ADAM — Onchain AI Agent Platform (Work in Progress)

Note: This project is in early development. The current codebase provides foundational Python automations and infrastructure for an advanced onchain AI agent. Core model integration and the tuning engine are not yet implemented. Expect rapid changes as we evolve toward a highly efficient onchain collaborator.

---

Vision

ADAM aims to be a modular, production-ready framework for building advanced onchain AI agents. The platform will combine a Live Context Engine (LCE), flexible storage, provider-agnostic embedding, and server-side automation—serving as the backbone for agentic applications that operate transparently and efficiently onchain.

---

Current Highlights

• Live Context Engine (LCE): Session-aware context ingestion, embedding, retrieval, and selection.
• Provider-Agnostic Embedding: Local and OpenAI examples included.
• Lightweight Storage: SQLite adapter with schema for sessions, atoms, events, profiles, and token accounting.
• Automation Endpoints: Managed Playwright browser pool for interactive flows.
• Testing: Smoke checks and examples to get started quickly.

Planned: Base model integration, onchain interaction modules, and a robust tuning engine for efficient, adaptive collaboration.

---

Repository Layout (High Level)

• av_agent/ — main package
  • live_context/ — LCE, DB adapter, embedders, policies, selectors, strategies, trackers, tests
  • web/ — Flask app and automation API
  • tools/ — utility tools (Playwright pool, helpers)
  • storage/ — key-value and vector DB adapters
  • llm/ — LLM provider abstractions
  • tests/ — test suite
• docs/ — design docs, architecture, and developer guide
• requirements.txt — pinned dependencies

---

Quickstart (Developer)

1. Create and activate a venv (Windows PowerShell):

   python -m venv .venv
   .\.venv\Scripts\Activate.ps1
   pip install -r requirements.txt

2. (Optional) Install Playwright and browsers:

   pip install playwright
   playwright install chromium

   Or rely on the repository post-install step when running npm install in the project root (this runs npx playwright install chromium). See docs/PLAYWRIGHT.md for more details.

3. Run tests:

   pytest -q

4. Run the Flask app locally:

   $env:PYTHONPATH='L:\\worxpace\\ADAM'
   python -m av_agent.web.flask_app

---

Documentation

See the docs/ folder for:

• system-specs.md — runtime and environment requirements
• architecture.md — component diagrams and interactions
• lce-specs.md — detailed LCE data model and algorithms
• developer-guide.md — contribution, testing, and extension notes

---

Roadmap

• Python automation and context engine foundation
• Base model integration (coming soon)
• Onchain interaction modules
• Tuning engine for adaptive, efficient collaboration
• End-to-end agentic workflows

---

Contributing

• Run tests and keep them green.
• Add documentation for new modules under docs/.
• Open a PR with a clear description and tests for behavior changes.

---

License

This project is MIT licensed. See the LICENSE file for details.

---

This README is not finalized. We are actively building out the core agentic and onchain capabilities. Stay tuned for major updates as we progress toward a fully operational onchain AI collaborator.