Skip to content

Project Charter

  • Date: 2026-08-02 — baseline consolidated from README.md, Roadmap.md, and Architecture.md; created to resolve the existing references to this file.

Purpose

AI Engineering Lab is a personal research and development repository for understanding AI engineering systems, developer tooling, LLM infrastructure, observability, and workflow automation. It now contains the released Provider-Aware Statusline (v2.0.0, 2026-08-05) together with the evidence-first research, findings, and ADRs that produced it.

Active project

Claude Code + OmniRoute observability. Understand the complete path from Claude Code's runtime information to OmniRoute's routed provider/model so that a provider-aware statusline could be designed and built from observed behavior rather than hardcoded assumptions.

Scope

  • Evidence-first investigation across the phases in Roadmap.md: Discovery → Architecture → Design → Implementation → Validation → Documentation.
  • External research preserved verbatim in Research/; local observations in Findings/; verified conclusions in Knowledge-Base.md; material project and architecture choices in Decisions.md.
  • The statusline design and implementation only after the relevant data flow, interfaces, constraints, and failure modes are understood and approved (ADR-0001: investigate first, design second, implement last).

Constraints

  • Read-only during discovery: do not modify Claude Code configuration, OmniRoute, or the statusline unless a separately approved experiment requires it (operating rule 5).
  • Secrets never surface in outputs, logs, or commits. .claude/settings.local.json contains a live token and is gitignored; only the placeholder example is committed.
  • External claims — including external AI research — are leads, not verified facts, until confirmed locally or against an authoritative source (operating rule 6).
  • No project Skill is created automatically; the organic lifecycle in CLAUDE.md governs Skill creation.

Success criteria

  • Phase 1 (met 2026-08-02): external research recorded verbatim with provenance, and the major statusline/OmniRoute fields, integration points, and the generic model label verified locally.
  • Phase 2: architecture approved only after the data source for each displayed value is identified and the design does not depend on unverified behavior (Roadmap Phase 2 gate).
  • Phase 3–6: an approved design; the smallest production-quality implementation consistent with it; validation against real and failure scenarios; and clear documentation and maintenance notes.

Non-goals

  • No production code during the discovery phases.
  • No modification of the live environment without separate approval.
  • No display of tokens, prompts, or sensitive request content in the statusline.
  • No unverified or fabricated facts in project documentation.