moteva/task_plan.md

# Task Plan: Server-Backed Brand Kit Canvas Context

## Goal
Persist Brand Kits by authenticated user, bind one optional Brand Kit to each project/canvas, expose a canvas-header selector, and make every Agent Chat turn and newly created thread use the selected kit as authoritative creative context.

## Current Phase
Phase 35

## Phases

### Phase 1: Architecture Discovery
- [x] Map Brand Kit frontend state, project persistence, chat/thread creation, and generation context flow
- [x] Confirm API generation, storage-driver, schema, and test patterns
- **Status:** complete

### Phase 2: API and Domain Design
- [x] Update the `.api` contract before generated code
- [x] Define user-owned Brand Kit storage and project binding invariants
- [x] Define server-side prompt injection behavior for all project chat sessions
- **Status:** complete

### Phase 3: Backend Implementation
- [x] Generate handlers/types from the validated API specification
- [x] Implement memory and PostgreSQL Brand Kit persistence keyed by user ID
- [x] Persist and expose project `brandKitId`
- [x] Inject selected Brand Kit into Agent Chat and new-thread context
- [x] Add backend tests and API documentation
- **Status:** complete

### Phase 4: Frontend Integration
- [x] Replace local-only Brand Kit persistence with authenticated API persistence
- [x] Add canvas-title Brand Kit selector with None and user-owned kits
- [x] Add a Brand Kit selector to the Home prompt composer, including an empty create action
- [x] Keep Brand Kit editor, defaults, and project binding synchronized
- **Status:** complete

### Phase 5: Verification
- [x] Run API validation, generation checks, Go tests/build, frontend typecheck/build
- [ ] Verify CRUD, user isolation, project binding, chat context, and selector behavior
- [ ] Inspect desktop and mobile browser layouts
- **Status:** in_progress

### Phase 6: Delivery
- [ ] Review scope, generated files, and prohibited product references
- [ ] Provide preview URL and concise implementation notes
- **Status:** pending

### Phase 7: Device Management Discovery and Design
- [x] Trace the existing account dialog, authentication gateway, and device API behavior
- [x] Define production-grade loading, empty, error, current-device, and destructive-action states
- **Status:** complete

### Phase 8: Device Management Implementation
- [ ] Implement the device management view and responsive styling in the existing account surface
- [ ] Wire list/remove-all behavior to the authenticated device API without changing unrelated account flows
- [ ] Make desktop/mobile session limits configuration-driven and return the effective policy from the device API
- [ ] Enforce each configured device-type limit atomically and render last-active timestamps in the browser timezone
- [ ] Add focused backend or frontend coverage where the repository supports it
- **Status:** in_progress

### Phase 9: Device Management Verification
- [ ] Run focused tests, frontend typecheck/build, and Go verification if backend code changes
- [ ] Inspect the authenticated desktop and mobile UI in a real browser
- **Status:** pending

### Phase 10: Device Management Delivery
- [ ] Review scope and final diff
- [ ] Provide the working preview URL and concise implementation notes
- **Status:** pending

### Phase 11: Visual Annotation Discovery and Design
- [x] Map image toolbar/action generation, canvas transforms, and reusable panel conventions
- [x] Translate the supplied annotation reference into the repository's interaction and visual language
- **Status:** complete

### Phase 12: Visual Annotation Implementation
- [ ] Add an image-toolbar entry and visual annotation editing mode
- [ ] Support brush, circle, rectangle, and cross marks on the selected image
- [ ] Make each mark's AI edit instruction directly editable and removable
- [ ] Submit the annotated source through the existing image action generation flow
- [ ] Preserve the source node and write generation state/results only to a sibling copy
- **Status:** in_progress

### Phase 13: Visual Annotation Verification
- [ ] Add focused interaction/serialization coverage where supported
- [ ] Run frontend typecheck/build and inspect desktop/mobile behavior in a real browser
- **Status:** pending

### Phase 14: Visual Annotation Delivery
- [ ] Review scope and final diff without altering unrelated user changes
- [ ] Provide the working preview URL and concise implementation notes
- **Status:** pending

### Phase 15: CAD Import Discovery and Design
- [x] Trace the canvas asset import, upload, persistence, rendering, and export paths
- [x] Define the supported DWG/DXF conversion boundary and failure states
- **Status:** complete

