Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions AGENTS.md
1 change: 1 addition & 0 deletions CLAUDE.md
1 change: 1 addition & 0 deletions GEMINI.md
37 changes: 19 additions & 18 deletions HUMANS.md
Original file line number Diff line number Diff line change
Expand Up @@ -4,34 +4,35 @@ Welcome. This document is for humans who want a quick, honest overview.

## What is Aiome?

Aiome is an early-stage CLI exploring how multiple AI agents could collaborate on software work. The long-term vision is
role-based agents coordinating through artifacts like code, tests, and issues. The current reality is much smaller.
Aiome is a CLI that automatically generates, maintains, and serves multi-scale knowledge about a codebase — so that AI
agents get exactly the right context for any task.

## Current State

Right now the CLI is a simple REPL that echoes input. That’s it. Anything beyond that is a direction, not a promise.
The core idea: AI agents have limited context windows and no memory between sessions. Hand-written context files
(CLAUDE.md, .cursorrules) work for small projects but don’t scale. Aiome analyzes your codebase’s structure, identifies
natural subsystem boundaries, generates knowledge at multiple resolutions, and serves the right resolution for each
task.

## Why It Exists
## Current State

We want to see if a small set of lightweight principles can help AI agents do useful work without heavy process. We
expect the project structure to evolve based on real usage.
Early stage. The CLI is a REPL that echoes input. The architecture is designed
([docs/design-doc.md](./docs/design-doc.md)) but not yet implemented. Next step is Phase 1: static and temporal
coupling analysis.

## Quick Start (Today)
## Quick Start

```bash
# Run locally
cargo run

# Or install locally
cargo install --path .
cargo run # Start the REPL
cargo install --path . # Or install locally as `aiome`
```

This starts a prompt. Type a line, it echoes back. Use Ctrl+D (Unix/Linux/macOS) or Ctrl+Z then Enter (Windows) to exit.
Ctrl+D to exit.
Comment thread
rongou marked this conversation as resolved.

## Where to Look Next
## Where to Look

- **README.md**: Direction for agents and contributors
- **src/main.rs**: The current CLI entry point
- **[README.md](./README.md)** — Project constitution (conventions, commands, repo layout)
- **[docs/design-doc.md](./docs/design-doc.md)** — Architecture: Analyze → Generate → Serve
- **[docs/vision.md](./docs/vision.md)** — Foundational ideas (BRAIN pattern, stigmergy, cognitive scaffolding)
- **src/main.rs** — The current CLI entry point

## The Name

Expand Down
80 changes: 34 additions & 46 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -2,71 +2,59 @@

> **Human readers**: See [HUMANS.md](./HUMANS.md) for a gentler introduction.

Aiome is an early-stage CLI exploring how multiple AI agents can collaborate on software work. This README is
intentionally lightweight: it sets direction without pretending the project is further along than it is.
Aiome automates multi-scale knowledge infrastructure for codebases — so that any AI agent, at any moment, can load
exactly the context it needs at exactly the right resolution. See [docs/design-doc.md](./docs/design-doc.md) for the
full architecture. See [docs/vision.md](./docs/vision.md) for the foundational ideas.

---

## Current State (Reality Wins)
## Current State

Right now this repository is small and experimental. The CLI is a simple REPL that echoes input. Anything beyond that is
directional, not guaranteed.
The CLI is a REPL that echoes input. Everything below is direction, not implemented.

If this document drifts from the code, update the document.
The next milestone is Phase 1 (Analyze): static + temporal coupling analysis producing an entanglement map and partition
structure. No LLM needed. See the design doc’s "CLI" section for the phased build plan.

---

## Direction (Not Mandates)
## Repository Layout

These are the ideas we want to grow toward, without locking ourselves into a rigid process:

- **Roles are verbs.** Treat agents as functions, not personas.
- **Bounded work.** Keep scope small, context limited, and changes reversible.
- **Coordination via artifacts.** Prefer code, tests, issues, and notes over chatty handoffs.
- **Transparency.** If you’re unsure, say so and leave a clear trail.
- **Locality.** Touch what you must, avoid sweeping refactors.

---

## Potential Roles (Names May Change)

These are placeholders for how we might structure work over time:

- **plan** — identify work worth doing
- **implement** — make a focused change
- **review** — check correctness and scope
- **debug** — fix a concrete issue
- **refactor** — improve structure without behavior changes
- **retrospect** — reflect and propose process tweaks

Treat these as a vocabulary, not a contract.
```
├── Cargo.toml # Rust 2024 edition, rustyline dependency
├── src/
│ └── main.rs # CLI entry point: REPL loop
├── docs/
│ ├── design-doc.md # Architecture: Analyze → Generate → Serve
│ └── vision.md # Foundational ideas: BRAIN pattern, stigmergy, scaffolding
├── HUMANS.md # Human-oriented overview
└── .github/workflows/
└── ci.yml # CI: fmt, clippy, test
```

---

## Working Here (Lightweight)
## Conventions

- Prefer small, readable changes over perfect systems.
- If you add tests, run `cargo test`. If you add formatting or linting conventions, document them.
- If you want to introduce new process or guardrails, open an issue first so humans can weigh in.
- **Rust 2024 edition.** `cargo fmt`, `cargo clippy -- -D warnings`, `cargo test` must all pass. CI enforces this.
Comment thread
rongou marked this conversation as resolved.
- **Small changes.** Prefer focused commits over sweeping refactors.
- **Coordination via artifacts.** Code, tests, issues, and PRs over chatty handoffs.
- **Honesty over aspiration.** If this document drifts from the code, update the document.

---

## Repository Snapshot (Today)
## Commands

This is the current layout and may evolve:

```
repo-root/
├── Cargo.toml
├── src/
│ └── main.rs
├── HUMANS.md
└── README.md
```bash
cargo run # Start the REPL
cargo test # Run tests (none yet)
cargo fmt # Format code
cargo clippy --all-targets --all-features -- -D warnings # Lint (matches CI)
cargo install --path . # Install locally as `aiome`
Comment thread
rongou marked this conversation as resolved.
```

---

## Final Note
## What’s Next

Let the structure emerge from the work. Keep the docs honest. Make it easy for the next person to understand what
changed and why.
Phase 1 of the design doc: `aiome init` producing an entanglement map and partition structure from static analysis
(tree-sitter) and git history (git2). Key dependencies to add: `tree-sitter`, `git2`, `petgraph`, `clap`.
Loading