Update app.py

app.py

@@ -11,34 +11,26 @@ import gradio as gr
 from transformers import pipeline
 from transformers.pipelines.audio_utils import ffmpeg_read
 
-# --- Firebase config (try gradio secrets, otherwise environment) --
-# On HF Spaces you can set secrets; fallback to environment variables if needed.
-try:
-    FIREBASE_URL = gr.secrets.get("databaseURL")  # gradio secrets object; use .get for safety
-except Exception:
-    # older gradio versions expose a dict-like object; try attribute access
+# --- Firebase config (gradio secrets or environment) ---
+# On HF Spaces set secrets as: databaseURL and databaseSecret (or FIREBASE_API_KEY if using other flows)
+def _get_secret_safe(name):
     try:
-        FIREBASE_URL = gr.secrets("databaseURL")
+        return gr.secrets.get(name)
     except Exception:
-        FIREBASE_URL = None
+        try:
+            # older gradio: callable or dict-like
+            return gr.secrets(name)
+        except Exception:
+            return None
 
-FIREBASE_AUTH = None
-try:
-    FIREBASE_AUTH = gr.secrets.get("apiKey")
-except Exception:
-    try:
-        FIREBASE_AUTH = gr.secrets("apiKey")
-    except Exception:
-        FIREBASE_AUTH = None
+FIREBASE_URL = _get_secret_safe("databaseURL") or os.environ.get("FIREBASE_DATABASE_URL") or os.environ.get("FIREBASE_URL")
+FIREBASE_AUTH = _get_secret_safe("databaseSecret") or _get_secret_safe("apiKey") or os.environ.get("FIREBASE_API_KEY") or os.environ.get("FIREBASE_AUTH")
 
-# fallback to environment variables if secrets not provided
-if not FIREBASE_URL:
-    FIREBASE_URL = os.environ.get("FIREBASE_DATABASE_URL") or os.environ.get("FIREBASE_URL")
 if FIREBASE_URL:
     FIREBASE_URL = FIREBASE_URL.rstrip("/")
 
-if not FIREBASE_AUTH:
-    FIREBASE_AUTH = os.environ.get("FIREBASE_API_KEY") or os.environ.get("FIREBASE_AUTH")
+print("Firebase URL:", FIREBASE_URL)
+print("Firebase Auth present:", bool(FIREBASE_AUTH))
 
 # requests (required for Firebase REST calls)
 try:
@@ -66,12 +58,8 @@ SLEEP_BETWEEN_CHUNKS = 0.01
 
 
 def _merge_with_overlap(accumulated: str, chunk_text: str, max_overlap_words: int = 12) -> str:
-    """
-    Merge chunk_text into accumulated, removing duplicated overlap (word-based).
-    """
     if not accumulated:
         return chunk_text
-
     acc_words = accumulated.strip().split()
     chunk_words = chunk_text.strip().split()
     if not chunk_words:
@@ -93,27 +81,29 @@ def _merge_with_overlap(accumulated: str, chunk_text: str, max_overlap_words: in
 
 
 # ------------------------------
-# Firebase helpers (GET + PUT with merging) - using key "/transcribe"
+# Firebase helpers (GET + PUT with merging)
 # ------------------------------
 def _firebase_headers():
-    """Return headers for Firebase REST call (none required for Realtime DB default)."""
     return {"Content-Type": "application/json"}
 
 
 def _firebase_get(path: str) -> Optional[object]:
-    """
-    GET value from Firebase at given path (path without leading slash, e.g. 'transcribe').
-    Returns parsed JSON value or None if not found / error.
-    """
     if not FIREBASE_URL or requests is None:
         print("_firebase_get: Firebase not configured or requests missing.")
         return None
     try:
         url = f"{FIREBASE_URL}/{path}.json"
         params = {"auth": FIREBASE_AUTH} if FIREBASE_AUTH else None
-        resp = requests.get(url, params=params, timeout=6)
-        resp.raise_for_status()
-        return resp.json()
+        print(f"GET {url} params={params}")
+        resp = requests.get(url, params=params, timeout=8)
+        print("GET status:", resp.status_code)
+        try:
+            print("GET response:", resp.text)
+            resp.raise_for_status()
+            return resp.json()
+        except Exception as e:
+            print("GET parse/raise:", e)
+            return None
     except Exception as e:
         print("Firebase GET failed:", e)
         return None
@@ -121,29 +111,42 @@ def _firebase_get(path: str) -> Optional[object]:
 
