fix(privacy): require explicit private broker configuration - #174
fix(privacy): require explicit private broker configuration#174anantham wants to merge 5 commits into
Conversation
Context: Public defaults and operator records embedded private deployment identifiers. Changes: Remove implicit broker discovery, require runtime configuration, sanitize examples and records, and extend the existing artifact scan. Impact: Unconfigured clients stay offline; saved endpoints and existing authorization checks remain intact. Tests: 114 focused Node 24.19 tests, 31 bridge tests, TypeScript, production build, controlled Chromium acceptance, and Windows parameter/configuration checks pass. Docs: Public configuration boundary, SEC-001 amendment, Issues and WORKLOG. Historical copies require separate assessment.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
Remove remaining runtime-only records and link reviewed source, fresh CI and private historical-audit boundaries.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: d09b11da6d
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
@codex review Current head d5efee7 resolves both findings: cutover rejects missing/mismatched canonical launcher roots before hardening, and missing-variable guards run inside the existing logged launcher entrypoint. The native Windows regression probe passes three root cases and four log cases using disposable copies and a hardening sentinel. The bridge suite passes 31 tests. Please review the exact updated head; no runtime deployment was performed. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: d5efee746c
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 2061311a11
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| Promise.allSettled([ | ||
| getImageCapableModels(), | ||
| fetchIndrasNetWorkflows(endpointUrl), | ||
| endpointUrl ? fetchIndrasNetWorkflows(endpointUrl) : Promise.resolve([]), |
There was a problem hiding this comment.
Gate IndrasNet submission before prompt planning
When a saved indrasnet/... model remains selected after its broker endpoint is cleared or imported empty, this only skips catalogue discovery; the dialog still keeps that saved model selectable and leaves Generate enabled. Submission then reaches generateIllustrationForSelection, which calls IllustrationService.generateIllustrationForSelection before generateIndrasNetImage validates the endpoint, so the selected text and chapter context are sent to the configured text planner and the resulting illustration is persisted before image generation inevitably fails with INDRASNET_ENDPOINT_REQUIRED. Validate this configuration before prompt planning (or disable submission), as the new SEC-001 amendment explicitly promises that an unconfigured broker causes no prompt composition.
AGENTS.md reference: AGENTS.md:L15-L15
Useful? React with 👍 / 👎.
|
@codex review The dialog now uses the existing endpoint validator before allowing submission of the selected IndrasNet route. Empty, whitespace and malformed endpoints disable Generate with an actionable error; an explicit cloud alternative and configured offline saved model remain usable. Three red regressions now pass; 43 focused Node 24.19.0 tests, types and lint pass. Please verify the exact updated head. |
|
Codex Review: Didn't find any major issues. You're on a roll. Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
Summary
Private deployment identifiers were embedded in public defaults, examples and operator records. Require an explicit broker URL instead: a fresh or cleared configuration performs no broker discovery, while existing saved settings continue to work.
Changes
Testing
Deployment and history
Runtime operators must supply the documented private environment variables and explicit script parameters before adopting the new launchers. Authorization checks and loopback bindings are preserved. This PR changes current content; prior commits, PR diffs and cached copies remain a separate historical-cleanup decision.
Review checklist
d09b11da6dcfa5a8c5c54ea12925ca73df5c0d11: all five jobs and Vercel pass (run 33950735713).1acd19a: no blocking P1/P2 findings; subsequent changes only sanitize/update records. Smaller follow-up packets confirmed complete source receipt after the initial oversized packet was truncated.