feat(plan): iterative planning & plan review #8105
Draft
+1,834
−28
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
Add comprehensive plan mode improvements including:
- Plan state management with session-scoped storage at ~/.local/share/opencode/plan/
- ExitPlanMode tool that AI calls when plan is ready for review
- Plan review TUI component with approve/reject keybindings
- Session-scoped permissions using {sessionID} placeholder in patterns
- evaluateWithSession() for permission evaluation with session context
- Plan→build agent transition on approval with synthetic continuation
- Post-compaction reminder pointing AI to plan file location
- Server endpoints for plan approve/reject operations
…g instructions
- Update plan.txt prompt to include {PLAN_FILE_PATH} placeholder
- Make insertReminders async to get plan file path from Plan.getOrCreate
- Add clear instructions for writing to plan file early and often
- Emphasize incremental updates rather than waiting for complete plan
- Add planContent field to session revert info - Save plan file content before revert, delete the file - Restore plan file content on unrevert - Add Plan.deletePlan() and Plan.writeContent() functions
Previously, revert would delete the plan file and save its content for unrevert. This caused the plan to be lost if the user proceeded with the revert instead of undoing. Now we save the content for unrevert but leave the file intact during revert. This preserves the plan file regardless of whether the user proceeds or undoes.
- Add PlanPatchPart to track plan content changes (only stored when content changes) - Capture plan content at step-start, compare at step-finish, create PlanPatchPart if different - Restore plan content from collected PlanPatchParts during revert - Fix ExitPlanMode to throw RejectedError when dismissed (interrupts loop) vs return result when rejected with feedback (allows revision) - Filter plan-patch parts from share sync (internal state only)
- Add PlanReviewInline component for inline plan review UI - Add plan_review state and event handlers to global-sync - Add initial loading of pending plan reviews on bootstrap - Add /plan-review/:requestID/content endpoint to fetch plan content - Render plan review UI above prompt input when pending - Regenerate SDK to include new plan review content endpoint
Contributor
|
Thanks for your contribution! This PR doesn't have a linked issue. All PRs must reference an existing issue. Please:
See CONTRIBUTING.md for details. |
Contributor
|
The following comment was made by an LLM, it may be inaccurate: Based on my search, I found one potentially related PR: Related PR:
This existing PR implements a foundational piece of the functionality (exiting plan mode), while the current PR (#8105) adds the broader iterative planning and plan review features that build upon or extend that capability. All other search results are unrelated to the planning/plan review features. |
Author
|
I'm publishing this now as a draft PR to get visibility and garner some feedback around the topic. It's a big PR with multiple different additions so I wanna make sure to get it right. Will be working further on it after I get to do some testing in the coming days |
Restore the permission-task.test.ts that was emptied during the
iterative plan mode feature implementation. Add new tests for the
evaluateWithSession function that handles {sessionID} placeholder
expansion in permission patterns.
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.
What does this PR do?
This PR adds two new features:
Iterative Planning, something which for example Claude Code and I believe Cursor as well which will keep iterating on one and the same plan document instead of writing it to the conversation history, losing context between iteration, etc...
Plan Review / Summary, after iterating on the plan, it will request to exit planning mode. When it does this it'll let you approve/reject the plan, give an optional reason, or dismiss it entirely and interrupt the flow. Gives you an easy way of looking through the entire plan before executing on it.
How did you verify your code works?
Basic tests in plan mode via TUI and Web interface. Will be doing more extensive testing on real projects (hence it being a draft PR)