Odysseus Workspace
Self-hosted AI workspace & agent environment — local-first alternative to ChatGPT & Claude with deep research, model benchmarking, and persistent memory.
Python / FastAPI
Node.js
ChromaDB
vLLM / Ollama
Docker GPU
Autonomous Agent Engine
Equip agents with toolchains (MCP, file ops, shell, skills) to execute complex, multi-step tasks autonomously on local hardware.
Deep Research Pipeline
Multi-step research runs that systematically query, parse, and synthesize complex web/document sources into structured reports.
Blind Model Compare
Benchmarking suite to test local & API models side-by-side without bias, complete with automated synthesis and ranking.
ChromaDB Memory & Skills
Vector embedding memory store (ChromaDB + fastembed) providing persistent contextual recall and dynamic skill evolution over time.
# Odysseus GPU Docker Compose stack initialization
services:
odysseus:
build: .
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: all
capabilities: [gpu]
ports:
- "8080:8080"
Technical Specifications
| Feature | Implementation | Benefit |
|---|---|---|
| Inference Serving | vLLM / llama.cpp / Ollama | VRAM-aware hardware optimization |
| Vector Storage | ChromaDB (ONNX Embeddings) | Instant semantic document/memory RAG |
| Agent Protocols | Model Context Protocol (MCP) | Standardized secure tool integration |
| Mail & CalDAV Integration | IMAP / SMTP / CalDAV Sync | AI triage, reminders, & calendar agent |