Repository navigation
fix(mistral): bound image pruning memory and normalize malformed references - #8434
Conversation
Preserve both the current Mistral Chat wire-policy link and bounded image-pruning documentation. Validated locally: 95 tests / 739 assertions; typecheck 2334 files; structure, privacy, diff checks.
Co-Authored-By: Epinephrine <luvs01@hanmail.net> # Conflicts: # structure/providers-and-adapters.md
Co-Authored-By: Epinephrine <luvs01@hanmail.net>
…mistral-image-limit-vulnerability
Merging dev pushed the merged providers.md to 2004 lines, tripping the file-size ratchet's NEW_OVERSIZED verdict. Condense the image-limit note to two lines (same content) so the merged file stays below 2000. Co-Authored-By: Epinephrine <luvs01@hanmail.net>
Devin Review flagged the new Mistral image-cap paragraph: it claimed a file-ID-only input_image is always refused by the translated-adapter media guard, but the guard only inspects message content, tool output and top-level input blocks. A replayed agent_message is not scanned, so the parser degrades a file-ID-only input_image inside it to an [image: <file id>] text marker that reaches Mistral as ordinary text. Scope the refusal claim to scanned positions and describe the agent_message marker path in all eight locales. Co-Authored-By: Epinephrine <luvs01@hanmail.net>
|
No actionable comments were generated in the recent review. Walkthrough
Merge Risk:
|
|
✅ Deterministic PR hygiene checks passed. |
Codex Review Summary: ❌ FailedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
The account paying for this security review has reached its Codex usage limits. The payer can check the Codex usage dashboard. For personal accounts, using credits requires enabling “Use credits for security reviews” in Code review settings. If you do not manage the paying account, contact this repository's admins. |
Maintainer judgmentIntegrate at Existing first-party Mistral users retain the configured image cap and ordering, with malformed references normalized. Other destinations and persisted configuration are unaffected. The implementation is platform-neutral TypeScript; Windows and macOS shards were skipped, so no cross-platform execution is claimed. The structure documentation and translations were updated; the independent reviews checked file-size, Lab boundary and privacy. Fresh independent code and separate security reviews passed this exact head. The code reviewer reported 10 focused tests passing, a base/head regression check and clean typecheck; the security reviewer reported hostile-input and boundary probes passing. I inspected the diff and verified CI run 42575505025 on this head: gates, test changed, docs build, structure and aggregate succeeded. No change-caused CI failure is known. The security reviewer also reported 150 failures in a broad local adapters/providers run under contention, outside the changed Mistral area; one Kiro file passed alone. Those failures were not all reproduced on base and remain unclassified, so they are not represented as a passing run. NOT RUN by maintainer: local test suite, full suite, Windows/macOS matrix and live Mistral request. Reviewer-reported local commands and broad-run failures were not independently rerun. Maintainer integration decision: use the dev-only maintainer integration policy if GitHub blocks only on required approval; the two independent reviews remain separate from that decision. |
Summary
Bound Mistral image-limit traversal bookkeeping by walking messages backwards instead of collecting a coordinate for every image. Keep the newest images, copy only changed messages/content, and replace missing or malformed image references without consuming the valid-image cap. Returned copy-on-write arrays still scale with input size.
Carries luvs01/opencodex#814 at
259d7471cabb849ddbc59d3cd11fd117ec43bc1b, including large frozen-input tests and corrected documentation for parser/media-guard boundaries in the canonical reference and seven translations. Commit history and authorship are preserved.Verification
Checklist
Co-authored-by: Epinephrine luvs01@hanmail.net
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Summary by CodeRabbit