Houtini LM
by shotaromehta
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| HOUTINI_LM_MODEL | No | Pin the model to use, because on a big catalogue every unknown model scores the same in routing and the first listed wins. Required in practice for OpenAI, which lists dozens of models that all score the same (e.g. gpt-5.2). | |
| HOUTINI_LM_API_KEY | No | API key for the endpoint, if it needs one (e.g. sk-... for OpenAI). Not required for a local LM Studio instance. | |
| HOUTINI_LM_THINKING | No | Controls thinking for models that support the toggle. 'auto' (default) switches thinking off for detected thinking models, 'off' forces thinking off on every call (needed when a backend serves a thinking model under a name detection can't recognise), and 'on' forces thinking on, which is worth it for bug-hunting or checking an argument, at the cost of time and tokens. | |
| HOUTINI_LM_SERIALISE | No | Set to 0 to turn off request serialisation. Calls are queued one at a time by default, because a single GPU can only serve one request anyway; disable this for OpenAI, a router or a batching backend like vLLM (e.g. HOUTINI_LM_SERIALISE=0). | |
| HOUTINI_LM_MIN_TOKENS | No | Minimum output budget. MCP clients habitually pass tiny caps like 256 that strangle reasoning models mid-thought, so anything under 4,096 is ignored unless you set HOUTINI_LM_MIN_TOKENS=0 to disable the floor. | |
| HOUTINI_LM_STRUCTURED | No | Set to 1 and each inference result also carries a structuredContent block (the answer, model, tokens, timing, quality flags) to read as JSON. Leave it off for Claude Code, which shows the model only that block when it's present. | |
| HOUTINI_LM_ENDPOINT_URL | No | URL of the OpenAI-compatible endpoint to delegate to (LM Studio, Ollama, vLLM, SGLang, LiteLLM router, OpenRouter, OpenAI, DeepSeek, Groq, etc.). LM Studio listens on localhost:1234 by default, which is where houtini-lm looks first, so this is only needed when the endpoint is elsewhere (e.g. http://192.168.1.50:1234 or https://api.openai.com). |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Server capabilities have not been inspected yet.
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
No tools | |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
This server cannot be deployed
Maintenance
ActivityActive
ResponsivenessNo issues