Getting Started - Web Dashboard¶
The URSA web dashboard provides a browser-based interface for running URSA workflows.
Screenshots forthcoming
A complete dashboard walkthrough should include screenshots. This page currently covers installation and launch commands; a fuller visual guide can be added later.
Install dashboard dependencies¶
Install URSA with the dashboard extra:
uv pip install "ursa-ai[dashboard]"
python -m pip install "ursa-ai[dashboard]"
Launch the dashboard¶
ursa-dashboard
By default this serves on 127.0.0.1:8080.
You can set the host, port, group, and initial config file:
ursa-dashboard \
--host 127.0.0.1 \
--port 8080 \
--group default \
--config config.yaml
The config file initializes the dashboard LLM endpoint settings.
Configure API credentials¶
Open Settings -> LLM and choose an API-key source:
- Secure system storage stores the key in macOS Keychain, Windows Credential Manager, or the available Linux keyring service. The key field is always blank when Settings opens; the dashboard reports only whether a usable key is configured.
- Environment variable retains the existing headless and automation workflow. Enter the variable name, not its value.
- No API key is appropriate for endpoints that do not require one.
Embedding credentials are configured independently under Settings -> Embedding/RAG, or can explicitly reuse the saved LLM key when both configurations resolve to the same provider or endpoint origin. Saved keys are bound to the configured provider or endpoint origin. After changing the endpoint host, save the key again to approve its use with that host.
The raw key is never written to dashboard settings, sessions, run records, or worker configuration files. In remote dashboard mode, credential changes must be served over HTTPS.
Choose a session workspace¶
Every new dashboard session requires an explicit workspace choice. Select a
folder you can find and reuse, or choose Temporary workspace for disposable
work. Temporary mode behaves like ursa --workspace tmp: the dashboard creates
the workspace in the operating system's temporary directory and removes it when
the session is deleted or the dashboard stops.
The dashboard no longer creates a hidden default session workspace under
~/.cache/ursa. Older sessions without an explicit workspace remain available,
but the dashboard prompts for a folder or temporary workspace before their next
run. When an older session still has its former UUID-named cached workspace, the
folder field is prefilled with that path so existing work remains easy to
recover. A user-selected folder is never deleted when its dashboard session is
deleted.
Launch an agent team or symposium¶
Open Environment runs from the dashboard sidebar. The page provides New team and New symposium actions that let you edit a starter YAML definition, optionally choose a unique Run ID, enter the task prompt, validate the configuration, and launch it directly from the browser. Reuse the same environment name with a new Run ID for follow-on work that should retain the team or symposium workspace while creating a separate replay.
The run is queued in the background and appears on the page immediately. Open it to follow the environment graph and work timeline live, inspect the final result, or cancel a dashboard-launched run. The dashboard uses the LLM, credential, and timeout settings configured under Settings. Member-specific model blocks may refer to API-key environment variable names, but literal API keys are rejected and never stored in YAML or run metadata.
Headless Linux
Secure system storage requires an available desktop keyring service. Headless deployments should continue to use environment variables or a deployment-managed secret provider.