stay backward-compatible with the currently deployed app.
Leave Migrate Staging Database deploy_sha empty (or type latest) to use the newest successful backend-quality-gate push. Filled 40-hex still takes the original path. Production workflows stay required.
Add Apache-2.0 @4n6h4x0r/stem-branch 0.8.0, a seven-governors adapter,
and POST /api/qizheng, /api/western, /api/ephemeris_events.
BUG-700 remains blocked (do not call --pillars). BUG-701 and BUG-702 are
resolved. BUG-703 is investigating (panchanga Lahiri). Skill is not bumped.
Engine calls already returned markdown; the worker failed because local postgres upsert required onConflict and the appendix write omitted it.
Co-authored-by: Cursor <cursoragent@cursor.com>
Staging PostgREST rejects the executiveSummary JSON-path alias, so GET /api/reports 500s. Persist a plain card_summary column from the Markdown excerpt instead.
Co-authored-by: Cursor <cursoragent@cursor.com>
Docs-only staging HEAD blocked Migrate Staging Database even when an ancestor already had a successful exact-SHA gate, so pending SQL could not be applied.
Co-authored-by: Cursor <cursoragent@cursor.com>
Twenty-four of the last sixty staging pushes were pure documentation, yet each
one ran (and cancelled) the full gate and image publish. Introduce
deploy/gated-paths.txt as the single source of truth for what must rerun the
gate: every Dockerfile COPY source, the Python package inputs, the workflow and
build-context files, and the repository files frontend/tests read at gate time.
Both triggers of backend-quality-gate.yml now carry that exact list; pushes
that touch none of it neither run the gate nor cancel a running code gate.
Because staging head may then legitimately sit ahead of the last tested SHA,
add deploy/is-docs-only-range.sh: it proves <base> is an ancestor of <head>
and that no changed path matches a gated glob, from local history when it is
available and otherwise from the Gitea compare API (per-commit `files`, parent
walk for ancestry, total_commits cross-checked). The publish dispatch and the
deploy-staging head checks accept an advanced head only when that script
succeeds; diverged, older, or code-bearing heads are still refused. In
deploy-staging the check runs after the gate-attested controller bundle is
extracted so only the tested checker and path list are ever executed; the
manual rollback branch is unchanged.
AGENTS.md §6.3/§6.4 describe the new contract: `.deployment.gitCommit` must
equal the latest staging commit that touched a gated path, not staging head.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VawU7Xfd5jS9wUEXz1XYmS
deploy/reclaim-runner-disk.sh ran `docker builder prune --force --all` at the
start of every gate job, so each publish rebuilt the Dockerfile `npm ci` layer
from scratch behind the registry mirror (7-minute image builds became 47).
Reclaim build cache in tiers instead: prune entries unused for 72h, re-measure
free space, and escalate to `--all` only while still below MINIMUM_FREE_GIB.
The final threshold check and the "only unheld resources" rules are unchanged.
Contract test asserts the aged prune runs first and that `--all` is only
reachable inside the re-measured conditional.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VawU7Xfd5jS9wUEXz1XYmS
Gitea (git.copse.top) is the only CI/CD control plane; GitHub is a
read-only mirror whose Actions are being disabled in repository settings.
- Delete all 11 `.github/workflows/*.yml` (stale copies of the old design).
- Delete unused Gitea manual workflows `ci.yml`, `test.yml`,
`publish-pypi.yml`, `apply-supabase-profile-migrations.yml`
(0-1 historical runs, no remaining target).
- Fold the full `python -m pytest` tree and `tests/run_all.py` into
`release-quality-gate.yml`, which previously only ran the curated
release profile; update the `run_quality_gate.py` comment accordingly.
- Port `reset-staging-account.yml` to Gitea: `runs-on: xiaoxin`, bounded
exact-SHA checkout, `refs/heads/staging` only, `staging-mutation`
concurrency, same email/confirmation/host/port/user/path assertions.
- Repoint frontend workflow tests at `.gitea/workflows/`, drop the
GitHub-only assertions, add coverage for the new reset workflow, and
remove `tests/test_supabase_profile_migration_workflow.py`.
- Update AGENTS.md §6.8, README.md, and deploy/README.md to the current
production/mirror facts and document the staging account reset.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VawU7Xfd5jS9wUEXz1XYmS
Keep async job and chart-cache files across API recreates, freeze jyotish_api_server.py growth, and fail fast with 429 when rectification or high-rigor compute is saturated.
Co-authored-by: Cursor <cursoragent@cursor.com>
Homepage bootstrap never escaped the loading shell when a new web image changed chunk hashes. Stamp deploymentId, stop caching chat HTML, and hard-reload failed chunk loads once.
Co-authored-by: Cursor <cursoragent@cursor.com>
Staging publish failed resolving python:3.12-slim through m.daocloud.io; the web image already uses the reachable SWR docker.io/library path.
Co-authored-by: Cursor <cursoragent@cursor.com>
Manual SKILL.md updates were blocked by registry sha256 and a byte-equal versions/ gate. Consult now loads the operator-maintained tree; rectification and personal-report stay hashed.
Co-authored-by: Cursor <cursoragent@cursor.com>
xiaoxin's 20-core npm test opened one Docker network per database file and exhausted default address pools. Limit concurrent fixtures and remove unused jyotisha-postgres networks before the gate.
Co-authored-by: Cursor <cursoragent@cursor.com>
Next's output file tracing pulls skills/ into .next/standalone with symlinks
resolved into real directories. The final stage copies the standalone bundle
into /app first, so the later explicit `COPY skills /app/skills` tries to put
the tracked symlink skills/jyotish-vedic-astrology/assets -> ../../assets on
top of a real directory. Whether that is fatal depends on the BuildKit
snapshotter: GitHub's runners tolerate it, Docker 29.1.3 with
containerd-overlayfs refuses with "cannot replace to directory ... with file".
Dropping the traced copy after the build lets the explicit COPY own that tree
deterministically. /app/assets still comes from the standalone bundle, which
outputFileTracingIncludes guarantees, so the symlink resolves.
Verified by building the image and inspecting it: skills/ complete at 1208
files, assets still a symlink that resolves, container boots on Next.js
16.3.1 and serves / with HTTP 200. Reproduced the original failure on the
pre-upgrade commit too, so it predates the Next 16.3.1 upgrade.
Co-authored-by: Cursor <cursoragent@cursor.com>
The staging host currently has .state/deployed-revision unreadable by the
deploy user, so the forward-only revision check failed with 'Permission
denied' before it could compare revisions. Treat an unreadable state file
like a missing one and discover the running revision from the web
container's GITHUB_SHA, in both the workflow step and the server-side
deploy/migration scripts.