 def _firebase_put(path: str, value: object) -> bool:
     """
-    PUT value to Firebase at given path (overwrites that key).
-    Returns True on success, False otherwise.
+    Try PUT; if auth is present but fails, print response. If it fails and auth was provided, try unauthenticated attempt
+    (useful to detect whether your DB requires auth or not).
     """
     if not FIREBASE_URL or requests is None:
         print("Firebase PUT skipped: config or requests missing.")
         return False
+
+    url = f"{FIREBASE_URL}/{path}.json"
+    params = {"auth": FIREBASE_AUTH} if FIREBASE_AUTH else None
+
     try:
-        url = f"{FIREBASE_URL}/{path}.json"
-        params = {"auth": FIREBASE_AUTH} if FIREBASE_AUTH else None
-        resp = requests.put(url, params=params, json=value, timeout=6)
-        resp.raise_for_status()
-        return True
+        print(f"Attempting PUT {url} params={params} payload_len={len(value) if isinstance(value, str) else 'obj'}")
+        resp = requests.put(url, params=params, json=value, timeout=10)
+        print("PUT status:", resp.status_code)
+        print("PUT response:", resp.text)
+        if resp.ok:
+            return True
+        else:
+            # If auth was used and failed, try unauthenticated (to differentiate rules vs invalid auth)
+            if FIREBASE_AUTH:
+                print("PUT with auth failed. Trying unauthenticated PUT to help diagnosis...")
+                try:
+                    resp2 = requests.put(url, json=value, timeout=10)
+                    print("Unauth PUT status:", resp2.status_code)
+                    print("Unauth PUT response:", resp2.text)
+                    return resp2.ok
+                except Exception as e2:
+                    print("Unauth PUT attempt failed:", e2)
+                    return False
+            return False
     except Exception as e:
-        print("Firebase PUT failed:", e)
+        print("Firebase PUT failed (exception):", e)
         return False
 
 
 def send_to_firebase_merge(new_transcript: str) -> bool:
-    """
-    Merge new_transcript with existing /transcribe value in Firebase (if present),
-    using the same overlap dedupe logic as the app, then write merged result back.
-    Returns True if write succeeded, False otherwise.
-    """
     if not FIREBASE_URL:
         print("send_to_firebase_merge: FIREBASE_URL not configured.")
         return False
@@ -152,32 +155,32 @@ def send_to_firebase_merge(new_transcript: str) -> bool:
         return False
 
     try:
-        # Read existing transcribe (if any) under the key "transcribe"
-        existing = _firebase_get("transcribe")  # could be None or a string or structured JSON
+        existing = _firebase_get("transcribe")
         existing_text = ""
         if isinstance(existing, str):
             existing_text = existing
-        elif isinstance(existing, dict) and "transcribe" in existing and isinstance(existing["transcribe"], str):
-            # handle previous structure that stored object with transcribe key
-            existing_text = existing["transcribe"]
+        elif isinstance(existing, dict) and "text" in existing and isinstance(existing["text"], str):
+            # if previously stored as object like {"text": "..."}
+            existing_text = existing["text"]
         elif existing is None:
             existing_text = ""
         else:
-            # if other types, coerce to string safely
             existing_text = str(existing)
 
