"""
Preflight dashboard: serves audio files and transcription results for human evaluation.
Run: python preflight/dashboard.py
Then open http://localhost:8765
"""
import json
import os
import sys
from http.server import HTTPServer, SimpleHTTPRequestHandler
from pathlib import Path
from urllib.parse import unquote, parse_qs, urlparse
PREFLIGHT_DIR = Path(__file__).parent
RESULTS_DIR = PREFLIGHT_DIR / "results"
SEGMENTS_DIR = PREFLIGHT_DIR / "pF_BQpHaIdU" / "segments"
PORT = 8765
def load_results():
phases = {}
for name in ["phase1_single", "phase2_concurrent", "phase3_full"]:
p = RESULTS_DIR / f"{name}.json"
if p.exists():
phases[name] = json.loads(p.read_text())
combined = RESULTS_DIR / "combined_summary.json"
summary = json.loads(combined.read_text()) if combined.exists() else {}
return phases, summary
def build_dashboard_html():
phases, summary = load_results()
# Build segments table from the largest phase available
best_phase = None
for name in ["phase3_full", "phase2_concurrent", "phase1_single"]:
if name in phases and phases[name].get("results"):
best_phase = phases[name]
break
results = best_phase.get("results", []) if best_phase else []
phase_summary = best_phase.get("summary", {}) if best_phase else {}
# Compute aggregates
total = len(results)
success = sum(1 for r in results if r.get("status") == "success")
errors = sum(1 for r in results if r.get("status") != "success")
avg_quality = 0
quality_scores = [r["validation"]["quality_score"] for r in results if "validation" in r]
if quality_scores:
avg_quality = sum(quality_scores) / len(quality_scores)
tts_clean = sum(1 for r in results if r.get("validation", {}).get("tts_clean_eligible"))
asr_ok = sum(1 for r in results if r.get("validation", {}).get("asr_eligible"))
no_speech = sum(1 for r in results if r.get("validation", {}).get("is_no_speech"))
cache_hits = sum(1 for r in results if r.get("token_usage", {}).get("cache_hit"))
total_tokens = sum(r.get("token_usage", {}).get("input", 0) + r.get("token_usage", {}).get("output", 0) for r in results)
latencies = [r["latency_ms"] for r in results if r.get("latency_ms", 0) > 0]
avg_latency = sum(latencies) / len(latencies) if latencies else 0
# Quality distribution
q_buckets = {"0.0-0.3": 0, "0.3-0.5": 0, "0.5-0.7": 0, "0.7-0.9": 0, "0.9-1.0": 0}
for q in quality_scores:
if q < 0.3: q_buckets["0.0-0.3"] += 1
elif q < 0.5: q_buckets["0.3-0.5"] += 1
elif q < 0.7: q_buckets["0.5-0.7"] += 1
elif q < 0.9: q_buckets["0.7-0.9"] += 1
else: q_buckets["0.9-1.0"] += 1
# Phase comparison
phase_rows = ""
for pname in ["phase1_single", "phase2_concurrent", "phase3_full"]:
if pname in phases:
s = phases[pname].get("summary", {})
phase_rows += f"""
| {pname.replace('_', ' ').title()} |
{s.get('requests_sent', 0)} |
{s.get('responses_success', 0)} |
{s.get('responses_error', 0)} |
{s.get('responses_429', 0)} |
{s.get('avg_quality_score', 0):.3f} |
{s.get('avg_latency_ms', 0):.0f}ms |
{s.get('cache_hits', 0)} |
{s.get('total_time_s', 0):.1f}s |
"""
# Segments table rows
seg_rows = ""
for i, r in enumerate(results):
if r.get("status") != "success":
seg_rows += f"""
| {i+1} | {r['segment_id']} | {r.get('audio_duration_s',0):.1f}s |
ERROR: {r.get('error_message','Unknown')} |
- | - |
"""
continue
v = r.get("validation", {})
speaker = r.get("speaker", {})
quality = v.get("quality_score", 0)
q_class = "q-high" if quality >= 0.9 else "q-mid" if quality >= 0.7 else "q-low"
lang = r.get("detected_language", "?")
lang_class = "lang-match" if lang == "te" else "lang-mismatch"
seg_rows += f"""
| {i+1} |
{r['segment_id']} |
{r.get('audio_duration_s',0):.1f}s |
{quality:.3f} |
{lang} |
{speaker.get('emotion','')} |
{speaker.get('pace','')} |
{'Y' if v.get('tts_clean_eligible') else 'N'} |
{r.get('latency_ms',0):.0f}ms |
{v.get('num_event_tags',0)} |
"""
# Results as JSON for JS
results_json = json.dumps(results, ensure_ascii=False)
return f"""
Preflight Dashboard - pF_BQpHaIdU
{success}/{total}
Success Rate
{avg_quality:.3f}
Avg Quality
{avg_latency:.0f}ms
Avg Latency
{total_tokens:,}
Total Tokens
Phase Comparison
| Phase | Sent | OK | Err | 429 | Quality | Latency | Cache | Time |
|---|
{phase_rows}
Quality Distribution
{"".join(f'
' for k,v in q_buckets.items())}
"""
class DashboardHandler(SimpleHTTPRequestHandler):
def do_GET(self):
parsed = urlparse(self.path)
path = unquote(parsed.path)
if path == "/" or path == "/index.html":
html = build_dashboard_html()
self.send_response(200)
self.send_header("Content-Type", "text/html; charset=utf-8")
self.end_headers()
self.wfile.write(html.encode())
return
if path.startswith("/audio/"):
filename = path[7:]
filepath = SEGMENTS_DIR / filename
if filepath.exists():
self.send_response(200)
self.send_header("Content-Type", "audio/flac")
self.send_header("Content-Length", str(filepath.stat().st_size))
self.end_headers()
self.wfile.write(filepath.read_bytes())
return
self.send_error(404)
return
if path == "/api/results":
phases, summary = load_results()
self.send_response(200)
self.send_header("Content-Type", "application/json")
self.end_headers()
self.wfile.write(json.dumps({"phases": {k: v.get("summary") for k, v in phases.items()}, "summary": summary}, ensure_ascii=False).encode())
return
self.send_error(404)
def log_message(self, format, *args):
pass
def main():
print(f"Starting preflight dashboard on http://localhost:{PORT}")
print(f"Results dir: {RESULTS_DIR}")
print(f"Segments dir: {SEGMENTS_DIR}")
server = HTTPServer(("0.0.0.0", PORT), DashboardHandler)
try:
server.serve_forever()
except KeyboardInterrupt:
print("\nShutdown.")
if __name__ == "__main__":
main()