Add GPT-5 verbosity model options to Codex and OpenCode#3944
Conversation
|
Important Review skippedAuto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Repository UI Review profile: CHILL Plan: Pro Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 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 |
| ...(verbosity ? { verbosity } : {}), | ||
| ...(input.interactionMode !== undefined ? { interactionMode: input.interactionMode } : {}), | ||
| ...(codexAttachments.length > 0 ? { attachments: codexAttachments } : {}), | ||
| } as CodexSessionRuntimeSendTurnInput & { readonly verbosity?: string }; |
There was a problem hiding this comment.
🟡 Medium Layers/CodexAdapter.ts:1557
The verbosity option is read from modelSelection and placed into turnInput, but CodexSessionRuntimeSendTurnInput has no verbosity field and sendTurn never forwards it to buildTurnStartParams. The intersection type cast as CodexSessionRuntimeSendTurnInput & { readonly verbosity?: string } only silences the compiler, so selecting a verbosity level has no effect on the Codex turn/start request — the value is silently dropped. To make this functional, add verbosity to CodexSessionRuntimeSendTurnInput, destructure it in sendTurn, and include it in the params built by buildTurnStartParams.
🤖 Copy this AI Prompt to have your agent fix this:
In file @apps/server/src/provider/Layers/CodexAdapter.ts around line 1557:
The `verbosity` option is read from `modelSelection` and placed into `turnInput`, but `CodexSessionRuntimeSendTurnInput` has no `verbosity` field and `sendTurn` never forwards it to `buildTurnStartParams`. The intersection type cast `as CodexSessionRuntimeSendTurnInput & { readonly verbosity?: string }` only silences the compiler, so selecting a verbosity level has no effect on the Codex `turn/start` request — the value is silently dropped. To make this functional, add `verbosity` to `CodexSessionRuntimeSendTurnInput`, destructure it in `sendTurn`, and include it in the params built by `buildTurnStartParams`.
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes using high effort and found 2 potential issues.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Want fixes drafted automatically? Bugbot Autofix can create code changes for findings. A team admin can enable Autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit 99c307d. Configure here.
| ...(input.interactionMode !== undefined ? { interactionMode: input.interactionMode } : {}), | ||
| ...(codexAttachments.length > 0 ? { attachments: codexAttachments } : {}), | ||
| }) | ||
| .sendTurn(turnInput) |
There was a problem hiding this comment.
Codex verbosity never reaches turn
High Severity
The CodexAdapter passes a verbosity parameter to session.runtime.sendTurn, but CodexSessionRuntime doesn't recognize or forward it. This means user-selected verbosity for Codex models is dropped, making the new UI option ineffective.
Reviewed by Cursor Bugbot for commit 99c307d. Configure here.
| readonly model: ProviderListResponse["all"][number]["models"][string]; | ||
| readonly agents: ReadonlyArray<Agent>; | ||
| }): ModelCapabilities { | ||
| const hasGpt5Verbosity = isGpt5Model(input.model.id); |
There was a problem hiding this comment.
OpenCode GPT-5 check case-sensitive
Medium Severity
OpenCode's isGpt5Model check for the verbosity option is case-sensitive, unlike Codex's case-insensitive approach. This leads to inconsistent availability of the verbosity option for models with non-lowercase 'gpt-5' prefixes across providers.
Reviewed by Cursor Bugbot for commit 99c307d. Configure here.
ApprovabilityVerdict: Needs human review 1 blocking correctness issue found. This PR adds new user-facing verbosity options for GPT-5 models. Unresolved review comments identify that the Codex verbosity value is silently dropped (feature doesn't work) and there's case-sensitivity inconsistency between providers. These substantive bugs require human review. You can customize Macroscope's approvability policy. Learn more. |


Summary
verbositymodel option (low,medium,high, defaultmedium) for GPT-5 family models in both providers.reasoningEffort,serviceTier, and OpenCodeagentbehavior).verbosityin OpenCode and Codex turn requests.OpenCodeProvider.test.tsOpenCodeAdapter.test.tsCodexProvider.test.tsCodexAdapter.test.tsValidation
vp checkvp run typecheckpnpm run dist:desktop:dmg.dmgartifact directly fromrelease/Notes
codex/GPT-5.3-Codex-Spark.Note
Low Risk
Changes are additive UI/capability wiring and optional turn parameters; no auth or data-path changes, with coverage in provider and adapter tests.
Overview
Adds a Verbosity model option (low / medium / high, default medium) for models whose IDs start with
gpt-5, surfaced in Codex and OpenCode provider capabilities alongside existing reasoning, service tier, agent, and variant controls.Codex maps the new descriptor in
mapCodexModelCapabilitiesand forwards the selected value frommodelSelectionthroughCodexAdapter.sendTurninto the session runtime turn payload. OpenCode exposes the same capability when listing models and passesverbosityintosession.promptAsyncwhen present.Provider and adapter tests cover GPT-5 vs non–GPT-5 capability exposure and end-to-end forwarding for both drivers.
Reviewed by Cursor Bugbot for commit 99c307d. Bugbot is set up for automated code reviews on this repo. Configure here.
Note
Add GPT-5 verbosity model option to Codex and OpenCode providers
verbosityselect option (low,medium,high, defaultmedium) to model capabilities for GPT-5 models in both CodexProvider.ts and OpenCodeProvider.ts; non-GPT-5 models are unaffected.verbosityvalue frommodelSelectionto the respective runtime calls when it is present.📊 Macroscope summarized 99c307d. 4 files reviewed, 0 issues evaluated, 0 issues filtered, 0 comments posted
🗂️ Filtered Issues
No issues evaluated.