refactored real time audio pipeline

stt_test.py

@@ -1,58 +1,93 @@
-# -*- coding: utf-8 -*-
-"""STT_test
+# =========================================
+#  TabeebAI — FINAL STABLE VERSION without RAG
+# Urdu Speech → English Clinical Report
+# =========================================
 
-Automatically generated by Colab.
-
-Original file is located at
-    https://colab.research.google.com/drive/1i4LcOMK0nHo5PE0tZKM-yCfxyGmv_sJa
-"""
-
-# =========================
-#  TabeebAI FULL PIPELINE (ONE CELL)
-# Speech → Whisper → Symptoms → JSON → Gradio + Batch
-# =========================
-
-!pip -q install groq pydub gradio
+!pip -q install groq gradio soundfile numpy
 
 import json
+import tempfile
+import soundfile as sf
 from groq import Groq
-from pydub import AudioSegment
-from google.colab import files, userdata
+from google.colab import userdata
 import gradio as gr
 
-# =========================
-#  GROQ SETUP
-# =========================
+# =========================================
+#  SETUP
+# =========================================
 client = Groq(api_key=userdata.get("GROQ_API_KEY"))
-print("✔ Groq Ready")
-
-# =========================
-#  AUDIO CONVERT (M4A → WAV)
-# =========================
-def convert_audio(file_path):
-    audio = AudioSegment.from_file(file_path, format="m4a")
-    wav_path = file_path.replace(".m4a", ".wav")
-    audio.export(wav_path, format="wav")
-    return wav_path
-
-# =========================
-#  WHISPER TRANSCRIPTION
-# =========================
-def transcribe_audio(file_path):
-    with open(file_path, "rb") as f:
-        res = client.audio.transcriptions.create(
-            file=(file_path, f.read()),
-            model="whisper-large-v3-turbo"
+
+print(" TabeebAI FINAL Version Ready")
+
+# =========================================
+#  TRANSCRIPTION (FORCED URDU)
+# =========================================
+def transcribe_audio(audio):
+
+    if audio is None:
+        return "⚠️ No audio detected"
+
+    sr, data = audio
+
+    try:
+        with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as temp:
+            sf.write(temp.name, data, sr)
+
+            with open(temp.name, "rb") as f:
+                response = client.audio.transcriptions.create(
+                    file=f,
+                    model="whisper-large-v3-turbo",
+                    language="ur"   #  FORCE URDU
+                )
+
+        return response.text.strip()
+
+    except Exception as e:
+        return f" Transcription Error: {str(e)}"
+
+
+# =========================================
+#  TRANSLATION (URDU → ENGLISH)
+# =========================================
+def translate_to_english(text):
+
+    if not text or "Error" in text or "No audio" in text:
+        return text
+
+    try:
+        response = client.chat.completions.create(
+            model="llama-3.3-70b-versatile",
+            messages=[{
+                "role": "user",
+                "content": f"Translate the following Urdu medical text into clear English:\n{text}"
+            }],
+            temperature=0.1
         )
-    return res.text
 
-# =========================
-#  SYMPTOM EXTRACTION (SAFE JSON)
-# =========================
+        return response.choices[0].message.content.strip()
+
+    except Exception as e:
+        return f"Translation Error: {str(e)}"
+
+
+# =========================================
+#  SYMPTOM EXTRACTION (ENGLISH ONLY)
+# =========================================
 def extract_symptoms(text):
 
+    if not text or "Error" in text or "No audio" in text:
+        return {"error": text}
+
     prompt = f"""
-Extract medical symptoms and return ONLY valid JSON.
+You are a clinical AI assistant.
+
+STRICT RULES:
+- Input is translated into English
+- Output MUST be ONLY in English
+- Return ONLY valid JSON
+- No explanations, no extra text
+
+Extract structured medical symptoms.
 
