feat(vision): routed describer backend — any picker-visible image-capable model (#2188) - #2306
Merged
Merged
Conversation
…ds (#2188 roadmap 170) Any picker-visible, non-provably-text-only provider row can now be offered as the vision describer through the new 'routed' backend. Adds the VISION_BACKENDS descriptor table, namespaced routed option values, the four-family + namespaced provably-blind probe, PUT coherence rules (namespaced <-> routed), and claude-code override parity. Docs 170/180 revised with three audit rounds folded in.
…roadmap 180) The routed backend describes images by POSTing the proxy's own /v1/chat/completions with the namespaced describer model, so every provider wire the router speaks is a valid describer. Recursion fence: the request carries x-opencodex-vision-describe, detected at the chat surface (bridge rebuilds headers) and honored at the Responses plan site — marked requests strip instead of describing (depth cap 1). Native chat fast path now defers image-bearing text-only-model requests to the Responses pipeline so vision coverage is symmetric. Admission ladder: env token, service token file, first apiKeys entry, sent as x-opencodex-api-key.
…ber verbatim Splits VisionBackend from the legacy SidecarBackend pair (web-search keeps its own union), infers routed for namespaced values in the GUI fallback so a working describer is never rewritten on save, widens the claude-code override select, updates the CLI usage line, and makes GET report a routed backend's namespaced model instead of collapsing to the legacy default.
Contributor
|
✅ Deterministic PR hygiene checks passed. |
Contributor
|
Caution Review failedThe pull request is closed. ℹ️ Recent review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: ASSERTIVE Plan: Pro Plus Run ID: ⛔ Files ignored due to path filters (1)
📒 Files selected for processing (23)
📝 WalkthroughWalkthroughThe vision sidecar now supports a ChangesUnified vision routing
Estimated code review effort: 4 (Complex) | ~60 minutes Sequence Diagram(s)sequenceDiagram
participant Client
participant ResponsesCore
participant VisionIndex
participant RoutedDescribe
participant ChatCompletions
participant TextOnlyModel
Client->>ResponsesCore: Send image request
ResponsesCore->>VisionIndex: Plan routed vision description
VisionIndex->>RoutedDescribe: Describe image with provider/model
RoutedDescribe->>ChatCompletions: Send terminal multimodal request
ChatCompletions->>ResponsesCore: Propagate terminal marker
ResponsesCore->>TextOnlyModel: Forward caption without image
TextOnlyModel-->>RoutedDescribe: Return description text
RoutedDescribe-->>ResponsesCore: Provide extracted caption
Suggested reviewers: ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Contributor
✅ READY
Hygiene✅ Deterministic PR hygiene checks passed. |
lidge-jun
marked this pull request as ready for review
August 21, 2026 16:45
This was referenced Aug 21, 2026
4 tasks done
This was referenced Aug 21, 2026
[Bug]: Cursor Auto Intelligence echoes [Tool Result] as assistant chat after native tool calls
#2317
Closed
Merged
Merged
3 tasks done
agentHits
pushed a commit
to agentHits/opencodex
that referenced
this pull request
Sep 17, 2026
…sidecar feat(vision): routed describer backend — any picker-visible image-capable model (lidge-jun#2188)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds the "routed" vision-sidecar backend (#2188 roadmap 160-190, revised): any picker-visible, non-provably-text-only provider row can describe images. The describer dispatches through the proxy's own /v1/chat/completions on loopback, so every provider wire the router speaks (xai, Antigravity, volcengine, cursor, zenmux, ...) is a valid describer with ONE executor. Options carry namespaced provider/model values; the provably-blind write gate widens to all vendor families and learns namespaced ids; PUT coherence rejects namespaced-with-legacy-backend and bare-with-routed. Recursion fence: the loopback describe carries x-opencodex-vision-describe, detected at the chat surface and honored at the Responses plan site (depth cap 1); the native chat fast path defers image-bearing text-only-model requests to the Responses pipeline. GUI splits VisionBackend from the legacy pair and infers routed for namespaced values; CLI list/usage updated; GET reports a routed describer verbatim.
Design + 4-round adversarial audit recorded in devlog/_plan/260820_sidecar_selection_unification/160-190.
Verification
Checklist
Screenshot
Summary by CodeRabbit
New Features
provider/modelidentifiers in vision selections.Bug Fixes