Repository navigation
FAQ: How do I use codeaf with Ollama (local models)? #2876
|
This is a maintainer-written FAQ, seeded from the questions people ask most. The answer below is marked as the accepted answer. If your case is different, reply here or open a new Q&A. I want to run codeaf against models on my own machine through Ollama, without an OpenRouter account. How do I set that up, and which model will it use? |
Replies: 1 comment
|
Ollama is a built-in provider and needs no key. Connect it
The connection check asks the local runner for its model list first. When it answers, its models show up under the provider's heading in Which model you get With only Ollama connected, only its installed models appear, and the opening model is an installed Ollama model (the first available chat model in provider order). Pick another one with No OpenRouter needed codeaf works with only a direct provider. Ollama counts as connected without a key, so pressing Adding a cloud provider later Open LM Studio, vLLM, llama.cpp, or Ollama on a non-default address Choose Custom OpenAI-compatible API in Sources: the README ("Ollama needs none") and the manual pages |
Ollama is a built-in provider and needs no key.
Connect it
/connectand choose Ollama. codeaf uses Ollama's usual local OpenAI-compatible address and asks for nothing.codeaf connect ollama. As the terminal docs put it, DeepSeek and MiniMax take a key through the same checked connection as/connect; Ollama takes none.The connection check asks the local runner for its model list first. When it answers, its models show up under the provider's heading in
/model.Which model you get
With only Ollama connected, only its installed models appear, and the opening model is an installed Ollama model (the first available chat model in pr…