Repository navigation
In Codex 0.147.0, background helper requests use gpt-5.6-luna. Even with OpenCodex's Shadow Call Intercept enabled, some Luna requests still reached OpenAI. #1684
Description
Client or integration
Codex App
Area
Proxy and routing
Summary
The cause was the logic in:
src/lib/shadow-call.ts
OpenCodex skipped interception when the request metadata contained:
{"request_kind": "turn"}
Codex 0.147.0 can mark Luna helper calls this way, causing OpenCodex to misclassify them as normal user turns.
Fix
Reproduction
Change shouldInterceptShadowCall() from the metadata-based logic to:
export function shouldInterceptShadowCall(
modelId: string,
configured: unknown,
headers: Headers,
): boolean {
return isShadowSourceModel(modelId, configured);
}
This makes OpenCodex intercept every configured shadow source model such as:
gpt-5.6-luna
Then configure the interception target, for example:
"shadowCallIntercept": {
"enabled": true,
"model": "ollama/qwen3.5:9b"
}
Restart OpenCodex and the Codex app-server.
Result
Before:
Codex helper call
→ gpt-5.6-luna
→ OpenAI
After:
Codex helper call
→ gpt-5.6-luna
→ OpenCodex intercept
→ ollama/qwen3.5:9b
After applying this change, the unwanted gpt-5.6-luna requests to OpenAI stopped.
Version
2.15.1
Operating system
win11
Provider and model
No response
Logs or error output
Screenshots and supporting files
No response
Redacted configuration
Checks
- I searched existing issues and documentation.
- I removed secrets, tokens, account details, request credentials, and personal data.
Activity
coderabbitai commentedon Aug 14, 2026
🔗 Related PRs
#426 - feat(codex): add account-qualified model namespaces [open]
#625 - [WRONG BRANCH] fix(claude): stabilize GPT/Codex prompt cache [open]
#671 - feat(codex): add exact account routing [merged]
#1087 - fix(gui): shadow-call selects save the canonical namespaced model id (
#1075) [merged]#1089 - feat(shadow-call): drop gpt-5.4-mini from the default intercept source models [merged]
📝 Issue Planner
Check the box below or use the
@coderabbitai plancommand to generate an implementation plan and prompts that you can use with your favorite coding assistant.🧪 Issue enrichment is currently in open beta.
To disable automatic issue enrichment, add the following to your
.coderabbit.yaml:💬 Have feedback or questions? Drop into our discord!
lidge-jun commentedon Aug 14, 2026
Fixed in ba10784 — shouldInterceptShadowCall now intercepts all shadow source models unconditionally, ignoring request_kind metadata.
lidge-jun commentedon Aug 14, 2026
Fixed in ba10784 on
dev. The commit message carriesCloses #1684, so this issue will close automatically whendevis promoted tomain.fix(proxy): intercept all shadow source models regardless of request_…
lidge-jun commentedon Aug 14, 2026
Confirmed fixed by merged PR #1685. The merged change blocks Luna-family helper requests from leaking to the public OpenAI endpoint and adds regression coverage for the affected helper path. Closing as completed. Please reopen with a post-fix OpenCodex version and a redacted route trace if any helper request still bypasses the configured OpenCodex route.
release: v2.17.0
Merge dev into main: shadow-source interception fix (lidge-jun#1684)
fix(proxy): intercept all shadow source models regardless of request_…
Merge dev into main: shadow-source interception fix (lidge-jun#1684)
release: v2.17.0