mirror of
https://github.com/supabase/supabase.git
synced 2026-10-05 17:35:10 +03:00
codex/fix-tanstack-e2e
7
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 Alaister YoungandAlaister Young
|
f125126aec |
chore: make agent instructions agent-agnostic (#49941)
Makes the repo's AI-agent setup tool-agnostic: instructions live in
`AGENTS.md` files, skills live in `.agents/skills/`, and Claude Code,
Codex, Cursor, and Copilot all read the same sources. Also sweeps the
skills for stale and duplicated content while everything was being
moved.
**Changed:**
- Every `CLAUDE.md` (root, `apps/studio`, `apps/docs`, `apps/kb`) is now
a one-line `@AGENTS.md` import; the content moved verbatim into an
`AGENTS.md` beside it. The root one moved from `.claude/CLAUDE.md` to
the repo root for consistency.
- All skills now live in `.agents/skills/`; `.claude/skills` is a single
symlink to it (replacing the old mix of real dirs and per-skill
symlinks). Path references in `.coderabbit.yaml`, code comments, and
docs updated to match.
- `.github/copilot-instructions.md` keeps only the review policy and
points at `AGENTS.md` + `.agents/skills/`. Copilot code review reads
those natively now, so the per-topic
`.github/instructions/*.instructions.md` files were duplicates of the
skills.
- Stale skill content fixed: `studio-queries` imported a toast library
Studio doesn't use, `telemetry-standards` and `studio-testing` used
import paths that don't resolve, `safe-sql-execution` cited a boundary
test that doesn't exist, the ask-the-docs references described an
`AiPrompt` mechanism that was replaced by the ID-keyed registry, plus a
handful of wrong paths, a self-contradicting `waitForTimeout` rule, an
invalid Playwright signature, and a ConfigCat flag described as PostHog.
- `studio-error-handling` now explains when to use `AlertError` (the
default) vs `ErrorMatcher`.
**Added:**
- `apps/docs/AGENTS.md` (docs test requirements, from the old Cursor
rule)
- `studio-shortcuts` skill (from the old Copilot instruction file,
verified against the current registry)
- `ask-the-docs/reference/graphql-endpoint.md` and
`search-embeddings.md` (from the old Cursor rules, with the missing
resolver/registration/codegen steps filled in)
- Feature-flag measurement section in `telemetry-standards`
**Removed:**
- `.cursor/` (rules folded in as above; skill symlinks no longer needed)
and `.cursorignore`
- `.github/instructions/` (8 files)
- `vercel-composition-patterns/AGENTS.md` – a 946-line verbatim
concatenation of its own `rules/` directory, and a nested `AGENTS.md`
that agents could auto-load as repo instructions
- `edit-the-docs/reference/structure-and-flow.md` – word-for-word copy
of the skill's own Phase 2 text
## To test
- `readlink .claude/skills` → `../.agents/skills`, and `ls
.claude/skills/copywriting/SKILL.md` resolves
- Open a Claude Code session at the repo root and in `apps/studio` – the
imported `AGENTS.md` content should load as before
- `git diff master --stat -M` shows the skill moves as 100% renames
(content unchanged except the listed fixes)
- Spot-check a fixed claim, e.g. `import { toast } from 'sonner'` in
`studio-queries`, or the `logs.all` ESLint rule cited in
`clickhouse-logs-queries/references/codebase-integration.md`
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
- **Documentation**
- Expanded guidance for documentation workflows, GraphQL resources,
search, ClickHouse logs, React forms, Studio testing, shortcuts,
telemetry, accessibility, copywriting, and composition patterns.
- Clarified local testing, linting, build workflows, error handling, and
AI coding agent usage.
- Added contributor guidance for the knowledge base, documentation, and
Studio areas.
- **Chores**
- Consolidated agent instructions and skill references.
