feat: transparent anthropic api incl. native anthropic api backend

api/openai.py

@@ -665,10 +665,16 @@ async def openai_models_proxy(request: Request):
         fetch.endpoint_details(ep, "/models", "data", config.api_keys.get(ep), skip_error_cache=True, timeout=8)
         for ep in all_llama_endpoints
     ]
+    # 4. Query native Anthropic endpoints via /v1/models (auth headers picked by endpoint type)
+    anthropic_tasks = [
+        fetch.endpoint_details(ep, "/v1/models", "data", config.api_keys.get(ep), skip_error_cache=True, timeout=8)
+        for ep in config.anthropic_endpoints
+    ]
 
     ollama_models = await asyncio.gather(*ollama_tasks) if ollama_tasks else []
     ext_openai_models = await asyncio.gather(*ext_openai_tasks) if ext_openai_tasks else []
     llama_models = await asyncio.gather(*llama_tasks) if llama_tasks else []
+    anthropic_models = await asyncio.gather(*anthropic_tasks) if anthropic_tasks else []
 
     models = {'data': []}
 
@@ -702,6 +708,16 @@ async def openai_models_proxy(request: Request):
                     model['name'] = model['id']
                 models['data'].append(model)
 
+    # Add native Anthropic models (if any)
+    if anthropic_models:
+        for modellist in anthropic_models:
+            for model in modellist:
+                if not "id" in model.keys():
+                    model['id'] = model.get('name', model.get('id', ''))
+                else:
+                    model['name'] = model['id']
+                models['data'].append(model)
+
     # 2. Return a JSONResponse with a deduplicated list of unique models for inference
     return JSONResponse(
         content={"data": dedupe_on_keys(models['data'], ['name'])},