feat(usage): include bounded local OpenCode history

[rynfar]

OpenCode task-host usage history was absent from Pylon's usage view, even though local SQLite and legacy JSON stores can hold project-level token and cost records. This adds a bounded local reader and merges its records into Pylon's existing usage cache and v8 contract while retaining v7 source attribution and existing Antigravity behavior. External OpenCode serverUrl instances remain outside this local history scan; Cursor account-wide history remains a separate maintainer decision.

The reader caps root entries, files, bytes, nested entries, and records; reports partial coverage for limits or malformed source data; and does not cache corrupt files as complete. Canonical roots and record fingerprints prevent duplicate counting. The contract retains compatibility with v7 summaries.

Verification: focused reader and shared merge tests 25/25; targeted service integration 1/1, covering SQLite/legacy JSON, duplicate roots, remote exclusion, warm/restart cache and partial status; final reader regression tests 6/6; scoped typechecks for server, contracts, shared, client-runtime, web and mobile passed; scoped lint, format and diff checks passed. An independent pre-PR adversarial review rechecked the final head and reported no high-confidence issue. The final review revision changed only the reader and its tests, so the slower service test was not repeated after that revision.

Related: #865; frozen T3 source e5a46d6c5d00b89afba5274a94d42428c8d79763 (#10409).

Model: GPT-6 Sol (low), Pylon Codex harness.

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated
pylon-marketing	Ignored	Preview	Sep 27, 2026 11:48am UTC

[rynfar]

UI evidence attempt (isolated state): started this PR head with a new task-owned base directory, opened a fresh collaborative-browser tab using its one-use pairing URL, and attempted a snapshot. The preview returned PreviewAutomationExecutionError; the Vite server reported HTTP 431 (request header fields too large). No before/after image could be captured. I stopped only the task-owned dev server. The focused service and client typechecks remain the UI contract verification; this is an evidence gap, not a claim that the browser flow passed.

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	14.0 KiB	14.0 KiB	+20 B (+0.1%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.2 KiB	7.3 KiB	+13 B (+0.2%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.7 KiB	6.7 KiB	+7 B (+0.1%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	58.0 KiB	58.0 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	9	9	0 (0.0%)	21	✅
Claude	Total thread wire	14.0 KiB	14.0 KiB	+17 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.2 KiB	7.2 KiB	−1 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.7 KiB	6.7 KiB	+18 B (+0.3%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	58.8 KiB	58.9 KiB	+44 B (+0.1%)	66.4 KiB	✅
Claude	Live turn messages	8	9	+1 (+12.5%)	21	✅

Baseline: 73e5205 · PR result: 176a157 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 115.7 KiB
• Claude decoded thread snapshot: 116.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.