feat(commands): add @gittensory chat <question> — grounded LLM Q&A via local Ollama

[JSONbored]

Summary

Adds a new @gittensory chat <question> command that answers a contributor's real question in natural prose, sourced from the same cached decision-pack facts ask/preflight already read — but generated entirely by local Ollama (env.AI_ADVISORY), never the frontier chain.

• New generation service (src/services/ai-chat-qa.ts): enable-flag → shared daily neuron-budget gate → provider call → output-safety check, mirroring ai-summaries.ts's shape. Redacts private decision-pack terms before they ever reach the prompt.
• New chatQa capability flag on advisoryAiRouting (default false), threaded through the type, both normalizers (+ engine-package twin), the focus-manifest merge, OpenAPI schema, and both yml config examples. Unlike the 4 existing capabilities, this one never falls back to the frontier chain — it declines instead.
• Registered in DEFAULT_COMMAND_AUTHORIZATION_POLICY as ["maintainer", "collaborator"] (narrower than ask's default — excludes confirmed_miner for v1), and in AI_COST_BEARING_COMMANDS for the tighter per-command rate limit.
• New Q&A-catalog-only command, structurally isolated from the write-capable action catalog. A static import-isolation test asserts the generation module can never reach a write/action handler or an undeclared DB-mutation helper.
• Fixed, non-LLM disclaimer footer on every answer card: read-only, cannot change review outcomes or gate state.
• The model's own generated text is markdown-neutralized the same way ask's question text already is (neutralizePublicMarkdownText + sanitizePublicComment), since this is the first surface where model output — not just user input — is echoed into a trusted bot comment.

Closes #4595
Part of #4594

Test plan

• Unit tests for every ChatQaResult status (disabled/unavailable/declined ×3/quota_exceeded/unsafe/ok/error) and the pure grounding-redaction helpers (test/unit/ai-chat-qa.test.ts)
• Import-isolation architecture test (test/unit/ai-chat-qa-import-isolation.test.ts)
• Parser, authorization default, rate-limit-set membership, and full answer-card rendering incl. the injection-neutralization regression (test/unit/github-commands.test.ts)
• Full end-to-end webhook dispatch test proving processors.ts actually reaches generateChatQaAnswer for a real chat command (test/unit/queue-5.test.ts)
• Capability-flag normalization + focus-manifest merge tests (both app and engine-package copies)
• npm run typecheck, npm run test:coverage (unsharded), npm run test:ci, npm audit --audit-level=moderate all green
• npm run ui:openapi and npm run command-reference regenerated and committed
• npx tsx scripts/check-engine-parity.ts passes

[superagent-security]

Superagent didn't find any vulnerabilities or security issues in this PR.

[cloudflare-workers-and-pages]

Deploying with    Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	gittensory-ui	767f73c	Commit Preview URL Branch Preview URL	Jul 11 2026, 07:48 AM

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.33%. Comparing base (f175e0c) to head (767f73c).
⚠️ Report is 1 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #4985      +/-   ##
==========================================
+ Coverage   94.31%   94.33%   +0.01%     
==========================================
  Files         455      456       +1     
  Lines       39083    39185     +102     
  Branches    14256    14288      +32     
==========================================
+ Hits        36862    36965     +103     
  Misses       1572     1572              
+ Partials      649      648       -1     

Flag	Coverage Δ
shard-1	46.82% <33.98%> (-0.04%)	⬇️
shard-2	33.41% <30.09%> (+0.05%)	⬆️
shard-3	31.50% <6.79%> (+0.37%)	⬆️
shard-4	32.72% <69.90%> (-0.28%)	⬇️
shard-5	33.56% <6.79%> (-0.07%)	⬇️
shard-6	45.16% <8.73%> (-0.10%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
packages/gittensory-engine/src/focus-manifest.ts	99.17% <100.00%> (+<0.01%)	⬆️
...ry-engine/src/review/advisory-ai-routing-config.ts	100.00% <ø> (ø)
...nsory-engine/src/settings/command-authorization.ts	100.00% <ø> (ø)
src/github/commands.ts	98.27% <100.00%> (+0.09%)	⬆️
src/openapi/schemas.ts	100.00% <ø> (ø)
src/queue/processors.ts	95.30% <100.00%> (+0.03%)	⬆️
src/review/advisory-ai-routing-config.ts	100.00% <ø> (ø)
src/services/ai-chat-qa.ts	100.00% <100.00%> (ø)
src/settings/command-authorization.ts	100.00% <ø> (ø)
src/signals/focus-manifest.ts	99.64% <100.00%> (+<0.01%)	⬆️
... and 1 more

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[loopover-orb]

Warning

🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨

⏸️ Gittensory review result - manual review recommended

_{Review updated: 2026-07-11 07:51:06 UTC}

27 files · 1 AI reviewer · no blockers · readiness 100/100 · CI green · clean

⏸️ Suggested Action - Manual Review

• Touches a guarded path — held for manual review: This PR changes guardrail-protected path(s): src/queue/processors.ts (matched src/queue/**), src/services/ai-chat-qa.ts (matched src/services/**).

