Upload app.py with huggingface_hub

app.py

@@ -14,8 +14,10 @@ except ImportError:
     HAS_SPACES = False
 
 import gc
+import json
 import os
 import time
+import traceback
 from pathlib import Path
 from typing import Optional, Dict, Any, List, Tuple
 
@@ -55,112 +57,147 @@ SENTENCE_PRESETS = {
 }
 
 
+def _to_json_safe(obj: Any) -> Any:
+    """Recursively convert numpy/torch structures to JSON-serializable primitives."""
+    if isinstance(obj, dict):
+        return {str(k): _to_json_safe(v) for k, v in obj.items()}
+    elif isinstance(obj, (list, tuple)):
+        return [_to_json_safe(x) for x in obj]
+    elif isinstance(obj, (np.floating, float)):
+        return float(obj)
+    elif isinstance(obj, (np.integer, int)):
+        return int(obj)
+    elif isinstance(obj, np.ndarray):
+        return _to_json_safe(obj.tolist())
+    else:
+        return obj
+
+
 def _diagnose_speech_core(
     audio_path: Optional[str],
     target_phrase: str,
     healthy_baseline_path: Optional[str] = None,
-) -> Tuple[str, str, str, str, str, str]:
+) -> Tuple[str, str, str, str, str, str, dict]:
     """Run full diagnostic pipeline and return formatted clinical report for Gradio."""
+    empty_dict = {"status": "waiting_for_input"}
+    
     if not audio_path:
         return (
-            "Please record speech or upload an audio file to evaluate.",
+            "**Notice**: Please record speech or upload an audio file to evaluate.",
+            "NO AUDIO",
+            "N/A",
             "N/A",
-            "0 / 100",
-            "0.0%",
             "",
-            "{}",
+            "",
+            empty_dict,
         )
 
-    if not target_phrase.strip():
+    if not target_phrase or not target_phrase.strip():
         return (
-            "Please enter or select an expected Target Phrase.",
+            "**Notice**: Please enter or select an expected Target Phrase.",
+            "NO TARGET",
+            "N/A",
             "N/A",
-            "0 / 100",
-            "0.0%",
             "",
-            "{}",
+            "",
+            empty_dict,
         )
 
-    engine = get_engine()
+    try:
+        engine = get_engine()
 
-    # Execute Diagnostic Engine
-    diag_res = engine.diagnose_audio(
-        audio_input=audio_path,
-        target_phrase=target_phrase,
-        normal_calibration_audio=healthy_baseline_path,
-    )
+        # Execute Diagnostic Engine
+        diag_res = engine.diagnose_audio(
+            audio_input=audio_path,
+            target_phrase=target_phrase,
+            normal_calibration_audio=healthy_baseline_path,
+        )
+
+        if diag_res.get("is_silent") or diag_res["decision"].get("is_silent"):
+            return (
+                "**No Speech Detected**: The audio is silent or below acoustic energy thresholds. Please speak clearly into your microphone.",
+                "SILENT",
+                "0 / 100",
+                "0.0%",
+                "",
+                "- **Silence Guard Active**: No vocal signal detected in audio stream.",
+                _to_json_safe(diag_res),
+            )
 
