Load less JavaScript, cache shared reads, and bound server memory - #100
Merged
Merged
Conversation
Checkoff, Production, and Plan imported three.js statically through the part thumbnail generator and the plate 3D preview, and jszip through the sliced-file and xlsx parsers, so every visit to those pages downloaded both. Thumbnail generation, the plate preview, and both zip readers now load on demand; the plate preview shows its loading box meanwhile. JavaScript each route loads before rendering (uncompressed, gzip): Checkoff +884 KB / 246 KB -> +258 KB / 84 KB Production +842 KB / 230 KB -> +200 KB / 65 KB Plan +679 KB / 183 KB -> +143 KB / 48 KB Printers +179 KB / 59 KB -> +86 KB / 32 KB scripts/route-bundles.mjs prints that table from the Vite manifest and fails when three.js reaches any route's static import closure; npm run quality runs it after the build. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…anifest routes to 8 MiB Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ion bodies at 96 MiB Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The web renderer produces 256 by 256 PNGs, well under 1 MiB, so the route no longer buffers up to 64 MiB per request. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…them in memory Source ZIP and file uploads now stream into a staging directory beside the Source directories while the byte budget is enforced, then move into place. ZIP extraction reads the archive from disk in chunks. MCP file imports use the same staging path. A 96 MiB upload-zip request peaked at about 200 MiB of heap plus external memory before and about 35 MiB after. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…l document Source uploads no longer load each nested 3MF package whole. The reader streams the package from disk, refuses an oversized model document from its declared size, and stops once the model document is complete. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… copying it Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The file picker saves one rule per selected file and the route has no rule-count cap, so a large Source could exceed the new 1 MiB default. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Printers, Settings, Checkoff, and Production screens each fetched the printer fleet and host integrations on their own. They now read shared printers and integrations queries, and fleet writes in Settings update or invalidate those caches. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Settings printer card fetched plans on its own and the unmatched print card kept a fallback fetch that no screen used. The printer card now reads the profiles query, and the unmatched print card takes plans from its parent. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The role color picker, part color dialog, Settings printer card, and Review sheet each fetched role filaments, the filament catalog, or Spoolman spools on their own. They now read shared queries. Custom filament and Spoolman setting writes invalidate the catalog, and a Plan rebuild invalidates that Build's role filaments. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each printer's slicer profile section fetched the same option lists, so Settings made one request per printer. The sections now share one query. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
All Production fetched printer checkoff links and unmatched prints by hand. It now reads them through queries, and the Printers page shares the same checkoff links cache. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… events Printer reconcile results, claimed unmatched prints, and Plan workspace refreshes reloaded the summary of every Build. The events already carry the Build id, so All Production and the Plan workspace now refetch that one Build's summary. Unmatched-file and dismiss events refresh farm state without touching Build summaries, and one reconcile notifies each affected Build once. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Status and reconcile polls ran at the configured cadence even when a host kept failing. Each consecutive offline answer now doubles that host's poll interval up to five minutes, and the first answer from the host restores the configured cadence. Healthy hosts keep their cadence, and a host never starts a poll while its previous one is running. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
reloadProfiles invalidated the profiles query, which refetched it, and then refetched it again. It now refetches once. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A five-minute cap left a printer that was switched back on showing offline for up to five minutes. At the default five-second cadence an unreachable host is now probed at 10, 20, 40, then every 60 seconds. The backoff tests run at that default cadence. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The S3 and local storage ports were constructed but never read, so every blob went to local disk while /health reported artifact_store "s3". Remove the ports, S3 config, the @aws-sdk/client-s3 dependency, the RustFS development service, and the S3 deployment documentation. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…eir migrations Print Partner v3.3.0 ships schema 31, so it becomes the oldest release this one upgrades in place. SQLite and Postgres startup now refuse an older stored schema before touching the database and tell the operator to step through v3.3.0 first. The version-gated repairs below 31 (accepted revision and required-unit backfills, the v26 compatibility cutover, legacy print Plan removal, the filament and plate-edit repairs, the v17 Plan-input rebuild) and their tests are gone. Fresh installs build the same schema as before. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Navigate logical layers of code changes, visualize relationships, and explore their blast radius. Note Currently processing new changes in this PR. This may take a few minutes, please wait... ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Team Run ID: ⛔ Files ignored due to path filters (1)
📒 Files selected for processing (116)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Comment |
Contributor
Bugbot couldn't run - usage limit reachedBugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit. A user or team admin can review and increase usage limits in the Cursor dashboard. (requestId: serverGenReqId_249b912c-34cd-4146-84fb-3fca81321abd) |
Owner
Author
|
Independent pre-merge verification: PASS+NOTES for An independent Codex agent that did not author this PR inspected shared reads, invalidation, polling, dynamic imports, disk-backed uploads, archive boundaries and upgrade guards. All 78 fresh focused tests passed. A real 96 MiB ZIP upload succeeded with 50.1 MiB peak heap-plus-external growth. The existing built bundle passed its import-closure check with no static Three.js route imports. No concrete merge blocker was found. Limits: tests used integrated stack head |
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.
