feat: Handle HF Remote API Call Format (#751)

**Reason for Change**:
Remote HuggingFace API uses a particular format: 

```
curl 'https://api-inference.huggingface.co/models/HuggingFaceH4/zephyr-7b-beta/v1/chat/completions' \
-H 'Authorization: Bearer hf_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx' \
-H 'Content-Type: application/json' \
--data '{
    "model": "HuggingFaceH4/zephyr-7b-beta",
    "messages": [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	],
    "max_tokens": 500,
    "stream": false
}'
```

This needs to be adhered to in the code. 

Reference:
https://huggingface.co/HuggingFaceH4/zephyr-7b-beta?inference_api=true

Adding this code ensures we continue to support OAI, HF Remote and
Custom URL as our LLM backend

presets/ragengine/inference/inference.py

@@ -26,6 +26,8 @@ def complete(self, prompt: str, **kwargs) -> CompletionResponse:
         try:
             if "openai" in INFERENCE_URL:
                 return self._openai_complete(prompt, **kwargs, **self.params)
+            elif "api-inference.huggingface" in INFERENCE_URL:
+                return self._huggingface_remote_complete(prompt, **kwargs, **self.params)
             else:
                 return self._custom_api_complete(prompt, **kwargs, **self.params)
         finally:
@@ -39,6 +41,13 @@ def _openai_complete(self, prompt: str, **kwargs: Any) -> CompletionResponse:
         )
         return llm.complete(prompt)
 
+    def _huggingface_remote_complete(self, prompt: str, **kwargs: Any) -> CompletionResponse:
+        headers = {"Authorization": f"Bearer {INFERENCE_ACCESS_SECRET}"}
+        data = {"messages": [{"role": "user", "content": prompt}]}
+        response = requests.post(INFERENCE_URL, json=data, headers=headers)
+        response_data = response.json()
+        return CompletionResponse(text=str(response_data))
+
     def _custom_api_complete(self, prompt: str, **kwargs: Any) -> CompletionResponse:
         headers = {"Authorization": f"Bearer {INFERENCE_ACCESS_SECRET}"}
         data = {"prompt": prompt, **kwargs}