Repository navigation
[FEATURE]: mutiple servers of the same local provider (llama.cpp, vllm, etc) #51675
Description
Feature hasn't been suggested before.
- I have verified this feature I'm about to request hasn't been suggested before.
Describe the enhancement you want to request
I have mutiple llama.cpp servers configured: one is local on my laptop and another on AI workstation in my home network. So when I try to use config like this:
{
"$schema": "https://opencode.ai/config.json",
"mcp": {
"dash-api": {
"type": "local",
"command": ["uvx", "--from", "git+https://github.com/Kapeli/dash-mcp-server.git", "dash-mcp-server"],
"enabled": true
}
},
"plugin": ["superpowers@git+https://github.com/obra/superpowers.git", "@mohak34/[email protected]"],
"provider": {
"llama.cpp": {
"npm": "@ai-sdk/openai-compatible",
"name": "llama-server (local)",
"options": {
"baseURL": "http://127.0.0.1:8080/v1"
},
"models": {
"qwen3-6-27b": {
"name": "Qwen3.6: 27b (local)",
"limit": {
"context": 128000,
"output": 64000
}
},
"qwen3-8-27b-medium": {
"name": "Qwen3.8: 27b (medium) (local)",
"limit": {
"context": 128000,
"output": 64000
}
},
"qwen3-8-27b-xhigh": {
"name": "Qwen3.8: 27b (xhigh) (local)",
"limit": {
"context": 128000,
"output": 64000
}
},
"qwen3-6-27b-mtp": {
"name": "Qwen3.6: 27b MTP (local)",
"limit": {
"context": 128000,
"output": 64000
}
},
"qwen3-6-35b-a3b-mtp-thinking": {
"name": "Qwen3.6: 35B-A3B MTP (local)",
"limit": {
"context": 128000,
"output": 64000
}
},
"qwen3-6-27b-nr": {
"name": "Qwen3.6: 27b no-reasoning (local)",
"limit": {
"context": 128000,
"output": 64000
}
},
"qwen3-6-35b-a3b": {
"name": "Qwen3.6: a3b-35b (local)",
"limit": {
"context": 128000,
"output": 64000
}
},
"qwen3-6-35b-a3b-thinking": {
"name": "Qwen3.6: a3b-35b thinking (local)",
"limit": {
"context": 128000,
"output": 64000
}
},
"kat-coder-2.5-dev": {
"name": "KAT Coder 2.5 Dev (local)",
"limit": {
"context": 128000,
"output": 64000
}
},
"gemma-4-26b-a4b": {
"name": "Gemma4: 26b-a4b (local)",
"limit": {
"context": 128000,
"output": 65536
}
},
"gemma-diffusion": {
"name": "Gemma: Diffusion (local)",
"limit": {
"context": 128000,
"output": 65536
}
}
}
},
"llama.cpp": {
"npm": "@ai-sdk/openai-compatible",
"name": "llama-server (ai-ws)",
"options": {
"baseURL": "http://xxfatumxx-ai-ws.local:8080/v1"
},
"models": {
"qwen3-8-27b-6-k-m": {
"name": "Qwen3.8: 27b (6-bit) (ai-ws)",
"limit": {
"context": 128000,
"output": 64000
}
},
"qwen3-8-27b-4-k-xl": {
"name": "Qwen3.8: 27b (4-bit) (ai-ws)",
"limit": {
"context": 262144,
"output": 128000
}
},
"qwen3-6-35b-a3b": {
"name": "Qwen3.6: 35B-A3B (ai-ws)",
"limit": {
"context": 262144,
"output": 128000
}
},
"gemma-4-31b": {
"name": "Gemma4: 31b (ai-ws)",
"limit": {
"context": 262144,
"output": 128000
}
},
"gpt-oss-20b": {
"name": "GPT-OSS: 20b (ai-ws)",
"limit": {
"context": 128000,
"output": 64000
}
}
}
}
}
}It loads only ONE of them (last configured), so now I just commenting out the one I don't use to make other work, which is inconvenient.
Is there any way to make opencode configuration support multiple servers of the same local provider? I haven't tested it on vllm or sglang but I think they will have the same problem.
Reacted by programmin1
Metadata
Metadata
Assignees
Labels
No labels
Activity
github-actions commentedon Sep 27, 2026
This issue might be a duplicate of existing issues. Please check:
xxfatumxx commentedon Sep 27, 2026
Found workaround by using custom provider logic so if you rename
llama.cppinto let's sayllama-ai-ws/modelscommand now sees models from both servers. If there is no resource/intention to add extra functionality I guess the issue can be closed.programmin1 commentedon Oct 6, 2026
There really should be easy way to set up a local AI - no internet or account required!
neriousy commentedon Oct 20, 2026
Thanks for following up. You confirmed that giving each server a distinct provider ID makes models from both servers available, which is supported by the custom-provider configuration. I'm closing this as resolved.