Forward gr.Request to pipeline functions in session API endpoints

Enables usage logging with proper user ID hashing from HF headers
for all session-based API calls.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

src/api/session_api.py

@@ -13,6 +13,7 @@ import shutil
 import time
 import uuid
 
+import gradio as gr
 import numpy as np
 
 from config import SESSION_DIR, SESSION_EXPIRY_SECONDS
@@ -167,13 +168,14 @@ def _format_response(audio_id, json_output):
 # ---------------------------------------------------------------------------
 
 def process_audio_session(audio_data, min_silence_ms, min_speech_ms, pad_ms,
-                          model_name="Base", device="GPU"):
+                          model_name="Base", device="GPU",
+                          request: gr.Request = None):
     """Full pipeline: preprocess -> VAD -> ASR -> alignment. Creates session."""
     from src.pipeline import process_audio
 
     result = process_audio(
         audio_data, int(min_silence_ms), int(min_speech_ms), int(pad_ms),
-        model_name, device,
+        model_name, device, request=request,
     )
     # result is a 9-tuple:
     # (html, json_output, speech_intervals, is_complete, audio, sr, intervals, seg_dir, log_row)
@@ -193,7 +195,8 @@ def process_audio_session(audio_data, min_silence_ms, min_speech_ms, pad_ms,
 
 
 def resegment_session(audio_id, min_silence_ms, min_speech_ms, pad_ms,
-                       model_name="Base", device="GPU"):
+                       model_name="Base", device="GPU",
+                       request: gr.Request = None):
     """Re-clean VAD boundaries with new params and re-run ASR + alignment."""
     session = load_session(audio_id)
     if session is None:
@@ -205,7 +208,7 @@ def resegment_session(audio_id, min_silence_ms, min_speech_ms, pad_ms,
         session["speech_intervals"], session["is_complete"],
         session["audio"], 16000,
         int(min_silence_ms), int(min_speech_ms), int(pad_ms),
-        model_name, device,
+        model_name, device, request=request,
     )
     json_output = result[1]
     if json_output is None:
@@ -216,7 +219,8 @@ def resegment_session(audio_id, min_silence_ms, min_speech_ms, pad_ms,
     return _format_response(audio_id, json_output)
 
 
-def retranscribe_session(audio_id, model_name="Base", device="GPU"):
+def retranscribe_session(audio_id, model_name="Base", device="GPU",
+                          request: gr.Request = None):
     """Re-run ASR with a different model on current segment boundaries."""
     session = load_session(audio_id)
     if session is None:
@@ -237,7 +241,7 @@ def retranscribe_session(audio_id, model_name="Base", device="GPU"):
         session["intervals"],
         session["audio"], 16000,
         session["speech_intervals"], session["is_complete"],
-        model_name, device,
+        model_name, device, request=request,
     )
     json_output = result[1]
     if json_output is None:
@@ -247,7 +251,8 @@ def retranscribe_session(audio_id, model_name="Base", device="GPU"):
     return _format_response(audio_id, json_output)
 
 
-def realign_from_timestamps(audio_id, timestamps, model_name="Base", device="GPU"):
+def realign_from_timestamps(audio_id, timestamps, model_name="Base", device="GPU",
+                             request: gr.Request = None):
     """Run ASR + alignment on caller-provided timestamp intervals."""
     session = load_session(audio_id)
     if session is None:
@@ -265,7 +270,7 @@ def realign_from_timestamps(audio_id, timestamps, model_name="Base", device="GPU
         intervals,
         session["audio"], 16000,
         session["speech_intervals"], session["is_complete"],
-        model_name, device,
+        model_name, device, request=request,
     )
     json_output = result[1]
     if json_output is None: