Skip to content

HF TGI Gateway (ailab-app)

What It Is

A protocol-real Hugging Face TGI-facing gateway for the guided credential chain. Discovery endpoints remain reachable, while /generate requires the HuggingFace token found in MLflow run params/tags.

This models a common production shape: TGI itself stays simple, and auth is enforced by a fronting inference gateway.

Surface

Endpoint Purpose
GET /health Basic health check
GET /info TGI model metadata
GET /v1/models OpenAI-compatible model listing
GET /metrics Prometheus-style telemetry
POST /generate Auth-gated text generation

Port & Unit

Parameter Value
Host 172.16.50.40
Port 8180
Unit tgi-gateway.service
Runtime python3 server.py
Required header Authorization: Bearer <HF token from MLflow>

Guided Chain Role

  1. Ray leaks the MLflow Basic credential.
  2. MLflow exposes the HF/TGI token in run params/tags.
  3. The TGI gateway rejects missing or wrong tokens.
  4. The TGI gateway accepts the MLflow-looted HF token.