### Phase 16: CAD Import Implementation
- [x] Confirm no backend API change is required for browser-side conversion
- [x] Add DWG/DXF selection, conversion, SVG/canvas rendering, and user feedback
- [x] Add focused tests and documentation
- **Status:** complete

### Phase 17: CAD Import Verification and Delivery
- [x] Run relevant frontend tests/builds
- [x] Inspect CAD import and canvas behavior in a browser with the exact R12 file
- [x] Review scope and deliver concise support/limitation notes
- **Status:** complete

### Phase 18: General CAD Compatibility Follow-up
- [x] Reproduce the block-heavy DWG timeout and isolate its dominant output path
- [x] Replace recursive INSERT/DIMENSION expansion with reusable SVG block definitions and full insert transforms
- [x] Cover nested blocks, repeated inserts, MINSERT arrays, malformed references, and recursion cycles
- **Status:** complete

### Phase 19: Vector Image Toolbar Follow-up
- [x] Recognize uploaded SVG and CAD-derived SVG nodes as vector images
- [x] Hide the raster image toolbar and show the compact vector/download toolbar for vector nodes
- [x] Preserve original SVG download behavior
- **Status:** complete

### Phase 20: Follow-up Verification and Delivery
- [x] Verify both user-provided CAD files plus representative DWG/DXF regression fixtures
- [x] Run focused tests, typecheck, and production builds
- [x] Verify vector toolbar, rendering, download, and persistence in a real browser
- **Status:** complete

### Phase 21: Vector Reference Message Rendering
- [x] Reproduce CAD-derived SVG references falling back to raw prompt directives
- [x] Parse SVG/CAD reference names with nested parentheses into the existing message capsule
- [x] Verify SVG directives enter the existing thumbnail/hover/copy component path; run tests and production build
- **Status:** complete

### Phase 22: Model-Compatible SVG References
- [x] Identify every canvas frontend submission path that can pass an SVG image reference to image generation
- [x] Rasterize SVG references to transparent WebP only at the model submission boundary
- [x] Cache and replace submitted reference URLs without mutating SVG canvas nodes or display names
- [x] Add focused regression coverage and verify prepared model content receives a WebP asset URL
- **Status:** complete

### Phase 23: CAD Semantics and Crop Discovery
- [x] Map CAD entity/layer/block/text data through conversion, canvas persistence, references, and Agent Chat generation
- [x] Map the existing crop overlay, sibling-node placement, vector toolbar, and upload/save paths
- [x] Define bounded semantic metadata and crop-output contracts for large drawings
- **Status:** complete

### Phase 24: CAD Semantics and Crop Implementation
- [x] Preserve bounded CAD semantics during DWG/DXF conversion and canvas persistence
- [x] Attach CAD semantic context to model-bound references without exposing it in user-visible chat text
- [x] Add vector crop action that preserves the source and creates a cropped sibling asset
- [x] Make cropped CAD regions attachable and reusable in Agent Chat
- **Status:** complete

### Phase 25: CAD Semantics and Crop Verification
- [x] Add conversion, persistence, semantic-payload, crop, and source-preservation regression coverage
- [x] Verify both supplied CAD files, focused Go/frontend tests, typecheck, and production builds
- [x] Attempt real-browser crop/reference inspection and record the unavailable browser backend
- **Status:** complete

### Phase 26: CAD Semantics and Crop Delivery
- [x] Review scope and generated files without altering unrelated changes
- [x] Document the semantic contract, crop behavior, and preview URL
- **Status:** complete

### Phase 27: Non-Passive Canvas Wheel Regression
- [x] Add a deterministic regression test for non-passive element-level wheel listener registration and cleanup
- [x] Replace the React synthetic canvas wheel handler with a native `{ passive: false }` listener
- [x] Verify zoom behavior, DWG upload flow, focused/full tests, and production build
- **Status:** complete

### Phase 28: Workspace Toolbar Crop Action
- [x] Add a persistent Crop icon immediately after the Text tool
- [x] Enable it only for a selected, ready image/vector node and reuse the existing crop overlay
- [x] Verify icon order, disabled/active states, crop entry, tests, and production build
- **Status:** complete

### Phase 29: Enlarged Single-Image Crop
- [x] Add regression coverage for translating the visible viewport intersection into normalized crop coordinates
- [x] Let workspace Crop automatically target the only visible image when none is selected
- [x] Anchor the crop selection and controls inside the visible viewport for oversized images
- [x] Verify enlarged-image crop entry, multi-image ambiguity, tests, and production build
- **Status:** complete