Summary
Performance and scaling work, stacked on #99. Pages load less JavaScript, repeat fewer requests, and the server holds less memory under uploads. It also removes S3 support that nothing used and sets v3.3.0 as the oldest version you can upgrade from.
npm run qualitypasses.What changes for users
JavaScript a page downloads before it renders:
scripts/route-bundles.mjsprints this table from the Vite manifest and fails if three.js reaches any page's initial load.npm run qualityruns it after the build./plansrequest instead of two.src/queries/readRequests.test.tsxcounts them.Server
src/upload-memory.test.ts).S3_BUCKETmade/healthreportartifactStore: "s3"while every blob stayed on local disk.@aws-sdk/client-s3is gone, along with the RustFS service indocker-compose.saas.yml.Upgrade floor (breaking for very old installs)
A database below schema 31 (v3.3.0, the only tagged release) is refused before any backup or DDL, with instructions to install v3.3.0, start it once, then upgrade. This removed about 3,200 lines of pre-31 migrations and repairs. A v3.3.0 database upgraded by the old and the new code produces identical SQLite
sqlite_masterand rows, and identical Postgres catalogs and rows (checked in PGlite). CHANGELOG and OPERATIONS describe the step-through upgrade.Verification
lint, knip, typecheck,
npm test(contracts 101, domain 141, web 1320, server 2131), workflow smoke, build, route bundle gate, runtime and browser tests all pass.Not included
Moving the contracts to
zod/miniis onperf/zod-miniand held back. It would save about 14 KB gzipped per visit, but zod/mini overflows the stack when an object field collects more than about 120k issues, turning some malformed requests from 400 into 500, and it parses large Plan saves about twice as slowly on the server because it skips classic's compiled object path. The overflow comes fromissues.push(...)spreads in zod's shared core, still present in zod 4.6.5, and classic hits it too for nested arrays, records of arrays and tuples. Revisit once zod fixes that upstream.🤖 Generated with Claude Code
Note
High Risk
Breaking upgrade floor and removed pre-31 migration code can brick old databases; upload limit changes may reject previously accepted large MCP/JSON payloads.
Overview
This release removes unused S3 blob storage end to end:
@aws-sdk/client-s3,StoragePortwiring in self-host and SaaS adapters,S3_*config, RustFS fromdocker-compose.saas.yml, and related deploy/docs. Artifacts stay on local disk under the data directory.Upgrade policy is stricter. Startup now refuses SQLite or Postgres below schema 31 (v3.3.0) via
upgrade-guardbefore migrations run, and drops thousands of lines of incremental migration/repair paths (compatibility dirty repair, legacy print-plan removal, v19–v31 cutover modules). Operators must run v3.3.0 once, then this release. CHANGELOG and OPERATIONS document the step-through path.Request and upload limits are retuned in
upload-limits.tsand docs: default JSON 1 MiB, bulk Plan/manifest routes 8 MiB, assistant/MCP bodies 96 MiB, MCP inline files 64 MiB (down from 256 MiB forpropose_import_source_files). Source uploads stream to disk; MCP imports stage to temp files; 3MF parsing uses chunked reads (byte-chunks.ts). SECURITY and API docs describe what still buffers in memory vs streams.Tests and Postgres DDL are aligned with the slimmer migration story (fewer legacy upgrade scenarios, consolidated
plan_revision_inputsshape in fresh migrations).Reviewed by Cursor Bugbot for commit cc907fc. Bugbot is set up for automated code reviews on this repo. Configure here.