eeco

Vision

The AI-agnostic ecosystem for autonomous project housekeeping.

README · Vision · Cockpit · Usage · Architecture · Public API · Extending · Contributing · Upgrading · Versioning · Changelog · Security

--- eeco is the AI-agnostic ecosystem for autonomous project housekeeping — an intelligent, system-level framework that sits as a meta-layer above modern software development. Its purpose is to radically automate project housekeeping so developers can focus entirely on creative implementation and problem-solving. This document states where eeco is headed. The README and USAGE describe what it does today. ## The vision eeco is not a reactive tool — it is a proactive background process that continuously analyses, understands, and actively maintains a project. Through pattern recognition, eeco identifies recurring development and maintenance patterns and derives workflows from them automatically — proactively, or on explicit user request. ## Core concepts ### Model-agnostic interoperability — the context compiler eeco is a universal abstraction layer. It ensures that every AI — from specialised small models to high-end LLMs like Gemini, Claude, or GPT — receives exactly the project context it needs. eeco acts as a **context compiler** that translates complex project documentation into a form any AI can process, in a standardised shape. ### Autonomous housekeeping With built-in standard workflows — a bug finder for error analysis, a hygiene keeper for code-standards enforcement, drift detectors across memory and docs — eeco takes the administrative load. The system learns from the developer workflow and evolves through AI-assisted self-optimisation. ### Intelligent context steering — AI guardrails eeco is a **guardrail** for AI agents. It delivers cross-model binding rules — language policies, attribution rules, project conventions — that minimise hallucinations and keep project quality consistent without the user repeating explicit instructions on every call. ## The goal eeco aims to become the standard for AI-assisted project steering: the interface where human software architecture and AI-driven implementation meet. Through the handshake between eeco and any AI model, a seamless environment emerges in which the AI acts not just as a code generator, but as part of a self-organising, highly efficient development system. ## Operating principles The vision above is the destination. These principles hold on the way there and are never traded away: - **Human-in-the-loop.** eeco proposes; the operator decides. It never commits to, pushes, or activates a workflow on your project's tracked tree without explicit consent. - **Init exception.** `eeco init` is the single verb permitted to make one initial commit and one initial push to the HOST repo, both restricted to the `.gitignore` line that scopes eeco's workspace out of the tracked tree — plus the private workspace-history repo described below, which is local-only and confined to the gitignored workspace. Every other verb — `run`, `gc`, `hooks`, `gates`, `add`, etc. — continues to never commit to or push the tracked tree, and never write outside the workspace. - **Workspace history (the AI's logbook).** eeco may keep a private, **local** git repository inside its own gitignored workspace directory (`/`) to version its knowledge layer — memory, queue, decisions, manifests — over time. This repository **never leaves the machine** (no remote, no push by default) and **never touches the host project's tracked tree**: it records only what eeco already writes inside its own gitignored workspace. It is created opt-out at `eeco init` (`workspace_history`: `manual` by default, `auto` to commit on every change, `off` to disable) and removed at `eeco uninstall`. Inside this private repo eeco has a free hand — commit, squash, gc — because nothing there is shared or tracked by the host; every other verb still never commits to the tracked tree, pushes, or writes outside the workspace. - **Opt-in AI spend.** Every AI call is consent-gated and budget-capped. The default automation level is manual. - **Local-first and private.** eeco reads anywhere in the repo, writes only inside its gitignored workspace. No telemetry, no hosted service. - **Reversible.** Every escape from the workspace — the pre-commit hook, the session-start integration, the commit-msg hook, the commit-guard PreToolUse hook — is opt-in, ledgered, and removable byte-for-byte. ## Non-goals - Not a CI service, not a hosted product — it runs on your machine. - Not tied to one AI vendor — the knowledge layer is provider-agnostic. - Not a replacement for operator judgment — eeco surfaces and proposes; the operator approves. ## Roadmap eeco follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html) over the frozen surface in [`docs/PUBLIC_API.md`](docs/PUBLIC_API.md). The next stretch deepens the context compiler and grows the autonomous-housekeeping workflows. See [`CHANGELOG.md`](CHANGELOG.md) for what has shipped. ## See also - [README.md](README.md) — quick start and install - [docs/USAGE.md](docs/USAGE.md) — full user reference - [docs/ARCHITECTURE.md](docs/ARCHITECTURE.md) — architecture overview --- [← Prev: README](README.md) · [Next: Cockpit →](docs/COCKPIT.md)