### Phase 30: Native-Resolution Vector Crop Fidelity
- [x] Reproduce the CAD SVG crop stroke/text degradation with a deterministic regression seam
- [x] Crop SVGs by rewriting their viewBox and render once at the final WebP resolution
- [x] Preserve SVG structure, transparency, CAD semantics, source immutability, and raster crop behavior
- [x] Verify crop output dimensions/display sizing, focused tests, production builds, and browser behavior
- **Status:** complete

### Phase 31: Canvas Performance and LeaferJS Evaluation
- [x] Build a repeatable browser performance baseline for pan, zoom, selection, and large CAD nodes
- [x] Isolate React/DOM reconciliation, SVG rasterization, overlays, persistence, and event-path costs
- [x] Map LeaferJS capabilities and integration constraints against the existing canvas domain model
- [x] Recommend the smallest migration boundary with measurable acceptance criteria
- **Status:** complete

### Phase 32: Leafer Renderer and Vector Texture Cache
- [ ] Extend the API/domain/storage contract with an optional vector render-preview URL
- [ ] Generate and upload transparent WebP previews while preserving original SVG/CAD source and semantics
- [ ] Add a LeaferJS visual renderer adapter for image and manual-frame nodes behind the existing canvas domain
- [ ] Move wheel/pan and node drag hot paths to imperative engine/DOM updates with commit-on-end state synchronization
- [ ] Preserve crop, export, selection, context menus, overlays, Agent Chat references, and persistence behavior
- [ ] Add focused regression/performance coverage and run full frontend/backend/build/browser verification
- **Status:** in_progress

### Phase 33: Workspace Crop Drag Selection
- [x] Enter a crosshair preselection state after activating the workspace Crop tool
- [x] Select a connected color-bounded region on click, capped to the visible image area
- [x] Map a held left-button drag into normalized image coordinates across pan and zoom
- [x] Preserve the existing crop editor, vector semantics, and source-preservation behavior
- [x] Add focused regression coverage and run frontend verification
- **Status:** complete

### Phase 34: Reference-Anchored Generation Placement
- [x] Preserve ordered canvas-node identity from submitted image capsules for the active agent turn
- [x] Place newly generated image nodes beside the last canvas-backed image capsule
- [x] Keep backend/default placement when no submitted capsule maps to a canvas node
- [x] Add multiple-capsule and fallback regression coverage
- [x] Run frontend tests, typecheck, build, and preview verification
- **Status:** complete

### Phase 35: Pen Tool and Shape Path Editing
- [x] Trace the workspace shape tool, canvas-node representation, Leafer renderer, and persistence behavior
- [x] Install and integrate `leafer-x-path-editor` without breaking existing shape creation
- [x] Add a new Pen icon immediately after Shape while preserving the existing Brush tool unchanged
- [x] Enable double-click path editing for pen paths and shapes created by the Shape tool
- [x] Keep legacy/unmarked Brush strokes outside the vertex-editor path contract
- [x] Cover editor lifecycle, persistence, and interaction regressions
- [x] Run frontend tests, typecheck/build, and browser verification
- **Status:** complete

## Invariants
- Brand Kits are owned and queried only through the authenticated user ID from request context.
- A project may reference zero or one Brand Kit owned by the same user.
- Selecting None clears the binding without deleting the Brand Kit.
- Server-side Agent Chat context is derived from the persisted project binding, never trusted from a client-supplied prompt.
- Home may request a Brand Kit ID only from the authenticated user's list; the server revalidates ownership before binding it.
- Deleting a Brand Kit clears project bindings to it.
- Device sessions are always scoped from the authenticated server-side user identity.
- Bulk device removal preserves the current session unless the product contract explicitly requires otherwise.
- New access and refresh tokens carry an opaque random session ID; the server validates that session before accepting the bearer token.
- Legacy access tokens without a session ID are migrated once from a stable token digest after user validation, preserving existing signed-in users.
- Revoked and expired sessions never authenticate or appear in the active-device list.
- Device metadata is derived from request headers for display only and never used as an authorization signal.
- Device timestamps cross the API boundary as RFC3339 UTC instants and are localized only in the browser's timezone.
- Visual annotation actions never mutate the source image node; their copied result node is the only generation target.
- Desktop and mobile web-session limits come from server configuration; the API returns the effective values and the UI never hard-codes them.
- Imported CAD files are read without mutation; only a sanitized derived SVG is uploaded and rendered, with explicit size, timeout, empty-drawing, and conversion errors.
- Repeated and nested CAD block references are represented once and instanced with SVG references; conversion cost must scale with unique block geometry, not insert count.
- SVG-backed image nodes use vector-only actions and download their source SVG rather than exposing raster editing actions.
- User-message image directives render through one attachment capsule/preview path regardless of raster or SVG source format.
- SVG remains the persisted canvas source; transparent WebP derivatives exist only for model-compatible image inputs.
- CAD source nodes are immutable during crop; every crop is a sibling asset positioned beside the source.
- Pen and Brush remain distinct workspace tools: Pen creates vertex-editable paths, while Brush keeps the existing continuous freehand behavior.
- CAD semantic context is bounded, persisted with the relevant source/crop, and sent as model context without becoming visible user-message copy.

