Skip to main content

Axion: An open-source local-first AI system for programming.

Project description

Axion

PyPI Python License Tests Downloads


Analyze → Plan → Execute → Validate

A deterministic, local-first AI orchestrator for software engineers.

Axion is not a chat interface. It is not autocomplete. It is not “AI magic”.

It is an engineering tool.


What Axion is (and what it is not)

Axion treats Large Language Models as non-deterministic execution engines that must operate inside a strict, auditable pipeline.

Instead of asking an AI “please fix my code”, you force it to:

  1. Analyze the real project structure
  2. Plan concrete technical steps
  3. Execute changes as reviewable diffs
  4. Validate results with real tooling

No hidden prompts.
No blind edits.
No guessing.


Why this project exists

Most AI coding tools optimize for speed of output.

Software engineering optimizes for:

  • correctness
  • predictability
  • debuggability
  • long-term maintainability

That mismatch causes real problems:

  • Code is generated without understanding the project
  • Developers approve changes they don’t fully understand
  • Bugs are pushed faster, not fewer

Axion exists to slow AI down and force it to behave like a junior engineer working under strict supervision.


The core difference

Aspect Typical AI Tools Axion
Interaction Chat / Autocomplete Structured pipeline
Control Implicit Explicit and reviewable
Output Raw code Unified diffs
Context Prompt-limited Project-aware
Validation Manual Automated
Philosophy “Trust the model” “Trust the process”

Design principles

Local-first
Your code stays on your machine. Axion runs locally and only sends what is strictly necessary to the model.

No magic
Every decision is logged. Every step is inspectable. Every change is explicit.

Tool-driven
The AI uses tools (AST parsing, tests, linters). It does not replace them.

Human-in-the-loop
Nothing is applied without your approval.


What Axion can do today

  • 🔍 Structural code reviews
    Detect bugs, architectural risks, performance issues, and security problems.

  • 🧭 Technical planning
    Generate step-by-step implementation plans in Markdown.

  • 🧩 Diff-based solutions
    Propose changes as standard unified diffs — no direct file mutation.

  • 🧪 Local validation
    Run tests and tooling before applying changes.

  • 🔌 Extensible architecture
    Plugin system for custom tools and workflows.

  • 🤖 Model agnostic
    Works with OpenAI, Anthropic, Ollama, and any LiteLLM-compatible provider.


Installation

pip install axionflow

Python 3.10+ required.


Basic usage

Initialize / Review a project

axion review .

Generate a technical plan

axion plan "Refactor authentication to use JWT with refresh tokens"

Solve a concrete problem

axion solve "Fix silent failures in the reasoning engine error handling"

All commands follow the same pipeline:

Analyze → Plan → Execute → Validate

Extending Axion

Axion is designed to be extended by engineers.

  • Custom tools
  • Custom validators
  • Custom reasoning steps

See the Plugin Development Guide.


Contributing

Axion is not looking for volume. It is looking for engineering-quality contributions.

If you care about:

  • clean abstractions
  • predictable systems
  • readable diffs
  • testable behavior

You’ll fit right in.

See CONTRIBUTING.md.


“Understanding the internals is the first step to excellence.”

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

axionflow-1.4.0.tar.gz (37.7 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

axionflow-1.4.0-py3-none-any.whl (39.2 kB view details)

Uploaded Python 3

File details

Details for the file axionflow-1.4.0.tar.gz.

File metadata

  • Download URL: axionflow-1.4.0.tar.gz
  • Upload date:
  • Size: 37.7 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for axionflow-1.4.0.tar.gz
Algorithm Hash digest
SHA256 3abbd275c3c0144035d0a58bcaf6f0587887d476522442a2885294870d0da323
MD5 9a197b240acbca8fe1d900c27e8fdbe4
BLAKE2b-256 e00ae56a13e7bc1d6d59cc2f1413bb8c41f4f79e287bfe66c14e2a1ad9e77931

See more details on using hashes here.

Provenance

The following attestation bundles were made for axionflow-1.4.0.tar.gz:

Publisher: publish.yml on KerubinDev/AxionFlow

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file axionflow-1.4.0-py3-none-any.whl.

File metadata

  • Download URL: axionflow-1.4.0-py3-none-any.whl
  • Upload date:
  • Size: 39.2 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for axionflow-1.4.0-py3-none-any.whl
Algorithm Hash digest
SHA256 f8291e3a680d34584cdf51d3ca0244f17ad6694f6845ee4af8622037f50c0c9e
MD5 1fbf4749aad76691190c26b4ca5d6f73
BLAKE2b-256 342c1d3f97fc70eb724b69a80966160d8007551898e4e9cec1a6ad81a7e81349

See more details on using hashes here.

Provenance

The following attestation bundles were made for axionflow-1.4.0-py3-none-any.whl:

Publisher: publish.yml on KerubinDev/AxionFlow

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page