Odysseus

Why deploy odysseus render on Render?

Odysseus is a self-hosted AI workspace that bundles chat, agents, deep research, documents, email, notes, and calendar into a single instance you run with your own LLM API keys. This Render Blueprint deploys the hosted build along with bundled SearXNG (private web search) and ChromaDB (vector store for RAG), so you get a private, authenticated AI environment without wiring up separate search and memory infrastructure. Note that the Render deployment omits local-model serving, GPU inference, and host-Docker features from the upstream project, relying on cloud LLM APIs instead.

This template wires Odysseus together with its two dependencies already configured as private services—SearXNG for web search and ChromaDB for vector storage—so RAG, semantic memory, and Deep Research work on first boot without you standing up separate containers, networking them, or managing a search API key. Render's Blueprint provisions all three services from one click with a persistent disk mounted at /app/data, auth enabled by default with secure cookies and a generated admin password, and the helper services locked to internal-only access—so the app is the sole public endpoint without any manual firewall or reverse-proxy work. You just drop your LLM and search API keys into the deploy form as secrets and skip the entire Docker Compose, service discovery, and secrets-management setup you'd otherwise hand-roll.

Architecture

What you can build

After deploying, you'll have a self-hosted AI workspace running on your own Render instance, accessible at a private URL and locked behind admin authentication by default. You can open a chat and get replies immediately once you've added your own OpenAI key, run web-sourced Deep Research through a bundled private search backend, and work with documents, notes, email, and calendar in one place—all using API keys you control. Vector storage for RAG and semantic memory runs alongside the app as a private service, so retrieval-based features work without any external database setup.

Key features

• Three-service Blueprint: One-click deploy provisions the web app plus private bundled SearXNG and ChromaDB services, with only the web app publicly exposed.
• Bundled ChromaDB RAG: Includes a private ChromaDB vector store for retrieval-augmented generation and semantic memory with no external setup.
• Local embeddings via fastembed: Embeddings run locally through fastembed, so the OpenAI key only needs chat-completions permission and no embeddings endpoint is used.
• IP-based demo rate limiting: Demo mode enforces per-minute, per-session, and per-IP daily caps keyed on the trusted X-Forwarded-For client IP so cookie-clearing can't bypass limits.
• Auto-generated admin auth: Auth is enabled by default with secure cookies and a randomly generated 256-bit admin password exposed only via the Render environment, never logged.

Use cases

• Solo founder deploys private AI chat workspace with own OpenAI key
• Researcher runs sourced Deep Research reports via bundled SearXNG search
• Indie dev opens public no-signup chat demo with cost caps
• Privacy-conscious team self-hosts agents and RAG on isolated instance

What's included

Prerequisites

• OpenAI API Key: Powers chat, agents, and research by authenticating LLM calls to OpenAI.
• Brave Search API Key: Enables Brave web search as an optional search provider (SearXNG is bundled by default).
• Tavily API Key: Enables the Tavily search provider as an optional web search backend.
• Serper API Key: Enables the Serper search provider as an optional web search backend.
• Google API Key: Enables Google Programmable Search as an optional web search provider (used with the search engine ID).
• Google Programmable Search Engine ID: The custom search engine ID (CX) that pairs with your Google API key for Google Programmable Search.
• Hugging Face Token: Grants access to gated Hugging Face models when required.

Next steps

1. Open the odysseus service URL and log in as admin using the password from Render Dashboard → odysseus service → Environment → ODYSSEUS_ADMIN_PASSWORD, then change it from the app — You should land in the authenticated workspace and see your new password accepted on the next login
2. Open Chat and send a test message — You should get a reply from the seeded OpenAI endpoint (default model gpt-5.6-sol) with no model wiring needed, confirming your OPENAI_API_KEY is working
3. Click Deep Research in the left-hand tools menu, enter a question, and run it — You should get a sourced report generated from live web results via the bundled SearXNG, confirming the private search service is wired up without any extra key

Resources

• Odysseus official site