bitmosh.dev/Blog/Dev Log/2026-06-09
v111.5lumaweave

v111 arc close — semver 0.17.0 → 0.18.0

Closes the v111 test-infrastructure arc. Bumps semver, reconciles NOW.md and ROADMAP with honest v111 history (including the pulled CI E2E work), and appends 6 KNOWN_SHARP_EDGES entries from v111's learnings.

v111.4flumaweave

Pull CI E2E wiring, defer to post-v1.0 dedicated arc

Reverted ci.yml and playwright.config.ts to pre-v111.4 state after 6 failed amendments. Documented the investigation and proposed real-fix path in a known-bug entry.

v111.4elumaweave

CI E2E: switch to Vite production-preview server

Replaced webServer command in playwright.config.ts with a CI-conditional that builds a production bundle then serves via vite preview, eliminating per-request JIT compilation causing 30-60s page loads.

v111.4lumaweave

CI E2E wiring — required job, Playwright cache action

Adds e2e job to ci.yml gated on lint-css + typecheck; uses microsoft/playwright-github-action@v1 for browser caching; --workers=4; timeout-minutes=15. E2E now required for merge.

v111.3lumaweave

Flaker triage — color-tab to web-first, 3 documented

Converts all 5 waitForTimeout calls in color-tab-functional.spec.ts to web-first assertions (stable triple-run); adds audit-marker comments to settings-panel, graph-sources, and gwells C9.0 (architectural deferral).

v0.1.3cerebra

Bench v0.2 LoRA track; confirm Granite 4.1 3B instruct as v0.1.x classifier

Benching the v0.2 LoRA training track after Phase 3 audit revealed corpus imbalance as the resume-blocker (10 RELATIONAL training records covering 2 of 4 sub-categories). Granite 4.1 3B instruct via Ollama stays as production; calibration confirmed at 65% partial-credit, unchanged from v0.1.0.

v0.1.2cerebra

v0.2 LoRA training infrastructure — corpus, trainer, evaluator

Built and validated the complete v0.2 QLoRA training pipeline. Phase 1 diagnosed the format mismatch root cause (no EOS, no completion-only loss, key name bug); Phase 2 fixed all three and confirmed 0 parse failures. Adapters are not shipping; infrastructure is committed for future use.

v111.2lumaweave

qa.ts waitForTimeout → web-first assertions

Converts two race-condition Band-Aids in qa.ts (openAdvisoryTab + expandSection) to web-first assertions; contract-registry timeout flake resolved, stable across 3 consecutive runs.

v0.1.1cerebra

Post-v0.1.0 architecture brainstorm — counsel/swarm, topology, fingerprinting

Captured the design thinking session that followed v0.1.0 close-out. Updated counsel/swarm cognition and two reframe docs; added conceptual topology probing and model disposition fingerprinting as new architecture concepts. Archived nine completed workflow prompts.

v111.1lumaweave

GVI split — monolith into 3 parallel spec files

Splits 233-test graph-visual-inventory.spec.ts into Core/Probes/Theme for Playwright worker parallelism; wall-clock drops from ~8.7 min to 2m58s.

v110.3lumaweave

v110 arc close — semver 0.17.0, NOW.md + ROADMAP, sharp-edges

Formal close of v110 arc: semver 0.16.0→0.17.0, NOW.md reconcile, ROADMAP v110 LANDED annotation, KNOWN_SHARP_EDGES +3 entries.

v110.2lumaweave

Real-source mode — unified testid + panel overflow fix

Unified graph-viewport testid across overlay + AppShell, resolving real-source inspector positioning; 2 skipped E2E re-enabled and passing.

v110.1lumaweave

Hygiene cluster — identity rename + ErrorBoundary + StatusCluster cleanup

Renamed starmap→LumaWeave in tauri.conf.json + 5 TS files, added root-scope ErrorBoundary, removed hardcoded "settling" layout state.