Together AI
proxium calls Together AI with your own Together AI key. Your applications send their requests to proxium with a proxium virtual key. They never hold the Together AI key.
| Setting | Value |
|---|---|
| Preset id | together |
| Base URL | https://api.together.xyz/v1 |
| How proxium sends the key | an Authorization: Bearer header |
| Request format | the OpenAI Chat Completions format |
| Request type | chat |
Add Together AI to a project
- Get a key from Together AI at api.together.ai/settings/api-keys.
- In the console, open Providers and select Add a vendor.
- In What are you adding, select Together AI. proxium fills in the base URL and the models.
- Paste the key into API key, and select Add Together AI.
proxium stores the key encrypted. The console does not show the key again.
Models
A new Together AI vendor starts with a list of its models. You can change the list on the Providers screen.
Call a model
A vendor that you add has the provider id t.<project>.together, where <project> is the project slug. Send GET /v1/models to get the exact ids that your key can use, and put one in place of <model>.
curl https://proxium.tech/v1/chat/completions \
-H "Authorization: Bearer $PROXIUM_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "t.<project>.together/<model>", "messages": [{"role": "user", "content": "Hello"}]}'
You can also send a tier name, such as trial, in model. proxium then picks the model from the routing policy of the project. Refer to Routing.
Ready-made routing
proxium suggests a routing for a project with Together AI. When you choose it on Routing, a request with "model": "auto" works at once: a classifier picks the tier, and the tier picks the model. The Routing screen shows the models of each tier. A tier that you set yourself keeps your models. Refer to Automatic routing.