Create app.py

app.py

@@ -0,0 +1,224 @@
+# app.py
+import os
+import json
+import http.client
+from io import BytesIO
+
+import gradio as gr
+from dotenv import load_dotenv
+from elevenlabs.client import ElevenLabs
+
+# ----------------------------
+# Config & clients
+# ----------------------------
+load_dotenv()  # supports local .env; on HF Spaces, set secrets in the UI
+
+ELEVENLABS_API_KEY = os.getenv("ELEVENLABS_API_KEY", "")
+API_KEY_302 = os.getenv("API_KEY_302", "")
+
+# ElevenLabs client (only if key is present)
+elevenlabs_client = None
+if ELEVENLABS_API_KEY:
+    elevenlabs_client = ElevenLabs(api_key=ELEVENLABS_API_KEY)
+
+
+# ----------------------------
+# Prompt templates (placeholders)
+# ----------------------------
+PROMPT_TEMPLATE_1 = """\
+You are a speech-language assistant. Given the ORIGINAL script and the TRANSCRIPT (imperfect ASR),
+list words/phrases likely to trigger stuttering (e.g., consonant clusters, long multisyllabic words).
+Output a short, structured summary and diagnosis for easy-to-stutter scenarios.
+
+ORIGINAL:
+{original_text}
+
+TRANSCRIPT:
+{transcribed_text}
+"""
+
+PROMPT_TEMPLATE_2 = """\
+You are a speech-language assistant. Rewrite the ORIGINAL script to reduce stuttering risk, while
+preserving meaning and tone. Prefer simpler synonyms, shorter clauses, easier onsets. Keep it concise.
+
+Diagnosis notes on easy-to-stutter scenarios:
+{notes}
+
+ORIGINAL:
+{original_text}
+
+Only return the revised script.
+"""
+
+PROMPT_TEMPLATE_3 = """\
+You are a speech-language assistant. Rewrite the ORIGINAL script to reduce stuttering risk, while
+preserving meaning and tone. Prefer simpler synonyms, shorter clauses, easier onsets. Keep it concise.
+
+Diagnosis notes on easy-to-stutter scenarios:
+{notes}
+
+ORIGINAL:
+{original_text}
+
+Only return the revised script.
+"""
+
+
+# ----------------------------
+# Helpers: STT & LLM calls
+# ----------------------------
+def transcribe_audio(upload_path: str | None, record_path: str | None) -> str:
+    """
+    Prioritize uploaded file if both provided.
+    Returns the transcribed text (or an error message).
+    """
+    # Choose input
+    audio_path = upload_path or record_path
+    if not audio_path:
+        return "No audio provided. Please upload or record audio."
+
+    if not ELEVENLABS_API_KEY:
+        return "ELEVENLABS_API_KEY not set. Please configure your environment."
+
+    # Read file as bytes -> BytesIO
+    try:
+        with open(audio_path, "rb") as f:
+            audio_data = BytesIO(f.read())
+    except Exception as e:
+        return f"Failed to read audio: {e}"
+
+    try:
+        transcription = elevenlabs_client.speech_to_text.convert(
+            file=audio_data,
+            model_id="scribe_v1",
+            tag_audio_events=True,
+            language_code="eng",
+            diarize=True,
+        )
+        # Minimal output: just return text
+        return transcription.text or ""
+    except Exception as e:
+        return f"Transcription error: {e}"
+
+
+def call_llm_302(prompt: str) -> str:
+    """
+    Minimal wrapper around 302.ai /v1/chat/completions.
+    Returns assistant text or an error string.
+    """
+    if not API_KEY_302:
+        return "API_KEY_302 not set. Please configure your environment."
