Skip to main content
Odysseus is a self-hosted AI workspace designed to provide a privatized Web experience similar to ChatGPT and Claude. It not only supports local models but also seamlessly integrates with cloud APIs (such as Mixroute), featuring powerful Agent routing, Deep Research, collaborative document editing, persistent memory (ChromaDB), and an evolving skill system.

Core Features

Smart Agent Routing

Built-in autonomous Agents based on tool calling, web search, file read/write, and Shell execution, supporting MCP service extensions.

Deep Research

Multi-step, autonomous academic and factual search, synthesis, and extraction, ultimately outputting comprehensive, high-quality visual reports.

Self-Hosted AI Collaboration

Equipped with a rich-text Markdown editor where AI can collaboratively edit documents based on context, perfectly optimized for mobile devices.

Prerequisites

Before starting deployment, please ensure your system meets the following requirements:
1

Prepare Mixroute Credentials

  1. Obtain your Mixroute API endpoint URL, typically: https://api.mixroute.ai/v1
  2. Generate your API Key in the Mixroute console.
  3. Determine the model name you want to use, e.g., claude-sonnet-4-6 (make sure this model is enabled in the Mixroute console).
2

Environment Dependency Check

  • Docker Deployment: Requires Docker and Docker Compose.
  • Native Linux Deployment: Requires Python 3.11 or higher, and ensure git is installed on your system.

Deploying Odysseus

You can choose between the recommended Docker containerized deployment or native Linux deployment based on your system configuration.

First Login & Password Retrieval

By default, Odysseus automatically creates an admin account admin during the first startup and prints the generated random initial password to the startup logs.
1

Retrieve Initial Admin Password

Run the following command in your terminal to view the output:
  • Docker Deployment:
    docker compose logs odysseus | grep -C3 'password'
    
  • Native Linux Deployment: Check the temporary initial password output in uvicorn or your startup terminal.
2

Access Web Console

Log in using the username admin and the password you just retrieved.
  • Local access: http://127.0.0.1:7000
  • Remote cloud server access: http://<your-server-ip>:7000
By default, the container only listens to 127.0.0.1. If you need to configure a cloud virtual machine, please refer to the Nntwork errors or container listening issues below.

Configuring Mixroute

Once logged in successfully, follow these steps to configure Mixroute as a custom endpoint to power your Odysseus intelligent assistant.
1

Open Settings Panel

Click Settings in the bottom-left corner to access the global system settings.Odysseus Seting
2

Configure Provider

Add a new custom endpoint on the Providers/Endpoints page in Settings:
  • Base URL: https://api.mixroute.ai/v1
  • API Key: Enter your Mixroute API Key (e.g., sk-xxxxxxxx) Odysseus Add Provider
3

Enable & Manage Models

After adding the endpoint, Odysseus will automatically probe and load the available models returned by that endpoint. You can enable frequently used models in the list, or assign them to global default roles (such as default chat, Deep Research, Task scheduling, vision, etc.).
Testing shows that the GPT series performs exceptionally well in Odysseus; using it in conjunction with GPT models is highly recommended.
Odysseus Models Enabled
4

Enjoy your Odysseus

Return to the Chat home page to start conversations and launch powerful intelligent Agents utilizing Mixroute’s underlying models.Odysseus Finish

FAQ & Advanced Setup

If you need to call or manage local GPUs through the Cookbook inside self-hosted containers, we recommend:
  1. Running the built-in diagnostic and configuration script:
    scripts/check-docker-gpu.sh --install-nvidia-toolkit --enable-nvidia-overlay
    
  2. This script automatically appends COMPOSE_FILE=docker-compose.yml:docker/gpu.nvidia.yml to .env to enable NVIDIA Container Toolkit mapping.
By default, the container only listens on 127.0.0.1.
  • If you need direct access via IP over LAN or the public network, please modify APP_BIND=0.0.0.0 in .env and restart the container.
  • For production environments, it is highly recommended to keep the listening address on 127.0.0.1 and use Nginx, Caddy, or Cloudflare Tunnel as a reverse proxy with an SSL certificate.
If you encounter issues where ChromaDB fails to load or experiences client incompatibilities:
# We recommend manually uninstalling potentially conflicting lightweight clients, then force-reinstalling the full package:
./venv/bin/pip uninstall chromadb-client -y
./venv/bin/pip install --force-reinstall chromadb
Once configured, all of Odysseus’s subsequent features (Deep Research, file retrieval, calendar/todos, and email triage) will be powered by the Mixroute endpoint bound and configured in your Settings, delivering a robust and highly cost-effective cloud-based LLM experience.