Agentic AI Settings
Fine-tune AI credit limits, agent behaviour, guardrails, and AI usage for the agentic pipeline and workflow execution engine. Model and engine selection lives on the AI Models tab.
AI Credits
Every AI request costs credits according to how heavy the model is — not by its token count — so a cheap and a premium model are easy to compare. The limits below apply to your whole organization and are enforced on the server for every AI feature. A limit of 0 means unlimited.
| Model type | Credits per request |
|---|---|
| Fast / low-cost models, Semantic Search embeddings and rerank | 1 |
| Standard chat models | 5 |
| Premium / reasoning models | 10 |
| Option | Description |
|---|---|
| Per-request credit limit | The highest credit cost one request may have. Models that cost more than this are refused — for example, a limit of 5 blocks premium models. |
| Daily credit limit | Credits the organization can use per day. Resets at midnight UTC. |
| Monthly credit budget | Credits the organization can use per calendar month (UTC). |
The card also shows how many credits were used today and this month, with a progress bar when a limit is set. Everyone can see the usage; only tenant administrators can change the limits. A refused request is not charged, and only successful requests use credits.
These limits work inside your plan's monthly credits (credits per seat × members). Only hosted AI uses credits — local models (Ollama, Local LLM, in-browser) never do. For the credits each plan adds and the Admin → Usage screen, see AI Credits & Usage.
Screen preview
A quick look at the screens. Use the toggle in the corner of each preview to switch between the dark and light themes.
Runs entirely on your machine
Runs in the browser tab, zero install
Anthropic, OpenAI, Google & more
Self-hosted server on your own network
Provider keys stay on the server, never in the browser
Each workspace picks its own default engine
One engine setting for every AI feature
Agentic settings
Agent Behaviour
| Option | Values | Description |
|---|---|---|
| Retry on failure | 0 – 5 | How many times to automatically retry a failed AI call before surfacing an error to the user. |
| Temperature | 0.0 – 1.0 | Controls output creativity. Lower values produce more deterministic, factual text. Higher values introduce more variation and creativity. |
| Streaming | On / Off | When enabled, AI output streams token-by-token so the user sees text appearing in real time rather than waiting for the full response. |
| Enable logging | On / Off | Logs every AI request and response to the server log file. Useful for debugging but increases disk usage. |
Temperature guide
For factual tasks (data summaries, code analysis) keep temperature at 0.2 – 0.4. For creative writing (blog posts, stories) raise it to 0.7 – 0.9.
AI Governance policy
The General sub-tab ends with the AI Governance policy card. It sets tenant-wide controls that the server applies to every AI request your organization sends to a model provider. It replaces the former on-screen guardrail switches, which did not change any behaviour.
| Control | What it does |
|---|---|
| Sanitize prompts | Removes hidden characters, script blocks and known prompt-injection phrases before a prompt is sent. Choose to remove them and continue, or to block the request. |
| Mask personal data in prompts | Replaces emails, phone numbers, payment cards, IBANs, US SSNs and T.C. kimlik numbers with tokens before the text leaves TotalApp. Also covers embedding and rerank requests. |
| Mask personal data in responses | Masks the same types in the answer before it is shown. |
| Block listed terms | Declines requests, and withholds responses, that contain a term from your list (one per line). |
| Record AI activity | Writes a metadata record for every governed request. Optionally adds a masked 200-character prompt preview. |
| Export audit trail | Downloads the records for a date range as a CSV file. |
Off by default, administrators only
Every control starts switched off. Only tenant administrators can view or change the policy; other users see a notice. For what each control covers and its limits, see AI Governance & Access; for how the controls are enforced, see AI Security.
Where the model settings live
This tab holds the general AI settings: usage limits, agent behaviour and the AI Governance policy. Choosing the AI engine, the embedding model, the rerank model and the other model lists now happens on the separate AI Models tab, right next to this one.