🤖 feat: add model selection to /compact command #275
Merged
Conversation
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
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR.
- Add -m <model> flag to /compact command for choosing compaction model - Reuse MODEL_ABBREVIATIONS for consistent abbreviation handling - Support both abbreviated (sonnet, haiku, opus) and full model strings - Update StreamingBarrier to show model name during compaction - Falls back to workspace model when -m flag not specified - Add comprehensive tests covering all flag combinations Examples: /compact -m haiku /compact -m opus -t 8000 /compact -m anthropic:claude-opus-4-1 Generated with `cmux`
Generated with `cmux`
Move thinking policy enforcement from frontend to backend (agentSession.streamWithHistory) to ensure it's applied consistently regardless of where the request originates (compact command, manual override, etc.). This prevents bugs where custom model overrides might bypass policy requirements (e.g., gpt-5-pro requiring 'high' thinking). Single source of truth pattern - no need to remember to call enforceThinkingPolicy in multiple places. Generated with `cmux`
Haiku is the fastest good coding model, making it ideal for compaction. Generated with `cmux`
Replace dynamic import with static import to satisfy linter rules. Generated with `cmux`
Extract compaction transformation logic into shared utility to ensure consistency between initial send (ChatInput) and resume (useResumeManager). This eliminates the architectural fragility where custom model/token overrides could be lost during stream resumption. Now both code paths call the same applyCompactionOverrides() function. Key changes: - Add CompactionRequestData shared type (includes model field) - Create applyCompactionOverrides(baseOptions, compactData) utility - Update prepareCompactionMessage to use shared function - Update useResumeManager to detect compaction and apply overrides - Add comprehensive tests for transformation logic Benefits: - Single source of truth for option transformation - Impossible to have divergent behavior between paths - Custom model preserved across interruption/resume - Works retroactively (user message metadata already persisted) Generated with `cmux`
ammar-agent
force-pushed
the
compact-model
branch
from
fc9e193 to
c4154ca
Compare
ammario
approved these changes
Oct 16, 2025
Check if cached shfmt binary works before using it. If the cache contains a broken binary, reinstall it. Fixes CI failure where cache hit but binary was unusable. Generated with `cmux`
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Overview
Adds support for choosing the compaction model via
/compact -m <model>flag. Reuses existing model abbreviation infrastructure for consistency and shows the model name in the streaming barrier during compaction.Changes
-mflag parsing to/compactcommand with model abbreviation resolutionParsedCommandto include optionalmodelfield for compact commandprepareCompactionMessage()to use custom model when specifiedStreamingBarrierto show model name during compaction (e.g., "claude-haiku-4-5 compacting...")-mflag, updated 27 existing testsUsage
Design Notes
MODEL_ABBREVIATIONSmap for consistency with/modelcommandTesting
✅ All 34 slash command tests passing
✅ All 547 unit tests passing
✅ TypeScript compilation successful
✅ ESLint checks passing
Generated with
cmux