Get your first answer in an app
Connect a chat application to a model available through your Saylek account. Your first success is an answer in that application, not just a saved key or a model in a list.
This path covers apps with custom OpenAI-compatible chat settings. Using Claude Code? Follow the Claude Code setup, which uses a different address and configuration. A coding agent also needs a model that supports its tool use; a chat response alone does not prove that.
Before you start
- Sign in to Saylek. You do not need an invitation to create an account.
- Make sure a model is available to your account from your own machine or one shared with you. If you have no access yet, ask someone who runs a machine to invite you, or follow the hardware path. If you have already accepted an invitation, continue here.
- In your app, find the settings for a custom AI provider or API connection. You need fields for a base URL, API key and model. If it only supports its own service and offers no custom connection, this path does not apply.
You do not need to install Saylek for this hosted connection. Requests pass through Saylek's servers and the machine serving the model, even when that machine is yours. Use a non-sensitive prompt for setup. Privacy and egress explains the boundaries; a Saylek machine set to local only does not keep hosted API requests local.
1. Create a key for this app
Open Applications, choose Add application, and give it a recognizable name. Create the key and save its secret when shown; it is displayed once. Keep it out of screenshots, shared documents and source code.
2. Enter the connection settings
| App setting | Value |
|---|---|
| Provider or connection type | Custom OpenAI-compatible chat completions |
| Base URL | https://api.saylek.com/v1 |
| API key | The Saylek application key you just created |
| Model | An exact model ID available to this key, selected in the next step |
Field names vary by application. If your app requires a model before saving, retrieve its ID in step 3 first. Save or apply the connection, then refresh its model list if the app provides that control.
Example: an Open WebUI instance you administer. Its documented connection controls are under Settings → Admin → Connections → Manage OpenAI API Connections → Add Connection. Enter the URL and key above, verify the connection, then save. Use its discovered model IDs or enter the exact ID from step 3. These are administrator settings: on a shared instance, check who will be able to use the connection before adding your key. Follow Open WebUI's connection guide if your version's labels differ. The final test is still an answer in your app.
3. Choose an available model
Select an exact model ID returned by the Saylek connection. Do not leave a default model name from another provider selected.
If your app cannot list models, this optional terminal check retrieves the list for the same key. Replace YOUR_API_KEY with your key on your own computer; do not share the command after inserting the secret.
curl https://api.saylek.com/v1/models \
-H "Authorization: Bearer YOUR_API_KEY"
The response contains a data list. Copy an entry's id into the app's model field. An empty list means there is no model available through that key at that moment; changing app settings will not create capacity.
For this first check, choose a model intended for text chat. If you do not know what it supports, ask the machine owner. Model availability does not establish support for images, tool use or every app feature.
4. Get an answer inside your app
Start a new conversation using the Saylek connection and the selected model. Send:
Explain what a GPU does in one sentence.
Success: the selected connection returns an answer in the app, without an authentication or model error. Check that the app is still using the Saylek connection rather than another configured provider. This confirms basic text chat, not every feature of the application or the model's quality for your work.
Next, try one small, non-sensitive example of the task you actually need. For a coding agent, use the dedicated setup and a bounded tool-use check before trusting it with consequential work.
If you do not get an answer
| Symptom | Next check |
|---|---|
| Key rejected | Confirm you pasted the Saylek key into this connection. If it was lost or exposed, use Rotate API key in Applications and update the app with the new secret, or revoke the old key and create another. |
| Model not found | Refresh models using the same key and address; copy the exact ID. Check that the serving machine remains available. |
| Model listed, request fails or waits | A listing does not reserve capacity. Check the machine's availability and the actual error before retrying. |
| API test works, app does not | Recheck the app's selected provider, base URL and model. API success alone does not verify the app's configuration. |
| Text chat works, another feature fails | Check that feature's model and API support; do not assume the connection test covered it. |
For a persistent failure, use Troubleshooting. Record the app name, selected model, time and error text, without the key or private prompt.
To stop using this connection: switch your app back to its previous provider and remove the Saylek key from its configuration. Manage or revoke the application key in Applications; removing it from the app alone does not revoke it.
Last checked 2026-10-25 · read as markdown at /docs/first-answer-in-an-app.md