Manage users, app settings, and view usage statistics
| Name / Email | Role | Status | Joined | Actions |
|---|---|---|---|---|
| Loading... | ||||
Configure OpenID Connect for SSO with Azure AD, Okta, Keycloak, PocketID, Google, etc.
Callback URL:
Global Scribe instructions for the operations listed under each prompt. These are not Clinical Assistant prompts or Markdown formatting controls. Personal Memories and templates are separate and are not managed here.
Environment variables take precedence over DB settings. Storing password in DB is convenient but consider using env vars for production security.
Loading...
Click "Search API" to query your configured provider for all available models. Use the search box to filter results.
The result limit also bounds selected sources; visual queries separately request up to 8 multimodal candidates before selection. Excerpt size is not the conversation budget.
Loading server budget metadata...
Set by CLINICAL_ASSISTANT_CONVERSATION_CHARS, or the server default when unset; legacy saved settings are ignored. Counts all prior user/assistant text plus the new question using JavaScript string length (UTF-16 code units), not provider tokens. Over-budget requests stop before inference; no automatic clipping or summarization.
Text answer output limits: 2,600 tokens initially and 5,000 for completion retries. Image input is separate: existing image routes trim and clip the user prompt at 5,000 UTF-16 code units before adding poster/layout instructions. This is not a new image budget or output-token limit.
Overrides the default behavior for retrieved text answers (including streaming). Fixed greeting/no-source replies, search rewrites and explicit handoff instructions are separate, not controlled by this editor.
Fixed citation safeguards (read-only): Use only retrieved evidence for factual claims and keep the exact provided source numbers. Do not invent, renumber, merge or move citations. Runtime evidence/citation safeguards are appended separately; this editor does not replace them.
Overrides the default poster instruction appended to the image user prompt by both direct and background-job image generation. Existing automatic portrait/landscape layout suffixes still follow it. This does not change text answers or image-job storage.
Click Search to query available voices from your configured TTS provider.
Click Search to query available STT models from your configured provider.
Click Search to query available embedding models from your configured provider.
Reset all announcement, feature flag, and email settings back to factory defaults. SMTP configuration and custom models are preserved.