- Removed obsolete editor-specific guidance, duplicate links, and
superseded documentation.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Alaister Young <10985857+alaister@users.noreply.github.com>
|
||
|
Wen Bo Xie
|
2681a21f5c |
docs: add Personal Access Tokens guide with generated permission tables (#49732)
Add a guide that compares classic and scoped personal access tokens, explains how account roles constrain token permissions, and walks through creating and testing a project-scoped token. Include generated tables mapping permissions to Management API endpoints and MCP tools, and link the guide from docs navigation and Studio token sheets. Move the scoped-token permission catalog from Studio into shared-data. Studio and docs generation now share permission names, categories, descriptions, risk metadata, modes, scopes, and display order. Generate the tables from the shared catalog, OpenAPI x-fga-permissions, and the downloaded MCP permission map. Exclude Workers permissions until the feature is live. Run regeneration through the docs Makefile, verify checked-in output in CI, and refresh it in the weekly Management API workflow. Add Dashboard and Docs ownership plus contributor guidance so permission changes stay synchronized. |
||
|
Alaister YoungandAlaister Young
|
2c76bb371b |
chore(studio): gate dead code with knip in CI (#49721)
Makes knip a CI gate for Studio so dead files and unused dependencies fail the PR instead of piling up. Third PR in the stack, on top of #49719 (dead code) and #49720 (unused deps), which get Studio to a clean run. **Changed:** - knip `pnpx knip@~5.50.0` → root devDependency `knip@6.32.3`, `pnpm knip` now runs it. The old `pnpx` form was actually broken: it resolved knip's `typescript` peer to TS 7 and crashed with `ts.getDefaultLibFilePath is not a function`. (6.33.0 is newer but blocked by `minimumReleaseAge`.) - `knip.jsonc` rewritten for v6 with a `workspaces["apps/studio"]` block. Framework-convention files (`router.tsx`, `start.ts`, `routes/**`, `compat/**`, `api/server.js`) are `entry` rather than `ignore` — an ignored file's imports aren't traced, which is how `ShellFallback.tsx` (only imported from `routes/__root.tsx`) was being reported as dead. knip 6's Next.js plugin already covers `instrumentation*.ts`, `proxy.ts`, `pages/**`; its tanstack-router plugin only looks under `src/`, hence the manual entries. Narrow `ignoreIssues` for graphql-codegen output and the `CONSTRAINT_TYPE` enum; `ignoreDependencies` for the five implicit deps from #49720, each with a comment; `ignoreBinaries: ["vercel"]`. - `apps/studio/CLAUDE.md`: one bullet on the gate and where framework files go. **Added:** - `.github/workflows/studio-knip.yml` — path-filtered to `apps/studio/**` + knip/pnpm config, mirrors `studio-lint-ratchet.yml`'s setup (no sparse checkout: knip needs every workspace's `package.json` to resolve the graph). Runs `pnpm knip --workspace apps/studio --reporter symbols --reporter github-actions` so findings show up as inline PR annotations. ~5s locally. Scope notes: the gate is Studio-only — the full-monorepo run still has ~400 dead files in `www`/`docs`/`blocks`, which is a separate effort. `exclude: ["types", "exports"]` is kept, so unused exports aren't gated yet, but `enumMembers`/`duplicates` are (they caught real things in #49719). ## To test - `pnpm knip --workspace apps/studio` exits 0 on this branch - The `Studio Dead Code (knip)` workflow runs on this PR and is green - Sanity-check the gate bites: add a throwaway `apps/studio/lib/unused.ts`, run `pnpm knip --workspace apps/studio` → reports it and exits 1 <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * **CI** * Added automated dead-code and unused-dependency checks for the Studio workspace on relevant pushes and pull requests. * Results appear in workflow summaries and as inline pull request annotations. * **Maintenance** * Improved analysis of framework-convention files and Studio code. * Standardized the local code-quality check and updated its configuration support. <!-- end of auto-generated comment: release notes by coderabbit.ai --> --------- Co-authored-by: Alaister Young <10985857+alaister@users.noreply.github.com> |
||
|
Alaister YoungandAlaister Young
|
d845768fcf |
chore(claude): add react-hook-form skill (#48431)
Adds a Claude skill encoding correct React Hook Form usage, so
AI-written form code follows best practices instead of copying the
anti-patterns common in older Studio code (prop-form
`form.watch()`/`formState` subscriptions, subscription-only watches,
unguarded `valueAsNumber`, `?? undefined` controlled values, defaults
computed from unloaded queries).
