Project settings

Open a project and click Settings at the bottom of the sidebar. The settings menu has eight tabs.

General

• Project name: shown throughout the console.
• Project ID: the stable identifier used in the API and in URLs. Click Copy to copy it.
• Request handling: two optional switches, both off by default:
  • Response caching: reuse stored responses for repeated requests instead of calling the model again. See Caching.
  • PII protection: block requests that contain emails, phone numbers, SSNs or card numbers. Blocked requests appear in Logs. Paid plans. See PII protection.
• Delete project: permanently removes the project and its data. This can't be undone.

Budget

• Current month spend: what this project's AI usage has cost so far this billing period.
• Monthly budget: the amount you expect to spend. You get alerts as you approach it.
• Spend cap: a hard limit. Turn on Enforce spend cap and requests are blocked once the project reaches it.
• Budget alerts: emails at 50%, 80% and 100% of the budget.

Providers

• Default provider and default model: used when a request doesn't name a model.
• Rate limiting: requests per minute, tokens per day and concurrent requests for this project. See Rate limiting.
• Fallback: turn on automatic fallback to send a request to a backup provider when the first one fails. Pick the fallback provider and model (leave the model empty to match the original model automatically), and how many retries to try first. See Routing and failover.
• Circuit breaker: stop sending traffic to a provider that keeps failing. Set the failure threshold (failures in a row before it trips) and the recovery timeout (seconds before trying the provider again).
• Requests by geography: where this project's requests come from, by country.

Infrastructure

A read-only view of how this project is served: its endpoint, live availability and latency for each configured provider (click refresh to check), and the request limits and defaults in effect. Change limits and defaults on the Providers tab.

Networking

• API base URL and region: where your app sends requests.
• Access mode:
  • Public ingress (default): requests are accepted from anywhere, with a valid API key.
  • Restricted ingress: only requests from the IP addresses or ranges you list are accepted. Add one address or CIDR range per line. Everything else is refused before it reaches a model, so it costs nothing.

Integrations

Connect outside services to the project. Under FinOps, connect CostGraph to send usage and cost data. See CostGraph. Missing a tool? Use Suggest an integration.

Agents have their own integrations (calendars, stores, payments), set up on each agent's Tools page. See Agent tools.

API

Keys that let your app call Cencori.

• Secret keys (csk_…): for your server. Keep them private: never put one in browser code, a mobile app or source control.
• Publishable keys (cpk_…): safe for browser code, but only work from the allowed domains you set when creating the key, for example localhost, *.myapp.com.

Click New secret key or New publishable key, give it a name, and copy it from Key created: it's shown once. To replace a key, create a new one, update your app, then revoke the old one. If a key leaks, revoke it straight away.

Webhooks

Send events to your own endpoint as they happen. Create webhook, give it a name and an endpoint URL, and choose events: