aperion-cortex 0.1.0

The Cortex - Unified memory and state management service for Aperion

Aperion State Gateway — "The Cortex"

Unified memory and state management service for the Aperion AI platform.

The Cortex separates Reasoning (transient LLM processing) from Memory (persistent storage), providing a centralized service for context packing, vector search, and the event ledger.

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                        THE CORTEX                                │
│                     (Port 4949)                                  │
├─────────────────────────────────────────────────────────────────┤
│                                                                  │
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────────────────┐  │
│  │   HOT       │  │   WARM      │  │         COLD            │  │
│  │   STATE     │  │   CONTEXT   │  │        STORAGE          │  │
│  │             │  │             │  │                         │  │
│  │ - Sessions  │  │ - Token     │  │ - Vector DB (Chroma)    │  │
│  │ - Fast KV   │  │   Packing   │  │ - SQLite Archives       │  │
│  │ - TTL Cache │  │ - LLM-ready │  │ - Training Ledger       │  │
│  │             │  │   Output    │  │                         │  │
│  └─────────────┘  └─────────────┘  └─────────────────────────┘  │
│                                                                  │
└─────────────────────────────────────────────────────────────────┘
                            │
                            ▼
            ┌───────────────────────────────┐
            │     API Endpoints (REST)      │
            │                               │
            │  GET  /state/snapshot         │
            │  POST /vector/search          │
            │  GET  /db/schema/{name}       │
            │  POST /audit/record           │
            │  GET  /events/stream (SSE)    │
            └───────────────────────────────┘

Memory Tiering

Hot State

Active session data with low-latency access. In-memory with optional Redis backend.

Warm Context

The "Context Window" manager. Accepts raw messages/files and returns a token-optimized ContextPack ready for the LLM.

Cold Storage

Long-term persistence via Vector DB (ChromaDB/FAISS) and SQLite.

Quick Start

# Install dependencies
pip install -e ".[dev]"

# Run the service
aperion-cortex --port 4949

# Or with uvicorn directly
uvicorn aperion_cortex.service.app:app --host 0.0.0.0 --port 4949

API Endpoints

Endpoint	Method	Description
/state/snapshot	GET	Returns signed ContextPack
/vector/search	POST	Semantic search
/db/list	GET	List databases
/db/schema/{name}	GET	Database introspection
/audit/record	POST	Log agent execution
/events/stream	GET	SSE event stream
/healthz	GET	Health check

Docker

docker build -t aperion-cortex .
docker run -p 4949:4949 aperion-cortex

Environment Variables

Variable	Default	Description
CORTEX_PORT	4949	Service port
CORTEX_SIGNING_KEY	(required)	Key for signing ContextPacks
CORTEX_DB_PATH	data/cortex.db	SQLite database path
CORTEX_VECTOR_BACKEND	chroma	Vector store: chroma or faiss
CORTEX_EMBEDDING_MODEL	all-MiniLM-L6-v2	Sentence transformer model
CORTEX_LOG_LEVEL	INFO	Logging level

Development

# Install dev dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Run tests with coverage
pytest --cov=aperion_cortex --cov-report=html

# Lint
ruff check src tests

# Type check
mypy src

License

MIT