Review summary
Adds a self-contained `@​gittensory chat <question>` capability generated exclusively via local Ollama (env.AI_ADVISORY), with a hard decline-not-fallback design that is distinct from the four existing advisoryAiRouting capabilities, and threads the new `chatQa` flag through every required site (types, both normalizers/twins, focus-manifest merge, OpenAPI schema+doc, yml examples, command-authorization defaults, AI_COST_BEARING_COMMANDS). The generation module is import-isolated from the write/action path and DB-mutation helpers, verified by a dedicated static-analysis test, and both the contributor question and the model's own output are run through the same sanitizePublicComment/neutralizePublicMarkdownText pipeline `ask` already uses, plus a pre-prompt regex redaction and a post-generation forbidden-term check before anything model-authored reaches the public card. Tests cover every ChatQaResult status, the redaction/compaction helpers, focus-manifest override precedence, command authorization defaults, and an end-to-end queue dispatch for both the enabled and default-off paths — this is a well-scoped, issue-linked (#4595, part of #4594) change with no scope creep and no schema/migration gap since advisoryAiRouting is an existing schemaless JSON blob.

Nits — 7 non-blocking

• The pre-prompt redaction in `redactGroundingText` (src/services/ai-chat-qa.ts) is an explicit regex allow-list of known blocker codes/boundary terms — a newly added private blocker code elsewhere in the codebase won't be redacted here until this file is updated in lockstep; worth a comment or a shared constant so the two don't drift.
• Several new magic numbers (256/64/512 token clamps, 0.1 temperature, the 5/4/4/8 slice caps, the 4-chars-per-token and 0.035 neurons-per-token heuristics in src/services/ai-chat-qa.ts) would read clearer as named constants, especially since ai-summaries.ts is cited as the model this mirrors.
• test/unit/queue-5.test.ts:1284 was flagged by secret scanning for `GITHUB_APP_PRIVATE_KEY: await generatePrivateKeyPem()` — please confirm this is a freshly generated ephemeral test key and not a checked-in fixture, since the helper name alone doesn't rule out a hardcoded PEM.
• The OpenAPI description notes `commandRateLimitPolicy: hold` as a co-requisite for the tighter AI-cost rate limit to actually apply, but that's documentation-only — since the fleet default is `off`, an operator who enables `chatQa` without also flipping that policy gets an AI-cost-bearing command with no effective per-window ceiling; consider surfacing this more forcefully than a schema description (e.g., a warning in the normalizer) so it isn't missed.
• Consider a shared constant/list for the private blocker codes so `PRIVATE_DECISION_BLOCKER_PATTERN` and any public-redaction helper elsewhere in the codebase can't silently diverge as new blocker codes are added.
• PR author also opened the linked issue — Link an issue that was opened by a different contributor, or provide a rationale for why this self-authored issue represents genuine discovery work.
• Touches a guarded path — held for manual review — A maintainer must review and merge this change.

Signal	Result	Evidence
Code review	✅ No blockers	1 reviewer
Linked issue	✅ Linked	#4595
Related work	✅ No active overlap found	No same-issue or scoped active PR overlap found.
Change scope	✅ 20/20	Low review scope from cached public metadata (1 linked issue).
Validation posture	✅ 25/25	PR body includes validation/test evidence.
Contributor workload	✅ 10/10	Author activity: 48 registered-repo PR(s), 40 merged, 420 issue(s).
Contributor context	✅ Confirmed Gittensor contributor	JSONbored; Gittensor profile; 48 PR(s), 420 issue(s).
Gate result	⚠️ Not blocking	Advisory; not blocking this PR.
Improvement	✅ Minor	risk: clean · value: minor — Code changes are accompanied by test evidence.

Review context

• Author: JSONbored
• Role context: owner (maintainer lane)
• Public audience mode: oss maintainer
• Lane context: Repository is configured for direct PR review.
• Public profile languages: not available
• Official Gittensor activity: 48 PR(s), 420 issue(s).
• PR-specific overlap: none found.

Contributor next steps

• Treat this as maintainer-lane context rather than normal contributor-lane activity.

Signal definitions

• Related work = same linked issue, overlapping active PRs, or title/path similarity.
• Change scope = cached public metadata such as size labels, draft state, and review-burden hints.
• Validation posture = whether the PR provides enough public validation/test evidence for maintainer review.
• Contributor workload = public contributor activity and cleanup pressure, not a repo-wide quality failure.
• Contributor context = public GitHub/Gittensor identity context; non-Gittensor status is not a blocker.

Visual preview

Route	Viewport	Before (production)	After (this PR's preview)	Diff
/	desktop			—
/	mobile			—

Click any thumbnail to open the full-size screenshot. Before = production · After = this PR's preview deploy.

_{🟩 Safe / merged · 🟦 Advisory · 🟨 Held for review · 🟥 Blocked / closed}

---

💰 Earn for open-source contributions like this. Gittensor lets GitHub contributors earn for the work they already do — register to start earning →.

Checked by Gittensory, a quiet PR intelligence layer for OSS maintainers.

• Re-run Gittensory review