TODO_COMMIT.md

TODO: Commit Briefing — MetaMetaMeta

Date: 2026-03-19 Branch: master Last commit: 4b14a07 feat: add eval dashboard with E2-E7 results (84 runs, 7 experiments)

---

Files to Commit

1. visual_explainer_run_pipeline.html

• 1656-line interactive HTML visualization of the run pipeline
• Standalone file, no build step required
• Category: visualization / explainer

2. Visualisungen_improve/index.html

• Improved dashboard HTML (replacement or iteration on existing eval dashboard)
• Category: visualization / dashboard

3. Visualisungen_improve/data.json

• Data file backing the improved dashboard
• Should be committed alongside index.html — they belong together
• Category: data / dashboard

4. Visualisungen_improve/export_runs.py

• Python script to export run data (presumably to data.json)
• Category: tooling / scripts

---

Files to Ignore (Do NOT Commit)

agent_arena.db

• SQLite binary database — runtime artifact, not source
• Binary files have no meaningful diffs in git
• Would bloat the repo history permanently
• Must be added to .gitignore before committing anything

---

.gitignore Updates

Current .gitignore content:

.DS_Store
__pycache__/
*.pyc
.venv/

Add the following:

# Databases / runtime artifacts
*.db
*.sqlite
*.sqlite3

This covers agent_arena.db and any future SQLite files.

---

Commit Strategy

Two commits. Reason: the pipeline explainer and the Visualisungen_improve work are separate concerns.

Commit 1 — .gitignore update (do this first, always)

chore: ignore SQLite database files

Stage only: .gitignore

Commit 2 — Pipeline visualizer

feat: add interactive run pipeline visual explainer

Stage only: visual_explainer_run_pipeline.html

Commit 3 — Improved dashboard and tooling

feat: add improved eval dashboard with export script (Visualisungen_improve)

Stage: Visualisungen_improve/index.html, Visualisungen_improve/data.json, Visualisungen_improve/export_runs.py

Total: 3 commits.

---

Exact Commands

Copy-paste ready. Run in order.

Step 1 — Update .gitignore:

# Edit .gitignore manually to add *.db lines, then:
git -C /home/smlflg/Projekte/MetaMetaMeta add .gitignore
git -C /home/smlflg/Projekte/MetaMetaMeta commit -m "chore: ignore SQLite database files"

Step 2 — Verify db is now ignored:

git -C /home/smlflg/Projekte/MetaMetaMeta status
# agent_arena.db should NOT appear in the output

Step 3 — Commit pipeline explainer:

git -C /home/smlflg/Projekte/MetaMetaMeta add visual_explainer_run_pipeline.html
git -C /home/smlflg/Projekte/MetaMetaMeta commit -m "feat: add interactive run pipeline visual explainer"

Step 4 — Commit improved dashboard:

git -C /home/smlflg/Projekte/MetaMetaMeta add Visualisungen_improve/index.html Visualisungen_improve/data.json Visualisungen_improve/export_runs.py
git -C /home/smlflg/Projekte/MetaMetaMeta commit -m "feat: add improved eval dashboard with export script (Visualisungen_improve)"

Step 5 — Verify final state:

git -C /home/smlflg/Projekte/MetaMetaMeta log --oneline -5
git -C /home/smlflg/Projekte/MetaMetaMeta status

Expected: clean working tree, 3 new commits on top of 4b14a07.

---

Pre-commit Checks

Before running any commit command:

1. agent_arena.db must not appear in git status after .gitignore update. If it does, the .gitignore edit was not saved correctly.

2. No accidental binary files — run git status after each git add and read the output carefully. If anything unexpected appears, do not commit.

3. HTML files open in browser — optionally verify visual_explainer_run_pipeline.html and Visualisungen_improve/index.html open without errors before committing.

4. export_runs.py is not sensitive — confirm it contains no hardcoded credentials, API keys, or absolute paths that only work on this machine.

5. data.json size — if it is very large (>1MB), consider whether it belongs in git or should be regenerated from export_runs.py on demand.

---

Decision Required Before Starting

• Is Visualisungen_improve/data.json small enough to commit? Check with:

  ls -lh /home/smlflg/Projekte/MetaMetaMeta/Visualisungen_improve/data.json
  

  If over 500KB, ask Samuel whether to gitignore it and regenerate from export_runs.py instead.