-    if diag_res["is_silent"] or diag_res["decision"].get("is_silent"):
+        result = diag_res["decision"]
+        pron = diag_res["pronunciation"]
+        flaws = diag_res["flaws"]
+        artic = diag_res["articulation"]
+        p_stut = float(diag_res["stutter_probs"][1]) if (diag_res.get("stutter_probs") and len(diag_res["stutter_probs"]) > 1) else 0.0
+
+        overall_bucket = result["buckets"]["overall"].upper()
+        fluency_score = f"{int(result.get('fluency_100', 100))} / 100"
+        pron_acc = f"{max(0.0, min(100.0, (1.0 - pron.get('wer', 0.0)) * 100.0)):.1f}%"
+
+        # Build Word Alignment Chips HTML
+        alignment = pron.get("alignment", [])
+        chips_html = "<div style='display:flex; flex-wrap:wrap; gap:8px; padding:12px; background:rgba(15,23,42,0.6); border-radius:8px; margin:10px 0;'>"
+        for item in alignment:
+            status = item["status"]
+            exp = item["expected"]
+            spk = item["spoken"]
+            if status == "correct":
+                chips_html += f"<span style='padding:6px 12px; background:rgba(16,185,129,0.15); color:#34D399; border:1px solid rgba(16,185,129,0.3); border-radius:6px; font-weight:600;'>[MATCH] {exp}</span>"
+            elif status == "substitution":
+                chips_html += f"<span style='padding:6px 12px; background:rgba(239,68,68,0.15); color:#F87171; border:1px solid rgba(239,68,68,0.3); border-radius:6px; font-weight:600;'>[DIFF] {exp} (heard: \"{spk}\")</span>"
+            elif status == "omission":
+                chips_html += f"<span style='padding:6px 12px; background:rgba(245,158,11,0.15); color:#FBBF24; border:1px solid rgba(245,158,11,0.3); border-radius:6px; font-weight:600;'>[UNSPOKEN] {exp}</span>"
+            elif status == "insertion":
+                chips_html += f"<span style='padding:6px 12px; background:rgba(168,85,247,0.15); color:#C084FC; border:1px solid rgba(168,85,247,0.3); border-radius:6px; font-weight:600;'>[EXTRA] {spk}</span>"
+        chips_html += "</div>"
+
+        # Build Flaw Report Summary Markdown
+        flaws_md = "### Specific Speech Pathology Findings:\n\n"
+        if flaws["has_r_flaw"]:
+            for r_err in flaws["r_sound_issues"]:
+                flaws_md += f"- **Rhotacism Flaw**: {r_err['message']}\n"
+        else:
+            flaws_md += "- **'R' Sound Articulation**: Accurate (No R->W/L substitution detected).\n"
+
+        if flaws["has_s_flaw"]:
+            for s_err in flaws["s_sound_issues"]:
+                flaws_md += f"- **Sigmatism Flaw**: {s_err['message']}\n"
+        else:
+            flaws_md += "- **'S' Sound Articulation**: Accurate (No sibilant lisp detected).\n"
+
+        if p_stut >= 0.78:
+            flaws_md += f"- **Disfluency Detected**: Elevated probability of repetition/block ({p_stut*100:.1f}%)\n"
+        elif p_stut >= 0.60:
+            flaws_md += f"- **Mild Hesitation**: Minor syllable repetition observed ({p_stut*100:.1f}%)\n"
+        else:
+            flaws_md += "- **Fluency Flow**: Continuous cadence (No disfluent events detected).\n"
+
+        flaws_md += f"- **Voice Phonation Correlates**: Pitch F0={artic.get('pitch_f0_mean_hz',0):.1f}Hz, HNR={artic.get('hnr_db',0):.1f}dB, Jitter={artic.get('jitter',0)*100:.2f}%\n"
+
+        confidence_val = result.get("confidence", "high")
+        heard_summary = f"**Decoded Transcription**: *\"{pron.get('asr_hypothesis','')}\"*\n\n**Confidence Rating**: `{confidence_val}` | **Latency**: `{diag_res['latency_ms']} ms`"
+
+        return (
+            heard_summary,
+            overall_bucket,
+            fluency_score,
+            pron_acc,
+            chips_html,
+            flaws_md,
+            _to_json_safe(diag_res),
+        )
+    except Exception as ex:
+        err_msg = f"**Execution Error**: {ex}\n\n```\n{traceback.format_exc()}\n```"
         return (
-            "No Speech Detected: The audio is silent or below acoustic energy thresholds. Please speak clearly.",
-            "SILENT",
+            err_msg,
+            "ERROR",
             "0 / 100",
             "0.0%",
             "",
-            str(diag_res),
+            f"- **Internal Error**: {ex}",
+            {"error": str(ex), "traceback": traceback.format_exc()},
         )
 
