Litellm dev 12 28 2024 p3 (#7464)

* feat(deepgram/): initial e2e support for deepgram stt

Uses deepgram's `/listen` endpoint to transcribe speech to text

 Closes https://github.com/BerriAI/litellm/issues/4875

* fix: fix linting errors

* test: fix test

litellm/main.py

@@ -4867,7 +4867,30 @@ def transcription(
             api_base=api_base,
             api_key=api_key,
         )
-
+    elif custom_llm_provider == "deepgram":
+        response = base_llm_http_handler.audio_transcriptions(
+            model=model,
+            audio_file=file,
+            optional_params=optional_params,
+            model_response=model_response,
+            atranscription=atranscription,
+            client=(
+                client
+                if client is not None
+                and (
+                    isinstance(client, HTTPHandler)
+                    or isinstance(client, AsyncHTTPHandler)
+                )
+                else None
+            ),
+            timeout=timeout,
+            max_retries=max_retries,
+            logging_obj=litellm_logging_obj,
+            api_base=api_base,
+            api_key=api_key,
+            custom_llm_provider="deepgram",
+            headers={},
+        )
     if response is None:
         raise ValueError("Unmapped provider passed in. Unable to get the response.")
     return response