Ai/issue 2 devloop

[SmileJune]

Summary by CodeRabbit

• New Features

  • Added /ai revise command support for applying feedback to existing PRs.
  • DevLoop runner can now run continuously via systemd timer (executes every minute).
  • GitHub token is now optional for read-only operations.
  • Improved state tracking prevents redundant command reprocessing.

• Documentation

  • Updated documentation covering new commands, systemd integration, and environment variables.

[coderabbitai]

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: fd41f972-47a9-4f60-8f29-b5f5657e9cf8

📥 Commits

Reviewing files that changed from the base of the PR and between 9706d5f and d659e19.

📒 Files selected for processing (4)

• docs/devloop-runner.md
• scripts/devloop/home-1.service
• scripts/devloop/home-1.timer
• scripts/devloop/runner.py

---

📝 Walkthrough

Walkthrough

The DevLoop runner is extended to accept /ai revise feedback commands in addition to /ai implement and /ai approve. A persistent state model tracks processed comment IDs to avoid reprocessing. GitHub client token is now optional for read-only operations. New systemd service/timer enable continuous local execution.

Changes

DevLoop Runner Revise and State Management

Layer / File(s)	Summary
Documentation Updates docs/devloop-runner.md	Flow, environment variables, revise examples, mark-existing flag usage, systemd setup walkthrough, and rewritten safety notes reflecting new /ai revise, /ai approve, /ai implement capabilities and persistent state behavior.
Systemd Service and Timer Configuration scripts/devloop/home-1.service, scripts/devloop/home-1.timer	Systemd unit files for continuous execution: service runs runner once with network/environment setup; timer schedules every minute with persistence for missed runs.
Persistent State Model and Environment I/O scripts/devloop/runner.py	RunnerState dataclass with processed comment IDs set. State file path resolution and JSON load/save helpers. load_env_file() reads local environment from configurable file.
GitHub Client Enhancement and Command Selection scripts/devloop/runner.py	GitHubClient accepts optional token for read-only operations. Refactored request headers. list_devloop_issues() replaces older method. Default command constants and helpers for detecting implement vs revise based on comment content and approver eligibility.
Candidate Discovery with State Filtering scripts/devloop/runner.py	find_candidates() evaluates issue and PR comment threads, selects latest valid command per thread, and filters by already-processed IDs or existing runner markers after the triggering command.
Revise Command Execution scripts/devloop/runner.py	revise_feedback() extracts user feedback from revise comments. build_codex_prompt(), run_codex(), and commit_and_push() accept command kind and comment to tailor prompts and commit messages for implement vs revise.
Main Loop Orchestration and State Persistence scripts/devloop/runner.py	Loads environment file and persisted state on startup. --mark-existing flag marks candidate comment IDs without executing. mark_processed() persists comment IDs to state file when --execute-codex is enabled.

Estimated code review effort

🎯 4 (Complex) | ⏱️ ~50 minutes

Possibly related PRs

• SmileJune/techcase#7: Introduced the initial DevLoop runner for /ai implement command handling; this PR refactors command detection and execution to support /ai revise and /ai approve with persistent state tracking.

Poem

🐇 A runner that revises with care,
Remembers what comments were there,
With state files so keen,
And systemd so clean,
The DevLoop bounces through air!

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch ai/issue-2-devloop

---

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.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}