Deploy LiteLLM proxy

README.md

@@ -12,13 +12,15 @@ This Space runs the LiteLLM proxy, providing an OpenAI-compatible API that can r
 ## Secrets to Set (Space Settings -> Secrets)
 
 - `LITELLM_MASTER_KEY`: master key required in `Authorization: Bearer ...`
-- `OPENAI_API_KEY`: for OpenAI routing
+- `GMN_API_KEY`: for the third-party OpenAI-compatible gateway
+- `GMN_API_BASE`: set to `https://gmn.chuangzuoli.com/v1`
 - `ANTHROPIC_API_KEY`: for Anthropic routing
 
 ## Useful Endpoints
 
 - `GET /health/liveliness` (no auth) - good for keep-alive pings
 - `POST /chat/completions` (auth) - OpenAI-compatible chat completions
+- `POST /v1/responses` (auth) - OpenAI-compatible Responses API (recommended)
 
 ## Example Curl
 
@@ -36,6 +38,19 @@ curl -sS https://YOUR-SPACE.hf.space/chat/completions \
   }'
 ```
 
+Responses API (some gateways require this exact path + OpenAI-Beta header):
+
+```bash
+curl -sS https://YOUR-SPACE.hf.space/v1/responses \
+  -H "Authorization: Bearer $LITELLM_MASTER_KEY" \
+  -H "Content-Type: application/json" \
+  -H "OpenAI-Beta: responses=v1" \
+  -d '{
+    "model": "gpt-4o-mini",
+    "input": "Say hi"
+  }'
+```
+
 Anthropic via alias:
 
 ```bash

config.yaml

@@ -3,7 +3,13 @@ model_list:
   - model_name: gpt-4o-mini
     litellm_params:
       model: openai/gpt-4o-mini
-      api_key: "os.environ/OPENAI_API_KEY"
+      # Third-party OpenAI-compatible endpoint
+      api_base: "os.environ/GMN_API_BASE"
+      api_key: "os.environ/GMN_API_KEY"
+      # Some gateways/WAFs require specific headers to allow /v1/responses
+      extra_headers:
+        OpenAI-Beta: "responses=v1"
+        User-Agent: "curl/8.0"
 
   # Pick a modern, fast Claude Sonnet variant as the default alias.
   # You can still call any Anthropic model via `anthropic/<model>` because of the wildcard route.
@@ -16,7 +22,11 @@ model_list:
   - model_name: openai/*
     litellm_params:
       model: openai/*
-      api_key: "os.environ/OPENAI_API_KEY"
+      api_base: "os.environ/GMN_API_BASE"
+      api_key: "os.environ/GMN_API_KEY"
+      extra_headers:
+        OpenAI-Beta: "responses=v1"
+        User-Agent: "curl/8.0"
     model_info:
       health_check_model: openai/gpt-4o-mini