Configure the runtime just enough to start working. This page covers the practical minimum for cloud providers, local Ollama usage, and the optional setup commands developers usually need next.
You do not need to configure every provider or every tool on day one. One working provider, one working model, and a successful seshat chat session is enough to begin.
The most direct setup is to export an API key and tell Seshat which provider and model to use. Replace the provider or model with whatever is valid for your environment.
export ANTHROPIC_API_KEY=sk-ant-...
seshat config --provider anthropic --api-key "$ANTHROPIC_API_KEY"
seshat config --model anthropic:claude-sonnet-4-20250514Before debugging anything, print the current config. This is the fastest way to confirm provider, model, and stored settings.
seshat config --printIf you want a local model, start Ollama first and then point Seshat to an Ollama model name. The exact model is up to you.
ollama serve
seshat config --provider ollama
seshat config --model ollama:llama3.1Some document workflows rely on additional Python-backed tooling such as Docling. You do not need this for a simple first chat, but developers often want to verify it early.
seshat setup --check
seshat setupOne cloud provider and one known-good model are enough. Optimize for simplicity, not completeness.
Pick the model you actually want to use in projects, then validate it inside a real repo with the CLI.
Check provider, model, and optional tooling now so SDK and automation work later without hidden environment issues.
If you need per-user credentials, workspaces, or organization policies, that belongs to the SeshatOS platform layer.
Once configuration is stable, move to the first real workflow instead of adding more setup for its own sake.