airegistry-cli 2.3.2

A dependency-free CLI and versioned registry for reusable AI building blocks.

🌐 Live Registry · ⚡ Quick Start · 🧩 Components · 🐛 Issues · 🤝 Contribute

✨ What is AIRegistry?

AI projects often rebuild the same small pieces: retrievers, text chunkers, prompt builders, conversation memory, validators, citations, tool routers, and agent loops.

AIRegistry packages these capabilities as small, readable, versioned source components that you can inspect, add to your project, and own. Components are copied into your codebase instead of being hidden behind a runtime dependency or remote service.

search  →  inspect  →  add  →  build

Reuse the capability. Own the source. Pin the version.

🚀 Why AIRegistry?

⚡ Quick start

1. Install the CLI

_{Copy-paste version (the safest form, it always targets the Python you run):}

python -m pip install airegistry-cli

2. Search and inspect components

ai search rag
ai info keyword-rag

3. Preview an install, then add it

ai add keyword-rag --dry-run
ai add keyword-rag

Pin an exact version when reproducibility matters:

ai add context-builder@1.1.0

4. Manage your installed components

ai list
ai update
ai remove keyword-rag

Install from source

git clone https://github.com/MUdevelops/AIRegistry.git
cd AIRegistry
python -m pip install .

No pip available? From a repository checkout, you can invoke the CLI directly with python ai.py search rag.

🧪 Use a component in your Python project

By default, components are copied to ai_components/<name>/. Change that destination with --dir. Install records (component, version, directory and files) are stored in .airegistry/lock.json.

For example, after adding keyword-rag:

import sys

sys.path.insert(0, "ai_components/keyword-rag/src")

from keyword_rag import retrieve

matches = retrieve(
    ["Cats purr.", "Python is a programming language."],
    "python",
    top_k=1,
)

print(matches)

AIRegistry components provide building blocks, not a hosted AI model. For semantic search, model inference, or other capabilities outside a component's scope, connect the appropriate implementation in your own application.

🗂️ Component catalog

The registry currently documents 11 components across retrieval, memory, agents, prompts, text processing, tools, validation, and workflow composition.

Every component is documented and typed, and the project includes tests. Component dependencies are optional and explicit: ai add shows them but does not install them automatically.

Browse the full catalog in the Live Registry.

🔗 Compose components into a workflow

Combine small pieces into a larger pipeline while keeping each part independently inspectable:

query-rewriter ──▶ keyword-rag ──▶ context-builder ──▶ prompt-builder
                       │                  ▲
                       └── retrieval-pipeline ──────┘

conversation-memory       tool-router ──▶ agent       citation-builder

This is a conceptual composition map; each component can also be installed and used independently.

📦 How the registry is structured

Each component has a versioned directory containing its manifest, documentation, and source code:

registry/
└── keyword-rag/
    └── versions/
        ├── 1.0.0/
        │   ├── component.json
        │   ├── README.md
        │   └── src/keyword_rag.py
        └── 1.1.0/
            ├── component.json
            ├── README.md
            └── src/keyword_rag.py

Run the registry builder to validate published components and regenerate the registry indexes:

python scripts/build_registry.py

Use --check in CI to fail when generated files are stale:

python scripts/build_registry.py --check

Published releases are immutable. Bug fixes and improvements should be released as a new version. See ARCHITECTURE.md for the design details.

🤝 Community publishing

AIRegistry uses a review-first submission path:

GitHub sign-in
      ↓
Submit component or release
      ↓
Worker validation
      ↓
Branch and pull request
      ↓
CI checks + maintainer review
      ↓
Merge

Submissions can introduce a new component or publish another version of an existing one. The publishing flow is designed to avoid direct writes to main.

Read CONTRIBUTING.md for contribution rules and the component format. Questions and bug reports belong in GitHub Issues.

🛡️ Security model

Registry entries and community submissions must be treated as untrusted input. The CLI is designed to validate paths and checksums, reject executable file types, and avoid running installed component code.

Before using a component in a sensitive environment, review its source and permissions as you would any code dependency. Report suspected vulnerabilities according to SECURITY.md.

🧑‍💻 Development and checks

Run the project checks from the repository root:

python scripts/build_registry.py --check
python -W error -m unittest discover -s tests -v
python -m py_compile ai.py scripts/build_registry.py
node --check website/app.js
node --check cloudflare/worker.js

Useful operational documentation:

• Cloud authentication and deployment
• Monitoring
• Architecture
• Changelog

📄 License

AIRegistry is released under the MIT License. See LICENSE for the complete text.