Feat Request: LM Studio Easy Connect

[Franzferdinan51]

i'm posting this requesting someone who has better skills than my self to make an easy way to connect too LM Studio i mean the usefulness of it would be nice and its just outside my reach right now thank you all for anyone who reads this i appreciate you i'm trying to get this installed on an android phone as i feel its perfect for that

[OnlyXianzo]

"Hey! I just recently contributed a fix to the provider logic, and you can actually connect to LM Studio right now without any extra code.

LM Studio uses an OpenAI-compatible API. You just need to:

1. Start the Local Server in LM Studio.
2. In your PicoClaw config.json, use the openai section.
3. Set the api_base to your PC's IP address (e.g., http://192.168.1.5:1234/v1).
4. Set any dummy string for the api_key.

Because of the recent fix to prioritize prefixes, it should route correctly as long as you use the right base URL!"

[rlucasfm]

Trying to use it like you said but running "picoclaw agent" just returns "Error creating provider: no API key configured for model: qwen3-8b"

Heres the config file:
{ "agents": { "defaults": { "workspace": "~/.picoclaw/workspace", "model": "qwen3-8b", "max_tokens": 8192, "temperature": 0.7, "max_tool_iterations": 20 } }, "channels": { "whatsapp": { "enabled": false, "bridge_url": "ws://localhost:3001", "allow_from": [] }, "telegram": { "enabled": false, "token": "", "allow_from": [] }, "feishu": { "enabled": false, "app_id": "", "app_secret": "", "encrypt_key": "", "verification_token": "", "allow_from": [] }, "discord": { "enabled": false, "token": "", "allow_from": [] }, "maixcam": { "enabled": false, "host": "0.0.0.0", "port": 18790, "allow_from": [] } }, "providers": { "anthropic": { "api_key": "", "api_base": "" }, "openai": { "api_key": "apikey_for_lmstudio", "api_base": "http://192.168.3.11:1234/v1" }, "openrouter": { "api_key": "", "api_base": "" }, "groq": { "api_key": "", "api_base": "" }, "zhipu": { "api_key": "", "api_base": "" }, "vllm": { "api_key": "", "api_base": "" }, "gemini": { "api_key": "", "api_base": "" } }, "gateway": { "host": "0.0.0.0", "port": 18790 }, "tools": { "web": { "search": { "api_key": "", "max_results": 5 } } } }

"Hey! I just recently contributed a fix to the provider logic, and you can actually connect to LM Studio right now without any extra code.

LM Studio uses an OpenAI-compatible API. You just need to:

1. Start the Local Server in LM Studio.
2. In your PicoClaw config.json, use the openai section.
3. Set the api_base to your PC's IP address (e.g., http://192.168.1.5:1234/v1).
4. Set any dummy string for the api_key.

Because of the recent fix to prioritize prefixes, it should route correctly as long as you use the right base URL!"

[OnlyXianzo]

@rlucasfm

Hey! Good to see you trying out LM Studio. The error is happening because qwen3-8b isn't a recognized keyword for the OpenAI provider in the current code.

Since my recent fix added support for explicit prefixes, you can fix this instantly by updating your config.json:

Change:
"model": "qwen3-8b"

To:
"model": "openai/qwen3-8b"

This will force it to use the openai config block you set up for LM Studio. Give that a shot!

[binnohq]

Still not working. Anyone had it working that way?

[themethod]

It seems the provider key needs to be "openai"... e.g. you can't just make up your own, "my-local" provider.

[rlucasfm]

@rlucasfm

Hey! Good to see you trying out LM Studio. The error is happening because qwen3-8b isn't a recognized keyword for the OpenAI provider in the current code.

Since my recent fix added support for explicit prefixes, you can fix this instantly by updating your config.json:

Change: "model": "qwen3-8b"

To: "model": "openai/qwen3-8b"

This will force it to use the openai config block you set up for LM Studio. Give that a shot!

Using it as openai/qwen3-8b worked!

[bl4cks1d3]

Error creating provider: model "openai/gpt-oss-20b" not found in model_list: model "openai/gpt-oss-20b" not found in model_list or providers

{
"agents": {
"defaults": {
"workspace": "~/.picoclaw/workspace",
"model": "openai/gpt-oss-20b",
"max_tokens": 8192,
"temperature": 0.5,
"max_tool_iterations": 20
}
},
"providers": {
"openai": {
"aou_key": "dummy",
"api_base": "http://192.168.0.38:1234/v1"
}

{
"data": [
{
"id": "openai/gpt-oss-20b",
"object": "model",
"owned_by": "organization_owner"
},
{
"id": "text-embedding-nomic-embed-text-v1.5",
"object": "model",
"owned_by": "organization_owner"
}
],
"object": "list"

[nikolasdehor]

@bl4cks1d3 -- your error model "openai/gpt-oss-20b" not found in model_list is from the newer model_list system introduced in PR #492. Also note: your config has a typo: "aou_key" should be "api_key".

For LM Studio users, the current recommended setup after the provider refactors (#48, #213, #492):

"agents": {
  "defaults": {
    "provider": "openai",
    "model": "openai/your-model-name"
  }
},
"providers": {
  "openai": {
    "api_key": "lm-studio",
    "api_base": "http://192.168.x.x:1234/v1"
  }
}

If using the model_list system, the model must be registered there. Alternatively, setting the explicit provider field bypasses model_list lookup for direct provider routing.