M
Mosaic Eval

Model setup

DashboardRunsNew RunModelsTasks

This page reflects what is already loaded from `.env` and the running local services. Change the env file, restart, and the available models will update here.

Drizzle + SQLiteBun workflowProxy-task MVP
Back to runs
Auto-detected stack
One Bedrock key unlocks four demo agents, while direct provider keys stay available as separate options.
3 Bedrock-backed models are ready, and 2 direct models are ready.
LM Studio
Point `LM_STUDIO_BASE_URL` at a running local server and set `LM_STUDIO_MODEL_ID`.
The adapter uses an OpenAI-compatible chat completions endpoint, so local models work with the same run flow.
Readiness
5 ready, 2 need setup.
Bedrock OpenAIBedrock ClaudeBedrock Kimi K2.5Bedrock MiniMax M2.5Google gemini-2.5-proMistral mistral-large-latestLM Studio local-model
Model test bench
Test a model
This page reads configured environment variables. If Bedrock is enabled, the Bedrock models appear automatically.
Adapter status
Confirm what is configured before you queue a full run.
Bedrock OpenAIBedrock ClaudeBedrock Kimi K2.5Bedrock MiniMax M2.5Google gemini-2.5-proMistral mistral-large-latestLM Studio local-model
Bedrock OpenAI
Set OPENAI_API_KEY for direct OpenAI, or set BEDROCK_API_KEY plus OPENAI_BASE_URL=https://bedrock-mantle.<region>.api.aws/v1 to use Bedrock-hosted gpt-oss.
Current result

Run a test to inspect the live response payload.