**Added:**
- `.claude/skills/react-hook-form/SKILL.md` — subscription model
(`useWatch`/`useFormState` with `control`), canonical zod + `FormField`
composition (layout deferred to `studio-ui-patterns`), `values:` option
for async data, null normalization for controlled inputs, number-input
handling, dirty-state and gating rules, plus a fix-what-you-touch policy
aligned with the `no-use-watch` lint ratchet
**Changed:**
- `.claude/CLAUDE.md` and `apps/studio/CLAUDE.md` — register the skill
in the skill lists/table
- `.coderabbit.yaml` — add the skill to the existing Studio
code-guidelines entry so CodeRabbit applies it when reviewing Studio
code
Benchmarked on three real form tasks (adding a live-updating field to
`ThroughputField`, a new sheet form with async + nullable data, a
review-changes step in `EditBucketModal`), each run with and without the
skill: 13/13 assertions with the skill vs 8/13 baseline. The baseline
shipped a genuine bug in one task — a `null` server default flowed into
a `''` its own schema rejected, making Save unreachable — which the
skill run avoided.
## To test
- Ask Claude Code to add a field to any Studio form and check it loads
the skill (it's in the studio CLAUDE.md skill table) and uses
`useWatch({ control, name })` rather than `form.watch`
- Skim `SKILL.md` for anything that contradicts current form conventions
— `apps/design-system` demos remain the layout source of truth
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Documentation**
* Added a new monorepo “react-hook-form” skill guide with recommended
patterns for safe form subscriptions, wiring, default values,
reset/submission flows, and common anti-patterns.
* Updated Studio skills/load guidance to expand and reorder the skills
matrix, including form logic and copywriting guidance.
* Updated required skill coverage so `react-hook-form` is included for
any form-related work.
* **Chores**
* Expanded automated review enforcement so Studio form code is checked
against the new “react-hook-form” skill guidance.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Alaister Young <10985857+alaister@users.noreply.github.com>
|
||
|
Charis
|
d5436ae826 |
feat(studio): log date range domain + session logRange state (#48401)
## I have read the [CONTRIBUTING.md](https://github.com/supabase/supabase/blob/master/CONTRIBUTING.md) file. YES ## What kind of change does this PR introduce? Feature (+ a small refactor and a docs/convention note). PR 4 of the stacked SQL-editor query-source series (Database vs Logs). ## What is the current behavior? The SQL editor has no representation of a logs query's time range: `querySource.ts` only knows how to map a snippet type to a source (`getSnippetSource`), and session state (`sql-editor-session-state.ts`) tracks results and the row limit but not a per-snippet time range. The Logs date picker's pure range helpers (`parseCustomInput`, `generateDynamicHelper`, the `Unit` type) are trapped inside the `Logs.DatePickers.tsx` React component. ## What is the new behavior? - **Logs time-range domain** in `querySource.ts`: branded `IsoDateTimeString` + `isoDateTimeString()`, `RelativeTimeUnit`, a `LogDateRange` discriminated union (relative/absolute), `DEFAULT_LOG_DATE_RANGE`, a single date-picker parser (`datePickerValueToLogDateRange` / `logDateRangeToDatePickerValue` — handles the five presets *and* dynamic `2h`/`30m` helpers; `calcTo === ''` means "now"; unparseable helpers degrade to absolute), and `resolveLogRunRange` which re-resolves relative ranges against `now` at run time (reusing the existing `ResolvedLogDateRange` shape). - **Session state**: per-snippet `logRange` + `setLogRange` — session-only, never written to snippet content, so it works on read-only shared snippets and is cleaned up in `clearForSnippet`. - **Refactor**: extracted the picker's framework-free helpers into a new pure `Logs.datePickerHelpers.ts`; the logs domain now shares the `Unit` type and reuses `generateDynamicHelper` instead of duplicating them. Importers point at the new module directly (no re-export shim). Hardened the amount parse against `NaN`. - **Full unit coverage** in `querySource.test.ts`. Recorded the no-shim refactoring convention in the `studio-best-practices` skill. Verification: `pnpm typecheck` clean, lint ratchet improved, 43 tests pass (querySource + Logs.Datepickers), Prettier clean. ## Additional context <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit - **New Features** - Added robust Logs date-range modeling with support for relative (e.g., last N units) and absolute time periods. - SQL Editor sessions now remember log date ranges per snippet. - **Bug Fixes** - Safer handling of invalid or missing date inputs, with sensible fallback to default/current time. - **Tests** - Added/expanded automated coverage for date-range conversion, helper parsing, and resolution behavior. - **Refactor** - Centralized date-picker helper utilities for reuse across the Logs and SQL query experience. <!-- end of auto-generated comment: release notes by coderabbit.ai --> |
||