## Errors Encountered
| Error | Attempt | Resolution |
|-------|---------|------------|
| zsh rejected an unmatched `HomePage/*.css` glob | 1 | Replaced the glob with explicit files discovered via `rg --files`. |
| Assumed `CanvasWorkspace/index.css`, which does not exist | 1 | Locate the workspace stylesheet from its imports before reading it. |
| `rg` treated a pattern beginning with `-- name:` as a flag | 1 | Use `rg -- <pattern>` when searching SQLC query annotations. |
| Assumed `AuthProvider.tsx` instead of the component directory entry | 1 | Locate the file with `rg --files` before any further auth-provider reads. |
| PostgreSQL package test failed after adding project mapping because text conversion helpers were not yet defined | 1 | Add nullable `pgtype.Text` conversion helpers and rerun the package tests. |
| One patch to remove a temporary `err = nil` line used the wrong surrounding field name | 1 | Reapplied the patch against the exact current lines; no code change was lost. |
| Workspace import patch assumed a nonexistent `BrandMark` import | 1 | Located the actual import block and applied the selector change against current source. |
| First i18n patch used an inexact Chinese source string | 1 | Re-read the locale file and patched the exact key/value context. |
| Existing account-device test called removal without an authenticated session context | 1 | Update the test to resolve the issued bearer identity and assert the new current-session contract. |
| Preview backend rejected `/dev/stdin` because go-zero infers config format from the filename extension | 1 | Use a `.yaml` symlink to stdin so the transformed preview config retains a recognized extension without creating a repository config file. |
| Path editor typecheck treated the asynchronously initialized Leafer App as nullable | 1 | Use a non-null local `nextApp` after construction while retaining a nullable cleanup reference. |
| Node's TypeScript test loader could not resolve the extensionless `shapeNode` import from `editablePath.ts` | 1 | Use the explicit `.ts` extension, which is supported by the no-emit bundler configuration and Node test runner. |
| TypeScript rejected the explicit `.ts` import until the no-emit project opted in | 1 | Enable `allowImportingTsExtensions`, matching the Node test loader and bundler-only output model. |
| Pen-path test expected unchanged local coordinates after a control point expanded the node bounds | 1 | Keep the world-coordinate-preserving implementation and use an in-bounds control point for the write-back assertion. |
| Initially repurposed the existing freehand Brush button as Pen | 1 | Restore Brush unchanged and add Pen as a separate tool immediately after Shape. |
| A shell health-check used zsh's read-only `status` variable | 1 | Rename the variable to `http_status`; subsequent endpoint checks returned 200. |
| Next standalone preview was first launched with `next start`, then without copied static assets | 1 | Use the already-running project dev server for the final preview and Vite for isolated mocked browser verification. |
| A second Next development server refused to start because the existing repo-wide dev lock is active | 1 | Use the already-built Next production server on the alternate preview port. |
| Final frontend build found another active `next build` process | 1 | Inspect the active process and wait for it to finish before retrying once; do not remove build locks blindly. |
| Path-tool editor branch was first patched into the duplicate background handler | 1 | TypeScript caught undefined `node`; restore the background branch and scope the node regression to `handleNodePointerDown`. |
| Anchor-Pen follow-up again matched the background Path branch | 1 | Scoped regression and TypeScript caught it; patch using the node handler's unique text-node condition. |
| Browser kept an obsolete Pen module after concurrent exports changed | 1 | Vite reported Fast Refresh invalidation; perform a full page reload after typecheck/build settles before re-verifying. |
| BSD `sed` did not support the GNU `0,/pattern/` address used for the preview port substitution | 1 | Replace it with a portable anchored substitution; the backend now listens on 8889. |
| The in-app browser runtime reported no available browser backends | 1 | Keep the verified local preview running and report the browser-only visual inspection gap; do not substitute an unrelated browser surface against the browser skill rules. |
| Escaping the `psql \d` meta-command through Docker produced an invalid command | 1 | Query `information_schema.columns` and `pg_indexes` with regular SQL instead. |
