[Autoresearch] Document Experiment Governance Runbook #911

Closed
opened 2026-03-22 13:06:06 +00:00 by perplexity · 1 comment
Collaborator

Parent

Part of #904 (Autoresearch Integration Proposal v2) — Action Item #10

Goal

Write a governance runbook documenting the full experiment lifecycle: branch → tox → PR → review → merge/revert. This ensures all agents follow the same safety constraints.

Sections to Cover

  1. Experiment Lifecycle

    • How to create an experiment branch
    • How tox validates (pytest, lint, type-check)
    • How PRs are created and structured
    • How Timmy (account #2) auto-reviews
    • Merge criteria vs revert criteria
  2. Safety Rules (from KT document — hard-won)

    • Never create test issues on the real repo
    • Morrowind is the confirmed direction — no pivots
    • OpenGauss was rejected — don't reference
    • No cloud dependencies — Ollama on localhost required
    • All experiment work lives on branches and PRs only
  3. Agent Roles in Research Swarm

    • Forge (kimi) → codebase optimization
    • Echo (hermes) → memory retrieval
    • Seer (manus) → inference quality
    • Quill (perplexity) → prompt quality
    • Mace (replit) → security
    • Helm (Timmy) → orchestration
    • rockachopa → principal authority, program.md author
  4. Compute Budget Rules

    • Metabolic protocol: Lightning balance governs tier selection
    • Resting mode (local M3 Max) is default
    • VPS for heavier inference
    • Cloud burst only when funded
  5. Failure Modes & Mitigations

    • Runaway experiments → feature branches + tox + revert default
    • Compute escalation → Lightning-balance budgets
    • Tests pass but UX degrades → H2 LLM-as-judge benchmarks
    • Experiment pollution → branches/PRs only, never auto-create issues

Deliverable

Wiki page or docs/autoresearch-governance.md in the repo

Cross-references

  • #904 (Autoresearch Epic)
  • #875 (Docker Compose — reproducible environments)

Owner

Engineering

## Parent Part of #904 (Autoresearch Integration Proposal v2) — Action Item #10 ## Goal Write a governance runbook documenting the full experiment lifecycle: branch → tox → PR → review → merge/revert. This ensures all agents follow the same safety constraints. ## Sections to Cover 1. **Experiment Lifecycle** - How to create an experiment branch - How tox validates (pytest, lint, type-check) - How PRs are created and structured - How Timmy (account #2) auto-reviews - Merge criteria vs revert criteria 2. **Safety Rules** (from KT document — hard-won) - Never create test issues on the real repo - Morrowind is the confirmed direction — no pivots - OpenGauss was rejected — don't reference - No cloud dependencies — Ollama on localhost required - All experiment work lives on branches and PRs only 3. **Agent Roles in Research Swarm** - Forge (kimi) → codebase optimization - Echo (hermes) → memory retrieval - Seer (manus) → inference quality - Quill (perplexity) → prompt quality - Mace (replit) → security - Helm (Timmy) → orchestration - rockachopa → principal authority, program.md author 4. **Compute Budget Rules** - Metabolic protocol: Lightning balance governs tier selection - Resting mode (local M3 Max) is default - VPS for heavier inference - Cloud burst only when funded 5. **Failure Modes & Mitigations** - Runaway experiments → feature branches + tox + revert default - Compute escalation → Lightning-balance budgets - Tests pass but UX degrades → H2 LLM-as-judge benchmarks - Experiment pollution → branches/PRs only, never auto-create issues ## Deliverable Wiki page or `docs/autoresearch-governance.md` in the repo ## Cross-references - #904 (Autoresearch Epic) - #875 (Docker Compose — reproducible environments) ## Owner Engineering
claude was assigned by Rockachopa 2026-03-22 23:32:00 +00:00
claude added the rejected-direction label 2026-03-23 13:51:23 +00:00
Author
Collaborator

🧹 Closed — Superseded

Autoresearch direction has been superseded by the Sovereignty Loop spec (#953) and the harness-first architecture. The research pipeline now routes through the cascade router and local model stack.

Ref: Deep Backlog Triage #1076. Reopen if needed.

🧹 **Closed — Superseded** Autoresearch direction has been superseded by the Sovereignty Loop spec (#953) and the harness-first architecture. The research pipeline now routes through the cascade router and local model stack. Ref: Deep Backlog Triage #1076. Reopen if needed.
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Rockachopa/Timmy-time-dashboard#911