MojaCode

A personal studio
for AI coding.

On the deskIndex

02 / Own model

Choose the model. Know the arrangement.

“Own model” may mean a model you run yourself or an endpoint you choose. Start by making that arrangement explicit; the practical responsibilities are different.

Name each part

Record the model identifier, who operates the endpoint, and the interface that sends requests to it. Confirm the expected API format and any required authentication using that setup's documentation. A familiar model name does not establish compatibility, tool support, or context limits. Avoid assuming that two providers serve an identical configuration.

Prepare a controlled first request

Use a short, non-sensitive source file you can understand without assistance. Request an explanation with references to actual functions. Verify the result against the file before introducing more context or tools. If the endpoint fails, inspect the address, model identifier, credentials, and error response individually rather than changing every setting at once.

Separate inference from action

An endpoint generating text and an assistant executing commands are distinct capabilities. Check file, shell, and connector access in the application you select. A local model does not prove that every connected service stays local. Verify current data handling and configuration before including confidential material; this studio guide is not a privacy audit.

A local endpoint

Confirm the runtime is listening at the address you configured. Check the model is loaded and the hardware can run it. You maintain the model service; separately review any network access introduced by tools or connectors.

An endpoint record

Model identifier: [exact name]
Endpoint operator: [you or provider]
Base URL and API format: [verified in documentation]
Context and tool support: [verified or unknown]
First task: explain one non-sensitive source file
Continue the practiceContext desk