| Desktop limit regression test observed two valid desktop sessions when the configured limit was 1 | 1 | Confirmed the limit was presentation-only; enforce it atomically during session creation and reconcile pre-fix sessions during device listing. |
| Launching `mcp-zero` from `server/` used a repository-root relative path | 1 | Relaunch with `../.tools/bin/mcp-zero` while keeping generation output rooted in `server/`. |
| NPM lookup for a nonexistent `@loaders.gl/dxf` package returned 404 and stopped the chained lookup | 1 | Use independent package searches and validate the chosen converter package before installation. |
| Package inspection commands raced the tar extraction and looked under the wrong transient path | 1 | Inspect the extracted tree first, then run dependent file reads sequentially against the discovered path. |
| `npm install` inherited a parent workspace containing an unsupported `workspace:*` URL | 1 | Inspect local package-manager metadata and run installation with workspace discovery disabled or use the repository's actual package manager. |
| TypeScript rejected `.length` on the CAD library's entity collection | 1 | Use the collection's declared count/iteration API instead of assuming it is an array. |
| Next 16 Turbopack emitted the module worker as raw `.ts` instead of executable bundled JavaScript | 1 | Run Next dev/build through its supported webpack mode, which handles `new Worker(new URL(...))`; retain Vite's native worker bundling. |
| Used nonexistent Playwright CLI command `network` | 1 | Use the CLI's documented `requests` and `response-body` commands for network inspection. |
| First Vite DXF upload triggered dependency pre-bundling and a development-page reload | 1 | Retry after Vite finishes one-time dependency optimization; production worker output is already bundled. |
| Node's strip-only TypeScript test loader rejected a constructor parameter property in the Leafer adapter | 1 | Replace the parameter property with an explicit class field so production TypeScript and the repository's Node test runner share supported syntax. |
| The Leafer descriptor test could not resolve frontend `@/` runtime aliases under Node's native loader | 1 | Keep the descriptor dependency-free at runtime with local source/adjustment checks; type-only aliases remain erased and production bundling is unchanged. |
| DOM `window.setTimeout` returned `number` but the viewport timer ref inferred Node's `Timeout` type | 1 | Declare the browser-owned timer ref as `number | null`, matching the existing browser timeout usage. |
| Combined render-preview cleanup/thumbnail patch assumed `mainCanvasPreviewImage` lived in `project_thumbnail.go` | 1 | Locate the function in `design_service.go` and split the patch across exact current file contexts. |
| The first crop preselection integration patch used one stale keyboard-handler context and failed atomically | 1 | Split the integration into small patches against the exact current source; no production file received a partial edit. |
| A diagnostic `rg` command contained an unmatched quote around a template-literal search | 1 | Use separate fixed-string searches without shell-sensitive backticks. |
| Real upstream DXF fixture parsed to an empty model space in `@node-projects/acad-ts` | 1 | Validate the library's document assembly in Node, then replace the DXF path or reader entirely if real fixtures remain empty. |
| Directory-wide listing of the WeChat file folder stalled on container-managed storage | 1 | Use the exact user-provided path directly; metadata access confirms the file is readable. |
| The exact R12 DXF crashed while serializing the first inherited-color POINT | 1 | Resolve non-renderable ByLayer/ByBlock colors through the entity and layer before calling the third-party SVG writer; lock the case with a minimal R12 regression test. |
| Next production minification reduced valid DXF geometry to zero-valued entities | 1 | Restore `acad-ts` constructor names from stable export keys before its name-based metadata lookup; keep the normal minified build and cover the mangled-name path in the regression test. |
| Reusing one entity writer and slicing its cumulative string exhausted the Node heap on the exact R12 drawing | 1 | Give each entity a short-lived fragment writer so output assembly stays linear in total geometry size. |
| Node's strip-only TypeScript test runner rejected constructor parameter properties | 1 | Use explicit class fields and constructor assignments, which preserve the project test runner's supported syntax subset. |
| Vector reference regression suite failed on nested English and Chinese CAD filenames | 1 | Expected red phase confirmed: the directive regex stops before the final filename parenthesis; update it to bind the closing delimiter adjacent to the URL separator. |
