Text generation models
Models and providers
T3 Code writes thread titles, commit messages, PR titles and descriptions, branch names, and voice transcript cleanup. You choose the model in Settings → Threads → Text generation model.
T3 Code docs: text generation
What it’s for
Switching to a text generator, a model API that only writes text. It is usually faster and cheaper than starting a coding-agent session for each request.
T3 Code docs: text generation
When to use it
Use this when your main provider can't write this text, as with Registry ACP agents or fx.
T3 Code docs: providers acpproviders fx
How to use it
Go to Settings → Providers → Text generators and choose Add generator. Pick OpenAI, Anthropic, OpenRouter, or Custom (OpenAI-compatible).
T3 Code docs: text generation
In the model picker for each text task, you can switch that task to the new generator or keep what it uses. Use Test next to a picker to have the chosen model answer a tiny prompt, which proves the key works.
T3 Code docs: text generation
On a remote environment, localhost in a custom endpoint URL means the server machine, not your device.
T3 Code docs: text generation
Review guides always need a coding agent. Choose Codex or Claude in Settings → Source control & issues → Review guide model.
T3 Code docs: text generation
Theo’s tips
You can pick the harness, model, and effort level used for text generation, such as commit messages, thread titles, and PR descriptions.
You can even pick which harness model and effort level is being used for that text generation for things like commit messages, titles for threads, PR descriptions, and more.
What’s changed
- Model APIs can directly generate titles, commit and PR text, branch names, and transcript cleanup.Updated Nov 9
- Added instructions for generating thread titles.Updated Oct 23
Unofficial notes. Not affiliated with Theo or t3.gg.