Playground
Loading connection settings…
LLM
Model, backend, and chat history for this tenant.
Reload
Default language
BCP-47 locale injected into prompts (e.g. en-US).
LLM model
deepseek-r1:7b
deepseek-r1:latest
granite4.1:8b
qwen2.5:3b
qwen3.5:4b
qwen3.5:9b
qwen3:4b
Model id for this tenant. Must exist on the configured backend.
Backend
ollama (OpenAI /v1 on OllamaEndpoint)
vllm
lmstudio
openai
openai-compatible
custom (alias of openai-compatible)
ollama-native (/api/chat fallback)
LLM endpoint URL (optional)
LLM API key (optional)
Max history messages
Streaming enabled
Enable model thinking
Off by default. When on, thinking-capable models (e.g. Qwen3) may spend many tokens before the answer. Requires
reasoning_effort
on OpenAI
/v1
or
think
on Ollama native.
Save LLM