Question
Hello, I'm trying to get opencode to work with my local Ollama model. I'd like to set up two modes that can be toggled with keyboard shortcuts: thinking mode (with reasoning enabled) and non-thinking mode (reasoning disabled). Below is a partial configuration I've attempted:
"qwen3.5:9b-q4_K_M": {
"_launch": true,
"name": "qwen3.5:9b-q4_K_M",
"attachment": true,
"reasoning": true,
"modalities": {
"input": ["text", "image"],
"output": ["text"]
},
"variants": {
"thinking": {
"reasoning": true,
},
"fast": {
"reasoning": false,
},
},
}In my configuration, there are some "magic parameters" like attachment, which I found through online searches. I couldn't find them in authoritative documentation, such as Ollama's official docs or opencode's official docs. However, they run without any errors, which is confusing. I request that the documentation be improved to clarify these parameters.
Question
Hello, I'm trying to get opencode to work with my local Ollama model. I'd like to set up two modes that can be toggled with keyboard shortcuts: thinking mode (with reasoning enabled) and non-thinking mode (reasoning disabled). Below is a partial configuration I've attempted:
In my configuration, there are some "magic parameters" like
attachment, which I found through online searches. I couldn't find them in authoritative documentation, such as Ollama's official docs or opencode's official docs. However, they run without any errors, which is confusing. I request that the documentation be improved to clarify these parameters.