+
+    try:
+        conn = http.client.HTTPSConnection("api.302.ai")
+        payload = json.dumps({
+            "model": "gpt-4o-mini",
+            "messages": [
+                {"role": "user", "content": prompt}
+            ]
+        })
+        headers = {
+            "Accept": "application/json",
+            "Authorization": f"Bearer {API_KEY_302}",
+            "Content-Type": "application/json"
+        }
+        conn.request("POST", "/v1/chat/completions", payload, headers)
+        res = conn.getresponse()
+        raw = res.read().decode("utf-8")
+        conn.close()
+
+        output = json.loads(raw)
+        # Defensive parsing
+        msg = output.get("choices", [{}])[0].get("message", {})
+        text = msg.get("content") or msg.get("text") or str(msg)
+        return text.strip()
+    except Exception as e:
+        return f"LLM API error: {e}"
+
+
+# ----------------------------
+# Button handlers
+# ----------------------------
+def on_click_transcribe(upload_path, record_path):
+    """
+    Button 1: Transcribe audio -> fill Textbox1 (transcribed text, non-editable).
+    """
+    text = transcribe_audio(upload_path, record_path)
+    return gr.update(value=text)
+
+
+def on_click_analyze(original_text, transcribed_text):
+    """
+    Button 2: Analyze easy-to-stutter words -> fill Textbox3 using PROMPT_TEMPLATE_1.
+    """
+    prompt = PROMPT_TEMPLATE_1.format(
+        original_text=original_text or "",
+        transcribed_text=transcribed_text or "",
+    )
+    analysis = call_llm_302(prompt)
+    return gr.update(value=analysis)
+
+
+def on_click_rewrite(mode, original_text, transcribed_text, summary):
+    """
+    Button 3: Rewrite script -> fill Textbox4 using PROMPT_TEMPLATE_2 or PROMPT_TEMPLATE_3.
+    """
+    if mode == "mode1":
+        prompt = PROMPT_TEMPLATE_2.format(
+            original_text=original_text or "",
+        )
+    else:
+        prompt = PROMPT_TEMPLATE_3.format(
+            original_text=original_text or "",
+        )
+    revised = call_llm_302(prompt)
+    return gr.update(value=revised)
+
+
+# ----------------------------
+# Gradio UI
+# ----------------------------
+with gr.Blocks(title="DeStammerer: Script Revise (Minimal)") as demo:
+    gr.Markdown("### DeStammerer (Minimal Demo)\n"
+                "A minimal 3-row UI for transcription, analysis, and revision.")
+
+    # Row 1: [audio upload, audio record, button1]
+    with gr.Row():
+        audio_upload = gr.Audio(label="Upload Audio", sources=["upload"], type="filepath")
+        audio_record = gr.Audio(label="Record Audio", sources=["microphone"], type="filepath")
+        btn_transcribe = gr.Button("1) Transcribe")
+
+    # Row 2: [textbox1 (ASR, readonly), textbox2 (original input), button2]
+    with gr.Row():
+        txt_transcribed = gr.Textbox(label="Transcribed Text (ASR)", interactive=False, lines=6, placeholder="ASR output appears here.")
+        txt_original = gr.Textbox(label="Original Script (input)", lines=6, placeholder="Paste your original script here.")
+        btn_analyze = gr.Button("2) Analyze (Find Easy-to-Stutter Words)")
+
+    # Row 3: [textbox3 (LLM summary), textbox4 (revised script), selector, button3]
+    with gr.Row():
+        txt_summary = gr.Textbox(label="LLM Summary: Easy-to-Stutter Words", lines=8, placeholder="Analysis will appear here.")
+        txt_revised = gr.Textbox(label="Revised Script", lines=8, placeholder="Rewritten script will appear here.")
+        mode_selector = gr.Dropdown(choices=["mode1", "mode2"], value="mode1", label="Rewrite Mode")
+        btn_rewrite = gr.Button("3) Revise Script")
+
+    # Wiring
+    btn_transcribe.click(
+        fn=on_click_transcribe,
+        inputs=[audio_upload, audio_record],
+        outputs=[txt_transcribed],
+    )
+
+    btn_analyze.click(
+        fn=on_click_analyze,
+        inputs=[txt_original, txt_transcribed],
+        outputs=[txt_summary],
+    )
+
+    btn_rewrite.click(
+        fn=on_click_rewrite,
+        inputs=[mode_selector, txt_original, txt_transcribed, txt_summary],
+        outputs=[txt_revised],
+    )
+
+if __name__ == "__main__":
+    demo.launch()