aurora-actr 0.5.1

AURORA: Adaptive Unified Reasoning and Orchestration Architecture with MCP Integration

AURORA

Version 0.5.1 | PyPI | Commands | SOAR Reasoning | ML Models

Adaptive Unified Reasoning and Orchestration Architecture - Project-based AI memory and reasoning system.

What It Does

AURORA adds persistent memory and structured reasoning to AI coding tools (Claude Code, Cursor, etc.).

Three core capabilities:

1. Multi-turn SOAR Reasoning - 9-phase cognitive pipeline for complex queries with automatic escalation
2. Planning & Agent Discovery - OpenSpec-adapted workflow orchestration with agent registry
3. ACT-R Memory - Tree-sitter AST indexing with activation-based retrieval for code, knowledge, and reasoning patterns

Retrieval modes:

• Lightweight: BM25 keyword + Git signals + ACT-R activation (default, ~520KB)
• Enhanced: Add semantic embeddings with optional ML package (~1.9GB)

Storage: Project-local .aurora/ directory (SQLite database, no cloud required).

Installation

PyPI (Recommended)

pip install aurora-actr

With Semantic Search (Optional)

pip install aurora-actr[ml]  # Adds PyTorch + sentence-transformers (~1.9GB)

Development

git clone https://github.com/amrhas82/aurora.git
cd aurora
./install.sh

Quick Start

# Initialize in your project
cd your-project/
aur init

# Index your codebase
aur mem index .

# Search indexed memory
aur mem search "authentication logic"

# Multi-turn SOAR reasoning
aur soar "How does the payment flow work?"

# Create implementation plan
aur plan create "Add user authentication"

# Health check
aur doctor

What Gets Indexed

AURORA indexes three types of chunks:

• code - Python functions, classes, methods (tree-sitter AST parsing)
• kb - Markdown documentation (README.md, docs/, PRDs)
• soar - Reasoning patterns (auto-saved after aur soar queries)

Default exclusions: .git/, venv/, node_modules/, tasks/, CHANGELOG.md, LICENSE*, build/, dist/

Custom exclusions: Create .auroraignore (gitignore-style patterns):

# .auroraignore example
tests/**
docs/archive/**
*.tmp

Retrieval Strategy

Hybrid scoring (default, no ML required):

• 40% BM25 keyword matching
• 30% ACT-R activation (usage frequency + recency)
• 30% Git signals (modification patterns)

With ML option ([ml]):

• 30% BM25 keyword matching
• 40% Semantic similarity (sentence-transformers)
• 30% ACT-R activation

Speed: Sub-500ms on 10K+ chunks.

Documentation

• Commands Reference - Complete CLI command documentation
• SOAR Reasoning - 9-phase cognitive pipeline details
• ML Models Guide - Custom embedding model configuration
• MCP Deprecation - Why MCP tools were deprecated
• Migration Guide - Migrating from MCP tools to slash commands

Architecture

Cognitive Foundations:

• ACT-R activation-based memory (cognitive science)
• SOAR 9-phase reasoning pipeline (State/Operator/Result)
• Tree-sitter for accurate AST parsing

Design Principles:

• Project-local (no cloud, .aurora/ directory)
• Lightweight by default (BM25 + activation)
• Optional semantic search (ML package)
• No API keys for memory/search operations

Configuration

Indexing: aur init or aur mem index .

Excluding files: Create .auroraignore in project root

Changing ML model: See ML Models Guide

MCP integration: Deprecated in v0.5.0, use slash commands instead

Requirements

• Python 3.10+
• ~520KB (base install)
• ~1.9GB additional (with ML features)

License

MIT License

Contributing

See CONTRIBUTING.md