| First post-fix parser assertion still expected no `index` property | 1 | The production parser correctly emits `index: undefined`; align the regression fixture with the stable result shape. |
| In-app browser discovery returned no available browser backend | 1 | Followed the browser skill troubleshooting flow and confirmed an empty backend list; retain automated tests/builds and report the interactive verification gap. |
| Combined final planning patch targeted progress lines as task-plan context | 1 | Split the update by file and apply each change against its exact current section. |
| Local `mcp-zero` stdio initialization returned no JSON-RPC response | 1 | API validation succeeded; use the documented `goctl api go` fallback with the existing `go_zero` style. |
| TypeScript did not infer that a crop existed from the derived crop-bounds check | 1 | Guard the scope line with both `crop` and `cropBounds`; Go targeted tests already pass. |
| One combined patch contained a malformed hunk boundary | 1 | Reissued the same two-file edit with valid patch boundaries; no source change was lost. |
| Node's strip-types test runner could not resolve the new extensionless runtime import | 1 | Keep the shared semantic types as a stripped type-only import and define the stable metadata element ID locally in the worker module. |
| Synthetic text-only CAD fixture did not expose a library bounding box | 1 | Add a simple visible line to establish drawing bounds while keeping the text-encoding assertion focused. |
| A second planning/error patch initially repeated the malformed hunk separator | 1 | Removed the empty hunk marker and applied the intended fixture/error update cleanly. |
| Isolated Next dev preview was rejected by an existing repository-wide dev lock | 1 | Keep the updated backend on 8889 and run the verified production Next build on 5175 with its API proxy pointed at 8889. |
| zsh rejected the optional `frontend/.env*` inspection glob because no env file exists | 1 | Search the concrete Next/Vite config and source paths without an unmatched optional glob. |
| Combined toolbar patch expected the wrong `onChooseShapeKind` callback name | 1 | Re-read the exact `WorkspaceToolbar` prop block and apply the component and call-site changes against current code. |
| First browser wheel probe stayed at 100% because the stage listener was never attached | 1 | The initial effect ran while the project-loading view had no `.canvas-stage`; re-run listener binding when `project.id` mounts the canvas. |
| Playwright `run-code` rejected a bare `await page...` snippet despite the abbreviated skill example | 1 | Read command help and pass an `async (page) => { ... }` function as required by the installed CLI. |
| The first mocked route handler used `URL`, which is unavailable in the Playwright callback sandbox | 1 | Derive the pathname from `route.request().url()` with bounded string operations; the corrected route exposed the real canvas. |
| Initial combined Phase 30 test/findings patch assumed a generic `# Findings` heading | 1 | Re-read the file header and apply the regression test and findings as separate patches against the exact heading. |
| Exact-DWG performance fixture served the raw converter intermediate and Chromium rejected anonymous `</>` tags | 1 | Use the real browser upload path so `normalizeCadSvg` repairs and sanitizes the final SVG before measuring paint/interaction cost. |
| Playwright file-input locator matched both canvas upload and Agent Chat upload controls | 1 | Target the first canvas file input explicitly for the real DWG upload probe. |
| Playwright request context bypassed the page's mocked asset route and could not resolve `asset.local` | 1 | Fetch the captured normalized SVG inside the page context so the route handler supplies the bytes. |
| Combined performance findings patch matched entries in a different order than the current file | 1 | Re-read the exact section and apply the package/result additions against their real positions. |
| Direct CDN import used LeaferJS's package ESM build, whose bare `@leafer/core` specifier is not browser-resolvable | 1 | Load the official prebundled browser distribution for the throwaway benchmark. |
| TypeScript widened the hidden reference-anchor content discriminator to `string` | 1 | Declare the helper's `AgentContent[]` return type explicitly and rerun the focused test and typecheck. |
| Detached `nohup` preview processes were reclaimed by the terminal runner | 1 | Start the frontend and backend in persistent tool sessions, then verify all direct and proxied endpoints. |
| `pnpm dev -- --port 5175` passed `--port` as a Next project directory | 1 | Use the repository script's supported `pnpm dev --port 5175` argument form. |