|
Alaister YoungandAlaister Young
|
4b24cf028a |
chore(claude): improve CLAUDE.md files and skill triggering (#48261)
Improves the repo's agent guidance: distills the always-required `studio-best-practices` skill into `apps/studio/CLAUDE.md`, tunes every skill description for reliable triggering, and mechanically enforces the generated-files rule. Grounded in Anthropic's official CLAUDE.md guidance (see justifications below). ## The main change: Studio CLAUDE.md gets a Code style section **Why:** `studio-best-practices` was a skill that instructed agents to *always* load it before any Studio code work. Anthropic's guidance draws the line as: sometimes-relevant guidance → skill (loaded on demand); always-relevant guidance → CLAUDE.md. A skill that must always load has failed the test for being a skill — it costs a tool-call round trip and, worse, silently does nothing in sessions that forget to load it. Since `apps/studio/CLAUDE.md` is lazy-loaded only when an agent touches Studio files, inlining is properly scoped: non-Studio sessions never pay for it. **Why not verbatim:** the skill was 175 lines, mostly ❌/✅ worked examples teaching practices models already know. Inlining it whole would push the file past the ~200-line point where Anthropic warns rules start getting lost. Instead each section was distilled to the rule it exists to enforce — e.g. the loading/error/success section kept its code block because the *shape* (early returns at top level, flat `&&` chains inline) is the prescription, and prose loses it. **The framing that makes the generic rules earn their place:** models default to matching surrounding code, and not all existing Studio code follows these practices. The section opens with "older Studio code predates some of these conventions — follow them rather than mirroring nearby legacy patterns," which converts otherwise-redundant React advice into an explicit instruction to break from local precedent. One rule was added that the old skill lacked: `useEffect` is for external-system sync only (~364 Studio files contain effects, many in patterns we don't want copied). **Changed:** - `apps/studio/CLAUDE.md` — new Code style section (84 lines total, within budget); skills table no longer mandates a pre-load - `.claude/CLAUDE.md` — dropped `pnpm install` from commands (guessable; Anthropic's test: "would removing this cause mistakes?") **Removed:** - `.claude/skills/studio-best-practices/` — fully absorbed; its cross-references to other skills were already covered by the skills routing table ## Skill description tuning Descriptions are the only signal an agent sees before deciding to load a skill, and the observed failure mode is under-triggering on tasks that don't name the skill. Nine descriptions reworded: front-loaded matchable keywords, added incidental-trigger cases (e.g. a new feature that adds copy is a `copywriting` moment), and disambiguated overlaps (`vitest` is now the API reference deferring to `studio-testing` for strategy). The `safe-sql-execution` rewrite was additionally validated with skill-creator's trigger-eval loop against 20 realistic queries: held-out test accuracy 54% → 71%, with zero false triggers across all iterations. (`vitest` shows under `.agents/` because `.claude/skills/vitest` symlinks there.) ## Generated-files enforcement **Added:** `permissions.deny` rules in `.claude/settings.json` for the six generated-file globs the root CLAUDE.md already lists. CLAUDE.md prose is advisory; permission rules are mechanical and also gate sandboxed Bash writes. (Verified live: the rule blocked an unintended regeneration of `database-types.ts` during testing.) ## To test - CI: prettier + typos checks pass (docs-only + settings change, no app code) - In a fresh Claude Code session in the repo: ask it to edit `apps/studio/routeTree.gen.ts` — should be denied by the new permission rule - Ask it to do any Studio UI task — it should pick up the Code style rules from `apps/studio/CLAUDE.md` without loading a best-practices skill <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * **Documentation** * Updated development guidance for testing, copywriting, SQL safety, telemetry, queries, error handling, and toolbar reviews. * Restructured Vitest references into clearer tables and improved formatting across several guides. * Added Studio code-style conventions and clarified when task-specific guidance should be applied. * Removed outdated Studio best-practices guidance. * **Chores** * Added safeguards preventing edits to generated and protected files. * Simplified the documented development command sequence. <!-- end of auto-generated comment: release notes by coderabbit.ai --> --------- Co-authored-by: Alaister Young <10985857+alaister@users.noreply.github.com> |
||
|
Alaister YoungandAlaister Young
