Admin Console › General
Path: Avatar dropdown → Administration → General
Section: Runtime & Interface
Section titled “Section: Runtime & Interface”| Field | Type | Range / Values | Description |
|---|---|---|---|
| Tool Loop Timeout (Seconds) | Number | 60 – 3600 | Maximum duration the agent may spend on tool calls. Default: 600 s. |
| Request Rate Limit | Number | 1 – 500 | Limits provider calls per minute to avoid HTTP 429 errors. Default: 10. |
| System Timezone | Text | IANA format, e.g. Europe/Berlin | Default timezone for cron jobs and audit logs. |
These settings apply globally and override individual user settings. Save via [Apply].
Section: Response Streaming
Section titled “Section: Response Streaming”| Field | Type | Description |
|---|---|---|
| Stream responses | Switch | Streams agent responses token by token into the chat while the model generates them. Default: on. Takes effect immediately (no [Apply] needed). |
Applies to the Anthropic API path and all OpenAI-compatible providers. CLI providers always respond as a block. Individual providers without SSE support can be excluded via provider metadata (
"stream": false). Details: Runtime & UI.
Section: Prompt Caching (Anthropic API)
Section titled “Section: Prompt Caching (Anthropic API)”| Field | Type | Description |
|---|---|---|
| Enable prompt caching | Switch | Enables cache_control on the Anthropic API path. Default: on. Takes effect immediately (no [Apply] needed). |
Controls only the Anthropic API path. Disable for sporadic single-shot usage with no visible cache savings (⚡). Since gpt-5.6, OpenAI also bills a cache-write premium (1.25×) — but caching there is implicit and this switch has no effect on it; watch the cost via the ⚡ badge in the trace. Details: Runtime & UI.
Section: Prompt Optimizer
Section titled “Section: Prompt Optimizer”| Field | Type | Description |
|---|---|---|
| Provider | Dropdown | Selects the AI provider for automatic prompt improvement. |
| Model | Dropdown | Selects the model within the chosen provider. Only available after a provider is selected. |
Save via [Apply].
Section: Summarizer
Section titled “Section: Summarizer”| Field | Type | Description |
|---|---|---|
| Provider | Dropdown | AI provider for the summarizer LLM call. |
| Model | Dropdown | Model within the chosen provider. Only available after a provider is selected. |
| Token Threshold | Number | Total tokens of all chat messages above which compression triggers. Default: 32,000. |
| Minimum Plaintext Window | Number | Number of most recent messages always kept as full text. Default: 20. |
Save via [Apply]. Without provider and model configured, compression remains inactive. See Context Compression for details.
Section: Message of the Day
Section titled “Section: Message of the Day”| Field | Type | Description |
|---|---|---|
| (text area) | Textarea | Message displayed on the chat start page for all users. Supports Markdown. Leave empty to display the selected agent’s description instead. |
Has its own [Save Message] button — independent of the global Apply button.