mozilla / mozilla/blender

feat(auto-engineer): make plan/implement turn and budget limits configurable per run

Open
#101 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
6
Forks
2
Avg merge
1d 11h
Merged PRs (30d)
17

Description

Problem: The auto-engineer's per-phase turn and budget caps are hardcoded — config/defaults.yml (max_plan_turns: 20, max_plan_budget_usd: 1.50, etc.) with defaults baked into scripts/run-claude.sh (MAX_CLAUDE_TURNS, MAX_BUDGET_USD). They aren't exposed on the workflow_dispatch, so there's no way to raise them for a single run from the Actions UI.

A plan run on #99 hit the cap and exited with code 1 and 0 lines of output ("likely hit max-turns or budget"), with no self-serve way to give it more room — see run https://github.com/mozilla/blender/actions/runs/30034409270. This was the main thing blocking us from using the auto-engineer.

Proposal: Add optional workflow_dispatch inputs — e.g. max_turns and max_budget_usd — that override the per-phase defaults, plumbed through to MAX_CLAUDE_TURNS / MAX_BUDGET_USD in run-claude.sh. Empty/unset falls back to config/defaults.yml (current behavior).

Acceptance: Can bump the plan/implement turn and budget limits for a single dispatch from the Run-workflow form without editing the repo.

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with the workflow that defines workflow_dispatch, then trace its inputs into config/defaults.yml and scripts/run-claude.sh. Verify how MAX_CLAUDE_TURNS and MAX_BUDGET_USD are currently populated and test a dispatch with empty and overridden values. Done means the Run workflow form can change plan and implement limits for one run while empty inputs preserve current defaults.

Written by the indexing model from the issue text.

Assessment

Tech stack
github-actions, shell, yaml
Domain
ci-cd, devops, tooling
Issue type
Feature
Difficulty
3/5
Estimated time
1-2 days
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
70/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.