Get a model response in one call.
Use AI in your application to summarize text or extract information from it. Send Ember the content and instructions, and it returns the model’s answer. You choose cheap, balanced or strong instead of a specific model; Gomry selects the model for that level.
curl https://api.cloud.gomry.com/v1/inference \
-H "Authorization: Bearer $GOMRY_CLOUD_KEY" \
-H "Content-Type: application/json" \
-d '{
"capability": "cheap",
"system": "Summarize in one sentence.",
"user": "Update the address before dispatch.",
"maxOutputTokens": 100
}'Summarize text in your application.
A support application could pass a help article with instructions to summarize the relevant steps for a customer’s question, then display the summary alongside the original article. Long input can be trimmed to the model’s context window, so supply the material needed for the answer within the selected grade’s limits.
Why use Ember instead of calling a model directly?
With a direct model integration, your application selects a specific model. With Ember, you choose a level and Gomry selects the model behind it. Your application keeps using the same request format when that selection changes, so you can add a summary or text-extraction feature without tying the integration to one model ID.
Your team still needs to evaluate responses against the application’s acceptance criteria when choosing a level or reviewing a model change.
Send your content and instructions.
{
"capability": "cheap",
"system": "Summarize the supplied help text in two sentences. Do not add information.",
"user": "Customers can update their delivery address before an order ships. After dispatch, they should contact support.",
"maxOutputTokens": 200
}Each call contains one system prompt and one user message. Your application manages any conversation context and subsequent actions. Ember is a single-response API, rather than a tool-executing agent or a streaming conversation service.
For structured output, an optional jsonSchema guides the model where supported; validate the result in your application. Check usage.truncated when the output limit is reached, especially before parsing JSON.
Input and output tokens have different rates.
| Capability | Per 1M input tokens | Per 1M output tokens |
|---|---|---|
| Cheap | $1.10 | $9.00 |
| Balanced | $5.50 | $28.00 |
| Strong | $5.50 | $28.00 |
Input and output usage are counted separately and reported with the response. Set an output-token limit appropriate to the answer you need; a response can be truncated at that limit.
For illustration, a request using 1,000 input tokens and 200 output tokens at the Cheap rates costs $0.0029: $0.0011 for input + $0.0018 for output. These token counts are hypothetical; billing follows the measured usage.
Build with Ember.
Enable Ember in the project and issue a key with inference permission. Send the key as a Bearer token to POST /v1/inference. Start with a small request whose answer your application can validate before expanding the input.