# Self-Improvement Loop Automation
# SPEC: SPEC-automation.md
#
# Runs the self-improvement loop on a schedule with budget and iteration limits.
# Creates a PR for human review if improvements are integrated.
#
# Safety constraints:
# - Human approval required for integration
# - Fail-fast on evaluation gate failures
# - Rate limits enforced via budget cap
name: Self-Improvement Loop
on:
# Manual trigger with configurable parameters
workflow_dispatch:
inputs:
budget:
description: 'Budget in USD'
required: false
default: '10.0'
type: string
max_iterations:
description: 'Maximum iterations'
required: false
default: '5'
type: string
dry_run:
description: 'Dry run (no changes)'
required: false
default: false
type: boolean
# Scheduled runs (weekly on Sundays at 2am UTC)
schedule:
- cron: '0 2 * * 0'
# Prevent concurrent runs
concurrency:
group: self-improvement-loop
cancel-in-progress: false
jobs:
improvement-loop:
name: Run Self-Improvement Loop
runs-on: ubuntu-latest
timeout-minutes: 120 # 2 hour max
# Only run on main repo, not forks
if: github.repository == 'kastalien-research/thoughtbox'
permissions:
contents: write
pull-requests: write
steps:
- name: Checkout code
uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Setup Node.js
uses: actions/setup-node@v4
with:
node-version: '20'
cache: 'npm'
- name: Install dependencies
run: npm ci
- name: Build
run: npm run build
- name: Create improvement branch
id: branch
run: |
BRANCH_NAME="improvement/sil-$(date +%Y%m%d-%H%M%S)"
git checkout -b "$BRANCH_NAME"
echo "branch=$BRANCH_NAME" >> $GITHUB_OUTPUT
- name: Run Self-Improvement Loop
id: sil
env:
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
run: |
# Set parameters from inputs or defaults
BUDGET="${{ inputs.budget || '10.0' }}"
MAX_ITER="${{ inputs.max_iterations || '5' }}"
DRY_RUN="${{ inputs.dry_run || 'false' }}"
# Build command
CMD="npx tsx scripts/agents/run-improvement-loop.ts"
CMD="$CMD --budget $BUDGET"
CMD="$CMD --max-iterations $MAX_ITER"
CMD="$CMD --output improvement-results.json"
CMD="$CMD --verbose"
if [ "$DRY_RUN" = "true" ]; then
CMD="$CMD --dry-run"
fi
echo "Running: $CMD"
$CMD || echo "SIL completed with exit code $?"
# Check if results file exists
if [ -f improvement-results.json ]; then
echo "results_exist=true" >> $GITHUB_OUTPUT
# Extract summary
TOTAL=$(jq '.summary.totalIterations' improvement-results.json)
SUCCESS=$(jq '.summary.successful' improvement-results.json)
FAILED=$(jq '.summary.failed' improvement-results.json)
echo "total_iterations=$TOTAL" >> $GITHUB_OUTPUT
echo "successful=$SUCCESS" >> $GITHUB_OUTPUT
echo "failed=$FAILED" >> $GITHUB_OUTPUT
else
echo "results_exist=false" >> $GITHUB_OUTPUT
fi
- name: Generate Scorecard
if: steps.sil.outputs.results_exist == 'true'
run: |
# Generate scorecard from improvement history
npx tsx -e "
import { ImprovementEventStore, ScorecardAggregator } from './src/observatory/index.js';
const store = new ImprovementEventStore();
await store.initialize();
const aggregator = new ScorecardAggregator(store);
const scorecard = await aggregator.computeScorecard({
outputPath: 'scorecard.json',
recentCount: 20
});
console.log('Scorecard generated:', JSON.stringify(scorecard.metrics, null, 2));
" || echo "Scorecard generation skipped"
- name: Upload Artifacts
if: steps.sil.outputs.results_exist == 'true'
uses: actions/upload-artifact@v4
with:
name: improvement-results
path: |
improvement-results.json
scorecard.json
retention-days: 30
- name: Check for Changes
id: changes
run: |
# Check if there are any code changes
if git diff --quiet HEAD; then
echo "has_changes=false" >> $GITHUB_OUTPUT
else
echo "has_changes=true" >> $GITHUB_OUTPUT
git diff --stat
fi
- name: Commit Changes
if: steps.changes.outputs.has_changes == 'true' && inputs.dry_run != true
run: |
git config user.name "github-actions[bot]"
git config user.email "github-actions[bot]@users.noreply.github.com"
git add -A
git commit -m "feat(sil): Automated improvements from self-improvement loop
Run summary:
- Total iterations: ${{ steps.sil.outputs.total_iterations }}
- Successful: ${{ steps.sil.outputs.successful }}
- Failed: ${{ steps.sil.outputs.failed }}
Generated by Self-Improvement Loop workflow."
- name: Push Branch
if: steps.changes.outputs.has_changes == 'true' && inputs.dry_run != true
run: |
git push -u origin ${{ steps.branch.outputs.branch }}
- name: Create Pull Request
if: steps.changes.outputs.has_changes == 'true' && inputs.dry_run != true
env:
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
run: |
gh pr create \
--title "feat(sil): Automated improvements from self-improvement loop" \
--body "## Summary
This PR contains automated improvements generated by the Self-Improvement Loop.
### Run Statistics
- **Total Iterations:** ${{ steps.sil.outputs.total_iterations }}
- **Successful:** ${{ steps.sil.outputs.successful }}
- **Failed:** ${{ steps.sil.outputs.failed }}
### Artifacts
- See workflow artifacts for detailed results and scorecard
### Review Checklist
- [ ] Review all code changes
- [ ] Verify tests pass
- [ ] Check for unintended side effects
- [ ] Approve for merge
---
*This PR was automatically generated by the Self-Improvement Loop workflow.*" \
--label "automated,self-improvement"
- name: Summary
run: |
echo "## Self-Improvement Loop Results" >> $GITHUB_STEP_SUMMARY
echo "" >> $GITHUB_STEP_SUMMARY
if [ "${{ steps.sil.outputs.results_exist }}" = "true" ]; then
echo "| Metric | Value |" >> $GITHUB_STEP_SUMMARY
echo "|--------|-------|" >> $GITHUB_STEP_SUMMARY
echo "| Total Iterations | ${{ steps.sil.outputs.total_iterations }} |" >> $GITHUB_STEP_SUMMARY
echo "| Successful | ${{ steps.sil.outputs.successful }} |" >> $GITHUB_STEP_SUMMARY
echo "| Failed | ${{ steps.sil.outputs.failed }} |" >> $GITHUB_STEP_SUMMARY
echo "| Changes Made | ${{ steps.changes.outputs.has_changes }} |" >> $GITHUB_STEP_SUMMARY
else
echo "No results generated." >> $GITHUB_STEP_SUMMARY
fi
