LM Studio

LM Studio exposes an OpenAI-compatible endpoint on localhost:1234 for developer machines. Useful for prototyping against a local model while still routing through TapPass for audit and policy.

Requirements

Server-side (TapPass must be able to reach the developer’s LM Studio port — typically only true when TapPass is also running locally):

LMSTUDIO_BASE_URL=http://localhost:1234/v1

Option A — SDK

from tappass import Agent

agent = Agent("http://localhost:9620", "tp_dev_...")
response = agent.chat("Hello", model="lmstudio/qwen2.5-7b-instruct")

Option B — OpenAI SDK, zero-code

export OPENAI_BASE_URL=http://localhost:9620/v1
export OPENAI_API_KEY=tp_dev_...

from openai import OpenAI

client = OpenAI()
response = client.chat.completions.create(
    model="lmstudio/qwen2.5-7b-instruct",
    messages=[{"role": "user", "content": "Hello"}],
)

When to use this

• Prototyping governance policies on your laptop before shipping a rule.
• Offline dev environments where you can’t hit a cloud provider.
• Cost control during demos — swap the model without touching agent code.

For production-scale self-hosted inference, use vLLM or Ollama instead.