Browse documentation

Overview · Launch preview. EvaEsi has not launched.

Developer documentation

Guides and API references for working with EvaEsi: one model, rooms and keys, responses, sources, usage and credit.

On this page

Launch preview. EvaEsi has not launched. These guides describe the API workflow at launch; API accounts and keys are not available from this website. Contact Forlais about API access.

Documented launch contract: The canonical public inference route is POST /api/v1/responses. The older /api/v1/chat route is documented as a compatibility alias.

Key concepts#

One model, five levels

EvaEsi is one model, EvaEsi 1.0. Every request goes to it. The level field sets how much work goes into the answer: look, focus, duet, high or hyper. See EvaEsi 1.0.

Rungs, not tokens

EvaEsi bills and reports usage in rungs. A rung is one character of answer at Focus. Each level uses a set number of rungs per character of answer, and input is priced at 30% of output.

  • At Focus, a 1,200-character answer is 1,200 rungs
  • At Look it is 600 rungs; at Duet 3,000; at High 4,200; at Hyper 5,400
  • Input rungs and output rungs are reported separately on every response

Rooms and keys

The API works in rooms. A room is one EvaEsi, with its own memory, files and storage, shared by any number of API keys. Give each of your users a room and they keep their context across your products. Rooms are free, with 1 GB of storage each; more storage is 2p per GB a month. Rooms are created and managed in the API portal, and export or deletion of a room is free.

Prepaid credit

You pay from prepaid credit, with automatic recharge and spend limits you set. A new account starts with a spend tier of £100 a month, which rises with payment history and account age. The Developer plan includes £5 of credit every month. Prices are on the pricing page.

Session APIs vs bearer APIs

The documented launch workflow for the API portal at evaesi.dev uses signed-in session routes under /v1/*. Programmatic clients use bearer routes under /api/v1/*. Portal access opens with the platform at launch.

Public API#

EndpointDescription
POST /api/v1/responsesCanonical single-request inference route
GET /api/v1/sources/{request_id}Live source bundle for a prior response
GET /api/v1/usageCurrent billing-period usage totals
GET /api/v1/usage/historyHistorical usage periods, newest first
GET /api/v1/limitsLive request, rung and spend envelope
GET /api/v1/modelsThe model and its levels
GuideDescription
QuickstartA room, a key and your first request at launch
AuthenticationAPI keys, scopes and rotation
EvaEsi 1.0The levels, their rungs and prices
Responses APICanonical public inference route
Sources APILive source lookup for prior requests
Usage APIUsage, history and limits response shapes
Limits and creditRequests at a time, spend tiers and header behaviour
PricingPrices by level, rooms, storage and the Developer plan

Base URL#

Base URLBase URL
https://evaesi.dev/api/v1

Authentication#

All public API requests require a bearer key:

ExampleAuthentication
Authorization: Bearer esi_dev_YOUR_API_KEY
Product vs API auth: Browser sessions manage your account, credit, rooms and API keys inside the API portal. API calls from your code always use bearer keys, never cookies.