 TEXT:
 {text}
@@ -61,95 +96,82 @@ FORMAT:
 {{
   "chief_complaint": "",
   "symptoms": [
-    {{"name": "", "severity": "", "duration": ""}}
+    {{
+      "name": "",
+      "severity": "",
+      "duration": ""
+    }}
   ],
+  "possible_conditions": [],
   "urgency": "low/medium/high",
-  "language": ""
+  "language": "English"
 }}
 """
 
-    res = client.chat.completions.create(
-        model="llama-3.3-70b-versatile",
-        messages=[{"role": "user", "content": prompt}],
-        temperature=0.2
-    )
-
-    output = res.choices[0].message.content
-    output = output.replace("```json", "").replace("```", "").strip()
-
     try:
-        return json.loads(output)
-    except:
-        return {"raw_output": output}
-
-# =========================
-#  FULL PIPELINE (CORE ENGINE)
-# =========================
-def run_pipeline(file_name):
-
-    wav = convert_audio(file_name)
-    text = transcribe_audio(wav)
-    analysis = extract_symptoms(text)
-
-    return {
-        "file": file_name,
-        "transcript": text,
-        "analysis": analysis
-    }
+        response = client.chat.completions.create(
+            model="llama-3.3-70b-versatile",
+            messages=[{"role": "user", "content": prompt}],
+            temperature=0.2
+        )
 
-# =========================
-#  BATCH PROCESS
-# =========================
-def batch_process(file_list):
+        output = response.choices[0].message.content
+        output = output.replace("```json", "").replace("```", "").strip()
 
-    results = []
+        return json.loads(output)
 
-    for f in file_list:
-        try:
-            results.append(run_pipeline(f))
-        except Exception as e:
-            results.append({"file": f, "error": str(e)})
+    except Exception as e:
+        return {
+            "error": "Parsing failed",
+            "details": str(e),
+            "raw_output": output if 'output' in locals() else ""
+        }
 
-    return results
 
-# =========================
-# 🌐 GRADIO INTERFACE
-# =========================
-def gradio_run(file):
+# =========================================
+#  FULL PIPELINE
+# =========================================
+def pipeline(audio):
 
-    result = run_pipeline(file.name)
+    transcript_urdu = transcribe_audio(audio)
+    english_text = translate_to_english(transcript_urdu)
+    analysis = extract_symptoms(english_text)
 
     return (
-        result["transcript"],
-        json.dumps(result["analysis"], indent=2, ensure_ascii=False)
+        transcript_urdu,
+        english_text,
+        json.dumps(analysis, indent=2, ensure_ascii=False)
     )
 
-with gr.Blocks() as demo:
-
-    gr.Markdown("# 🏥 TabeebAI - Speech → Symptoms AI")
 
-    file_input = gr.File(label="Upload M4A File")
+# =========================================
+#  UI (AUTO TRIGGER — NO BUTTON BUG)
+# =========================================
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
 
-    btn = gr.Button("Run Analysis 🚀")
+    gr.Markdown("""
+    # 🏥 TabeebAI — Clinical Voice AI
+    🎤 Speak in Urdu → AI gives English Medical Report
+    """)
 
-    transcript_box = gr.Textbox(label="Transcript")
-    json_box = gr.Textbox(
-    label="Symptoms JSON",
-    lines=15,
-    max_lines=30,
-    show_copy_button=True
-)
-
-    btn.click(gradio_run, file_input, [transcript_box, json_box])
-
-demo.launch(share=True)
+    audio_input = gr.Audio(
+        sources=["microphone"],
+        type="numpy",
+        label="🎤 Speak in Urdu"
+    )
 
-# =========================
-#  OPTIONAL BATCH TEST (RUN AFTER UPLOAD)
-# =========================
-uploaded = files.upload()
-results = batch_process(uploaded.keys())
+    transcript_box = gr.Textbox(label="📝 Urdu Transcript", lines=5)
+    english_box = gr.Textbox(label="🌐 English Translation", lines=5)
+    json_box = gr.Textbox(label="🩺 Clinical JSON Output", lines=16)
 
-print("\n✔ FINAL OUTPUT:\n")
-print(json.dumps(results, indent=2, ensure_ascii=False))
+    #  AUTO PROCESS AFTER RECORDING
+    audio_input.stop_recording(
+        fn=pipeline,
+        inputs=audio_input,
+        outputs=[transcript_box, english_box, json_box]
+    )
 
+# =========================================
+#  LAUNCH
+# =========================================
+demo.launch(debug=True, share=True)