local

Overview ¶

VaultRun Local Gateway — OpenAI-compatible action plane for local inference.

Proxies /v1/chat/completions to Ollama / LM Studio / vLLM and executes VaultRun sandbox tools whenever the model returns tool_calls.

Usage:

LOCAL_GATEWAY_AUTH_TOKEN=... \
VAULTRUN_BASE_URL=http://localhost:8080 \
VAULTRUN_API_KEY=vr_... \
LOCAL_GATEWAY_UPSTREAM_URL=http://127.0.0.1:11434 \
./vaultrun-local

See docs/local-gateway.md.