-        # Merge using overlap dedupe
         merged = _merge_with_overlap(existing_text, new_transcript)
-
-        # Debug/logging
-        print(f"Existing length: {len(existing_text)}, New chunk length: {len(new_transcript)}, Merged length: {len(merged)}")
-
-        # Write merged plain string under /transcribe (overwrites)
-        success = _firebase_put("transcribe", merged)
+        # Save as object with timestamp to make debugging easier
+        payload = {
+            "text": merged,
+            "updated_at": datetime.utcnow().isoformat() + "Z",
+            "length": len(merged),
+        }
+        print("Prepared payload for /transcribe:", {k: ("<len>" if k=="text" else v) for k, v in payload.items()})
+
+        success = _firebase_put("transcribe", payload)
         if success:
             print("Successfully wrote merged transcribe to Firebase (/transcribe).")
         else:
-            print("Failed to write merged transcribe to Firebase.")
+            print("Failed to write merged transcribe to Firebase (/transcribe).")
         return success
     except Exception as e:
         print("send_to_firebase_merge: exception:", e)
@@ -185,25 +188,18 @@ def send_to_firebase_merge(new_transcript: str) -> bool:
 
 
 # ------------------------------
-# Transcription generator
+# Transcription generator (same as before)
 # ------------------------------
 def transcribe_file(filepath: str):
-    """
-    Generator that yields (transcript_text, continue_html_update).
-    After final chunk: attempt to merge+write to Firebase under /transcribe.
-    Show Continue only if write succeeds; otherwise show an error HTML message.
-    """
     if not filepath:
         raise gr.Error("No audio recorded! Please record audio (max 59s recommended).")
 
     try:
         sr = pipe.feature_extractor.sampling_rate
 
-        # Read audio bytes
         if isinstance(filepath, (bytes, bytearray)):
             audio_array = ffmpeg_read(filepath, sr)
         else:
-            # filepath is a local path provided by gradio when type="filepath"
             with open(filepath, "rb") as f:
                 audio_array = ffmpeg_read(f.read(), sr)
 
@@ -221,7 +217,6 @@ def transcribe_file(filepath: str):
         for i in range(num_chunks):
             start = i * chunk_samples
             end = min((i + 1) * chunk_samples, total_samples)
-
             if i > 0:
                 start = max(0, start - overlap_samples)
             if i < num_chunks - 1:
@@ -237,12 +232,10 @@ def transcribe_file(filepath: str):
 
             accumulated_text = _merge_with_overlap(accumulated_text, chunk_text)
 
-            # yield partial; keep Continue hidden
             yield accumulated_text, gr.update(visible=False)
-
             time.sleep(SLEEP_BETWEEN_CHUNKS)
 
-        # Final: attempt to write merged transcript into Firebase under /transcribe
+        # Final write
         write_ok = False
         try:
             write_ok = send_to_firebase_merge(accumulated_text)
@@ -250,7 +243,6 @@ def transcribe_file(filepath: str):
             print("Error while sending to Firebase:", e)
 
         if write_ok:
-            # Continue button only when Firebase write succeeded
             continue_html = """
             <div style="text-align:center; margin-top:20px;">
                 <a href="https://nolist-hospitalcentral.hf.space"
@@ -270,7 +262,6 @@ def transcribe_file(filepath: str):
             """
             yield accumulated_text, gr.update(value=continue_html, visible=True)
         else:
-            # Show error HTML and keep Continue hidden
             error_html = """
             <div style="text-align:center; margin-top:20px; color:#b91c1c;">
                 <strong>Failed to save transcribe to Firebase.</strong><br/>
@@ -298,8 +289,7 @@ with gr.Blocks(title="Record Option") as demo:
     output_text = gr.Textbox(label="Transcription", lines=12)
     continue_block = gr.HTML(visible=False)
 
-    # Auto-trigger transcribe when recording upload completes
     audio_input.change(fn=transcribe_file, inputs=[audio_input], outputs=[output_text, continue_block])
 
 if __name__ == "__main__":
-    demo.launch(server_name="0.0.0.0", server_port=7860)
+    demo.launch(server_name="0.0.0.0", server_port=7860)