|
2d5ec97df8 |
chore: split CLAUDE.md into root and studio-specific files (#48202)
Splits agent guidance into a lean monorepo-wide root file and a studio-specific file that Claude Code lazy-loads when working under `apps/studio/`. This keeps every session's baseline context small while giving studio work much richer, enforceable guidance. **Changed:** - `.claude/CLAUDE.md` — now monorepo-wide only: corrected pnpm version (10 → 11), expanded workspace table (design-system, ui-library, lite-studio, ui-patterns, api-types, pg-meta, shared-data), commands (`format`, `generate:types`, `api:codegen`), CI gates + never-hand-edit generated files, monorepo-wide conventions (incl. the named-exports rule, which lives in the shared eslint preset and applies to all six apps), and monorepo-wide skill triggers. Studio detail is replaced by a pointer to the nested file. Also corrects a long-standing error inherited from the old file: the `_Shadcn_` convention was inverted — `Button_Shadcn_` is the only suffixed export left and is rarely the right choice; primitives are unsuffixed. - `.claude/skills/studio-ui-patterns/SKILL.md` — removed the same stale `_Shadcn_` claim from the forms section (this skill also feeds CodeRabbit reviews). - `apps/studio/components/README.md` — component template now uses a named export, matching the lint-enforced convention (was the one doc still showing `export default`). - `apps/studio/TANSTACK_MIGRATION.md` — cleanup checklist gains an item to remove the migration section from `apps/studio/CLAUDE.md` when the migration finishes. - `.gitignore` — removed the blanket `CLAUDE.md` ignore rule (added in #40231 for personal local files, no longer used that way). Nested `CLAUDE.md` files are now tracked by default, so shared guidance can't silently fail to land. For *personal* notes, use `CLAUDE.local.md` (Claude Code loads it automatically alongside `CLAUDE.md`, and it's now gitignored here) — or `.git/info/exclude` if you prefer a different filename. **Added:** - `apps/studio/CLAUDE.md` — studio guidance, loaded on demand: mandatory skill routing (always load `studio-best-practices`, plus a task → skill table), TanStack Start migration rules (pages/routes mirroring, when a manual mirror is needed, never delete `pages/**` files), data-layer/state orientation, a default-to-shipping-tests-with-changes policy, and a "defaults that differ here" list (ESLint warning ratchet + local `lint:ratchet` command, `copyToClipboard` await rule, `useParams` from `common`, dayjs/sonner, `ui` vs `ui-patterns` import split, `@tanstack/react-table` over `react-data-grid`, etc.). ## Accuracy Every factual claim in both files (62 total) was verified against the code by parallel review agents instructed to refute each one. Results: 54 correct as written, 2 wrong (the inherited `_Shadcn_` inversion, and a fabricated `useExecuteSqlQuery` hook name — the real export is `useExecuteSqlMutation`), 6 imprecise (e.g. dayjs plugins load in both runtime entries, the ratchet counts occurrences regardless of severity). All fixed in this PR. ## Context cost | File | Size | When it loads | % of a 200k window | |---|---|---|---| | `.claude/CLAUDE.md` | 70 lines, ~1.2k est. tokens | every session | ~0.6% | | `apps/studio/CLAUDE.md` | 53 lines, ~1.6k est. tokens | only when touching studio files | ~0.8% | The always-loaded footprint grew only ~0.2k est. tokens vs the old 45-line file — everything studio-heavy sits behind the lazy load, so docs/www sessions pay nothing for it. Both files are well under Claude Code's large-file warning threshold (~40k chars) and the <200-line adherence guidance, with room to roughly double before it's worth worrying about. ## To test - Open a fresh Claude Code session from the repo root and read any file under `apps/studio/` — `apps/studio/CLAUDE.md` should get pulled into context automatically. - `git check-ignore apps/studio/CLAUDE.md` exits 1 (not ignored); `git check-ignore CLAUDE.local.md` exits 0 (ignored). - Skim both files — every claim has been code-verified (see Accuracy above), but a human sanity pass on the *judgment* calls (what's included/omitted) is welcome. <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * **Documentation** * Refreshed monorepo onboarding conventions with updated tooling requirements, expanded inventory, standardized common scripts, and clearer CI gating and checks. * Added/updated Studio contributor guidance, including the TanStack Start migration rules and Studio development/testing/UI conventions. * Updated Studio component documentation to use named exports. * Refreshed the “Forms” UI pattern guidance and adjusted the referenced UI primitives. * **Chores** * Updated ignore rules so the primary top-level onboarding document is tracked. <!-- end of auto-generated comment: release notes by coderabbit.ai --> --------- Co-authored-by: Alaister Young <10985857+alaister@users.noreply.github.com> |