-    result = diag_res["decision"]
-    pron = diag_res["pronunciation"]
-    flaws = diag_res["flaws"]
-    artic = diag_res["articulation"]
-    p_stut = float(diag_res["stutter_probs"][1]) if (diag_res["stutter_probs"] and len(diag_res["stutter_probs"]) > 1) else 0.0
-
-    overall_bucket = result["buckets"]["overall"].upper()
-    fluency_score = f"{int(result.get('fluency_100', 100))} / 100"
-    pron_acc = f"{max(0.0, min(100.0, (1.0 - pron.get('wer', 0.0)) * 100.0)):.1f}%"
-
-    # Build Word Alignment Chips HTML
-    alignment = pron.get("alignment", [])
-    chips_html = "<div style='display:flex; flex-wrap:wrap; gap:8px; padding:12px; background:rgba(15,23,42,0.6); border-radius:8px; margin:10px 0;'>"
-    for item in alignment:
-        status = item["status"]
-        exp = item["expected"]
-        spk = item["spoken"]
-        if status == "correct":
-            chips_html += f"<span style='padding:6px 12px; background:rgba(16,185,129,0.15); color:#34D399; border:1px solid rgba(16,185,129,0.3); border-radius:6px; font-weight:600;'>[MATCH] {exp}</span>"
-        elif status == "substitution":
-            chips_html += f"<span style='padding:6px 12px; background:rgba(239,68,68,0.15); color:#F87171; border:1px solid rgba(239,68,68,0.3); border-radius:6px; font-weight:600;'>[DIFF] {exp} (heard: \"{spk}\")</span>"
-        elif status == "omission":
-            chips_html += f"<span style='padding:6px 12px; background:rgba(245,158,11,0.15); color:#FBBF24; border:1px solid rgba(245,158,11,0.3); border-radius:6px; font-weight:600;'>[UNSPOKEN] {exp}</span>"
-        elif status == "insertion":
-            chips_html += f"<span style='padding:6px 12px; background:rgba(168,85,247,0.15); color:#C084FC; border:1px solid rgba(168,85,247,0.3); border-radius:6px; font-weight:600;'>[EXTRA] {spk}</span>"
-    chips_html += "</div>"
-
-    # Build Flaw Report Summary Markdown
-    flaws_md = "### Specific Speech Pathology Findings:\n\n"
-    if flaws["has_r_flaw"]:
-        for r_err in flaws["r_sound_issues"]:
-            flaws_md += f"- **Rhotacism Flaw**: {r_err['message']}\n"
-    else:
-        flaws_md += "- **'R' Sound Articulation**: Accurate (No R->W/L substitution detected).\n"
-
-    if flaws["has_s_flaw"]:
-        for s_err in flaws["s_sound_issues"]:
-            flaws_md += f"- **Sigmatism Flaw**: {s_err['message']}\n"
-    else:
-        flaws_md += "- **'S' Sound Articulation**: Accurate (No sibilant lisp detected).\n"
-
-    if p_stut >= 0.78:
-        flaws_md += f"- **Disfluency Detected**: Elevated probability of repetition/block ({p_stut*100:.1f}%)\n"
-    elif p_stut >= 0.60:
-        flaws_md += f"- **Mild Hesitation**: Minor syllable repetition observed ({p_stut*100:.1f}%)\n"
-    else:
-        flaws_md += "- **Fluency Flow**: Continuous cadence (No disfluent events detected).\n"
-
-    flaws_md += f"- **Voice Quality**: Pitch F0={artic.get('pitch_f0_mean_hz',0):.1f}Hz, HNR={artic.get('hnr_db',0):.1f}dB, Jitter={artic.get('jitter',0)*100:.2f}%\n"
-
-    heard_summary = f"**Decoded Transcription**: *\"{pron.get('asr_hypothesis','')}\"*\n\n**Inference Latency**: `{diag_res['latency_ms']} ms`"
-
-    return (
-        heard_summary,
-        overall_bucket,
-        fluency_score,
-        pron_acc,
-        chips_html + "\n\n" + flaws_md,
-        str(diag_res),
-    )
-
 
 # Apply ZeroGPU acceleration decorator if running on Hugging Face ZeroGPU
 if HAS_SPACES:
@@ -175,8 +212,8 @@ else:
 # Construct Gradio Modern Interface
 with gr.Blocks(title="Anvaya | Speech Pathology Diagnostics") as demo:
     gr.Markdown("""
-    # ANVAYA · Clinical Speech Pathology & Articulation Diagnostics
-    ### Multi-Modal Diagnostics: Neural Disfluency · Rhotacism ('r') · Sigmatism ('s' Lisp) · Praat Vocal Phonation
+    # ANVAYA · Speech Disfluency & Articulation Screening
+    ### Multi-Modal Screening: Neural Disfluency · Rhotacism ('r') · Sigmatism ('s' Lisp) · Phonation Acoustics
     """)
 
     with gr.Row():
@@ -243,6 +280,7 @@ with gr.Blocks(title="Anvaya | Speech Pathology Diagnostics") as demo:
 
             summary_box = gr.Markdown("### Clinical Assessment Summary\n*Results will appear here after analysis.*")
             alignment_html = gr.HTML(label="Word-Level Alignment")
+            flaws_box = gr.Markdown("### Specific Speech Pathology Findings\n*Sound checks will appear here.*")
 
             with gr.Accordion("Auditable Telemetry & Acoustic Evidence Trace", open=False):
                 raw_json = gr.JSON()
@@ -250,9 +288,8 @@ with gr.Blocks(title="Anvaya | Speech Pathology Diagnostics") as demo:
     diagnose_btn.click(
         fn=diagnose_speech_hf,
         inputs=[audio_input, target_text, healthy_baseline],
-        outputs=[summary_box, kpi_strat, kpi_fluency, kpi_acc, alignment_html, raw_json],
+        outputs=[summary_box, kpi_strat, kpi_fluency, kpi_acc, alignment_html, flaws_box, raw_json],
     )
 
 if __name__ == "__main__":
     demo.launch()
-