feat: typegen concurrency

[atilafassina]

Run all queries concurrently.

In Dev Playground app (9 queries), this speeds up typegen from ~93 seconds to ~3 seconds
That's about ~30x faster

Before

[appkit:server] Mode: development (Vite HMR)
  appkit:server Remote tunnel: blocked; inactive +0ms
  Describing 9 queries (1/9)       ✓ app_activity_heatmap
  Describing 9 queries (2/9)..     ✓ apps_list
  Describing 9 queries (3/9).      ✓ cost_recommendations
  Describing 9 queries (4/9)...    ✓ example
  Describing 9 queries (5/9).      ✓ spend_data
  Describing 9 queries (6/9)...    ✓ spend_summary
  Describing 9 queries (7/9).      ✓ sql_helpers_test
  Describing 9 queries (8/9)       ✓ top_contributors
  Describing 9 queries (9/9)..     ✓ untagged_apps
  Described 9 queries (9/9)
  appkit:type-generator:query-registry 9 new queries, 0 cached, 0 failed. Typegen: 93.28s +93.3s

Now

[appkit:server] Mode: development (Vite HMR)
  appkit:server Remote tunnel: blocked; inactive +0ms
  Describing 9 queries (5/9)..     ✓ app_activity_heatmap
    ✓ apps_list
    ✓ cost_recommendations
    ✓ example
    ✓ spend_data
    ✓ spend_summary
    ✓ sql_helpers_test
    ✓ top_contributors
    ✓ untagged_apps
  ✓ Described 9 queries
  appkit:type-generator:query-registry 9 new queries, 0 cached. Typegen: 3.02s +3.0s

Improvements breakdown

Reducing network latency

• Run queries in batches of 20 (avoids rate-limiting and sets good and safe concurrency out-of-the-box)

Reduce FS io

• Read all query files concurrently from system to prepare query batch
• Batches cache writing per query batch
• Debounce writing to the terminal (starting/stop spinner for cached queries are a waste of io - 2 per query)

Bonus
Add a perf measurement on debug to summarize cache HIT/MISS and time spent querying.

[Copilot]

Pull request overview

This PR improves AppKit’s type generator performance by parallelizing query schema discovery (DESCRIBE QUERY) and moving filesystem/cache operations to async APIs, reducing end-to-end type generation time in dev.

Changes:

• Add concurrent, batched DESCRIBE execution with cache writes per batch and consolidated spinner output.
• Switch cache, query reading, and output writing to node:fs/promises.
• Extend CLI spinner with an update() method and update unit tests for the new async/concurrent behavior.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
packages/appkit/src/type-generator/vite-plugin.ts	Replace fs.existsSync usage with existsSync import.
packages/appkit/src/type-generator/tests/generate-queries.test.ts	Update tests/mocks for async fs, spinner updates, and batching behavior.
packages/appkit/src/type-generator/spinner.ts	Add Spinner.update() for progress text updates during concurrent runs.
packages/appkit/src/type-generator/query-registry.ts	Implement parallel query file reads and batched concurrent DESCRIBE execution with cache updates.
packages/appkit/src/type-generator/index.ts	Use async writeFile for generated type declarations.
packages/appkit/src/type-generator/cache.ts	Convert cache load/save to async and use node:fs/promises.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.