refactor(backend): extract process_reel_url from extract_url handler

Prereq for the Supabase queue worker — same pipeline, callable in-process.

main.py

@@ -600,17 +600,13 @@ def list_clusters():
         logger.error(f"Failed to list clusters: {str(e)}")
         raise HTTPException(status_code=500, detail=f"Failed to list clusters: {str(e)}")
 
-@app.post("/extract/url", response_model=ExtractionResponse)
-async def extract_url(payload: dict):
-    """
-    Accepts Instagram Reel URL. Downloads video, extracts audio, transcribes audio,
-    combines description with transcript, extracts JSON highlights, and saves to database.
-    """
-    url = payload.get("url")
-    if not url:
-        raise HTTPException(status_code=400, detail="Missing required 'url' parameter.")
+def process_reel_url(url: str) -> dict:
+    """Full single-reel pipeline. Returns the saved DB record.
 
-    # Check if we already processed this URL to save API tokens
+    Returns the cached row if this URL was already processed. Raises
+    HTTPException(400) on silent-hook reels (no spoken content). Always
+    cleans up the temp audio file. Used by /extract/url and the queue worker."""
+    # Cache: skip inference if we already have this URL
     try:
         conn = sqlite3.connect(DB_PATH)
         cursor = conn.cursor()
@@ -626,41 +622,29 @@ async def extract_url(payload: dict):
                 "raw_transcript": row[3],
                 "post_caption": row[4],
                 "extracted_json": json.loads(row[5]),
-                "created_at": row[6]
+                "created_at": row[6],
             }
     except Exception as e:
         logger.warning(f"Failed to check existing SQLite URL cache: {str(e)}")
 
     mp3_path = None
     try:
-        # 1. Download video and extract audio MP3 & caption description
         mp3_path, post_caption, title = download_and_extract_audio(url)
-
-        # 2. Transcribe Audio
         raw_transcript = ""
         try:
             raw_transcript = transcribe_audio(mp3_path)
         except Exception as e:
             logger.warning(f"Transcription failed: {str(e)}. Proceeding using metadata/caption only.")
-
-        # 3. Check for empty transcription/silent hook
         guard_silent_hook(raw_transcript, post_caption)
-
-        # 4. Extract structured insights
         extracted_data = extract_structured_json(raw_transcript, post_caption)
-
-        # 5. Commit record to database
-        db_record = save_to_database(
+        return save_to_database(
             url=url,
             title=title,
             raw_transcript=raw_transcript,
             post_caption=post_caption,
-            extracted=extracted_data
+            extracted=extracted_data,
         )
-
-        return db_record
     finally:
-        # Cleanup temporary audio files
         if mp3_path and os.path.exists(mp3_path):
             try:
                 os.remove(mp3_path)
@@ -668,6 +652,15 @@ async def extract_url(payload: dict):
             except Exception as ce:
                 logger.warning(f"Could not delete temp file {mp3_path}: {str(ce)}")
 
+
+@app.post("/extract/url", response_model=ExtractionResponse)
+async def extract_url(payload: dict):
+    """Accepts an Instagram Reel URL and runs the full extraction pipeline."""
+    url = payload.get("url")
+    if not url:
+        raise HTTPException(status_code=400, detail="Missing required 'url' parameter.")
+    return process_reel_url(url)
+
 @app.post("/extract/file", response_model=ExtractionResponse)
 async def extract_file(
     file: UploadFile = File(...),

verify_pipeline.py

@@ -214,6 +214,36 @@ def test_cluster_merge_pass():
         main.CLUSTER_CHUNK_DELAY = orig_delay
 
 
+def test_extract_url_regression():
+    print("Testing POST /extract/url via process_reel_url (mocked)...")
+    import main
+    long_transcript = " ".join(["word"] * 30)  # clears the 15-word silent-hook guard
+    mocked = ReelExtraction(
+        core_topic="Topic", key_takeaway="Takeaway",
+        action_items=["a"], tools_or_resources=["b"])
+    orig_dl = main.download_and_extract_audio
+    orig_tr = main.transcribe_audio
+    orig_ex = main.extract_structured_json
+    orig_save = main.save_to_database
+    main.download_and_extract_audio = lambda url: ("/tmp/does_not_exist.mp3", "cap", "Title")
+    main.transcribe_audio = lambda p: long_transcript
+    main.extract_structured_json = lambda t, c: mocked
+    main.save_to_database = lambda url, title, raw_transcript, post_caption, extracted: {
+        "id": "id-1", "url": url, "title": title,
+        "raw_transcript": raw_transcript, "post_caption": post_caption,
+        "extracted_json": extracted.model_dump(), "created_at": "2026-06-06T00:00:00Z"}
+    try:
+        r = client.post("/extract/url", json={"url": "https://www.instagram.com/reel/REGRESSION1/"})
+        assert r.status_code == 200, r.text
+        assert r.json()["extracted_json"]["core_topic"] == "Topic"
+        print("[OK] extract_url regression passed!")
+    finally:
+        main.download_and_extract_audio = orig_dl
+        main.transcribe_audio = orig_tr
+        main.extract_structured_json = orig_ex
+        main.save_to_database = orig_save
+
+
 def test_delete_reel():
     print("Testing DELETE /reels/{id}...")
     import main, sqlite3, uuid, json
@@ -268,6 +298,7 @@ if __name__ == "__main__":
     test_recompute_clusters_mock()
     test_cluster_chunking()
     test_cluster_merge_pass()
+    test_extract_url_regression()
     test_delete_reel()
     test_list_clusters()
     test_reels_include_cluster()