Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
140 lines (75 loc) · 4.34 KB

CONTRIBUTING.md

File metadata and controls

140 lines (75 loc) · 4.34 KB
Raw

Contributing to ACI (Always-On Consciousness-Inspired AI)

First off, thank you for your interest in contributing!

This project explores a neuro-inspired, memory-centric architecture for an always-on, consciousness-inspired agent. There is a rough working implementation of the DMN-style loop already running in a Jupyter Notebook. The next major milestone is a reimplementation targeting:

  • Isaac Sim for grounded sensory experience and world interaction

  • Google Colab for accessible, reproducible experiments

Until that reimplementation lands, contributions that refine algorithms, interfaces, and evaluation strategies are very welcome.

Project Status

  • Current: A prototype DMN loop with module stubs and basic orchestration in a notebook.

  • In progress: Designing the memory system and grounding interfaces for Isaac Sim.

  • Near-term goals:

    • Define concrete data schemas for memory nodes/edges.

    • Specify z_self update equations and calibration/safety metrics.

    • Implement HC expansion controls and compute budgets.

    • Provide Colab and Isaac Sim scaffolding.

How to Contribute

  • License: MIT

  • Workflow: Fork → Branch → Commit → PR

  • Requirements:

    • Explain the purpose of your change clearly in the PR description.

    • Include a brief rationale: what problem it solves, how it affects the system, any trade-offs.

    • If you modify algorithms, add notes on assumptions, expected complexity, and defaults.

    • If you introduce new parameters, document sensible defaults and bounds.

Good First Contribution Ideas

  • Algorithm refinement

    • Clarify and/or implement scoring features (coherence, novelty, epistemic gain).

    • Propose/implement neuromodulator gating policies (e.g., beam width schedules).

    • Draft consolidation thresholds and causal edge confidence calculations.

  • Memory schemas

    • Propose a minimal node/edge schema for episodic → semantic → autobiographical layers.

    • Sketch a hybrid storage approach (vector + graph + columnar attributes).

  • Safety and calibration

    • Define a simple calibration loop (e.g., reliability diagrams, Brier/NLL updates).

    • Integrate safety penalties into decoding/selection with transparent logging.

  • Evaluation

    • Add ablation plans and metrics for identity coherence, abstraction emergence, and safety.

    • Provide small synthetic tasks for pipeline health checks.

Pull Request Guidelines

  • Keep PRs focused and reviewable.

  • Include:

    • Summary of change and motivation.

    • Any new configs, defaults, or interfaces.

    • Usage examples or test snippets if applicable.

    • Backwards-compatibility notes if relevant.

  • If introducing dependencies or affecting performance budgets, call that out explicitly.

Code and Documentation Style

  • Prefer clear, modular code over cleverness.

  • Document key functions/classes with short docstrings describing inputs, outputs, and side effects.

  • Log key signals where useful (e.g., candidate scores, safety penalties, neuromodulator vector) to support reproducibility and diagnosis.

  • For pseudocode/spec contributions, use concise, unambiguous descriptions and default values where possible.

Community Standards

  • Be respectful and constructive.

  • Critique ideas, not people.

  • Prefer proposals with testable claims, measurable metrics, or minimal examples.

Questions and Discussion

  • If uncertain about direction, open an issue before large changes.

  • For algorithmic suggestions, include a brief literature pointer or rationale when possible.

How to Submit a PR

  1. Fork the repository.

  2. Create a feature branch from main.

  3. Commit changes with clear messages.

  4. Open a Pull Request to main.

  5. In the PR:

    • Explain your change meaningfully.

    • Describe how you tested it (or how it can be tested).

    • Note any follow-up work you recommend.

Thanks again for contributing---thoughtful refinements now will significantly accelerate the Isaac Sim and Colab reimplementation and help make the architecture robust, testable, and useful to the broader research community.