WIP

scripts/batch_inference_v3.py

@@ -0,0 +1,92 @@
+#!/usr/bin/env python3
+"""Batch inference v3 - 30 random PDFs for Round 2 validation."""
+
+import json
+import os
+import random
+import time
+
+import requests
+
+PDF_DIR = "/mnt/c/Users/yaoji/git/Billo/Billo.Platform.Document/Billo.Platform.Document.AdminAPI/downloads/to_check"
+API_URL = "http://localhost:8000/api/v1/infer"
+OUTPUT_FILE = "/mnt/c/Users/yaoji/git/ColaCoder/invoice-master-poc-v2/scripts/inference_results_v3.json"
+SAMPLE_SIZE = 30
+
+
+def main():
+    random.seed(99_2026)  # New seed for Round 3
+
+    all_pdfs = [f for f in os.listdir(PDF_DIR) if f.lower().endswith(".pdf")]
+    selected = random.sample(all_pdfs, min(SAMPLE_SIZE, len(all_pdfs)))
+
+    print(f"Selected {len(selected)} random PDFs for inference")
+
+    results = []
+    for i, filename in enumerate(selected, 1):
+        filepath = os.path.join(PDF_DIR, filename)
+        filesize = os.path.getsize(filepath)
+        print(f"[{i}/{len(selected)}] Processing {filename}...", end=" ", flush=True)
+
+        start = time.time()
+        try:
+            with open(filepath, "rb") as f:
+                resp = requests.post(
+                    API_URL,
+                    files={"file": (filename, f, "application/pdf")},
+                    timeout=120,
+                )
+            elapsed = round(time.time() - start, 2)
+
+            if resp.status_code == 200:
+                data = resp.json()
+                field_count = sum(
+                    1 for k, v in data.items()
+                    if k not in (
+                        "DocumentId", "confidence", "success", "fallback_used",
+                        "bboxes", "cross_validation", "processing_time_ms",
+                        "line_items", "vat_summary", "vat_validation",
+                        "raw_detections", "detection_classes", "detection_count",
+                    )
+                    and v is not None
+                )
+                det_count = data.get("detection_count", "?")
+                print(f"OK ({elapsed}s) - {field_count} fields, {det_count} detections")
+                results.append({
+                    "filename": filename,
+                    "status": resp.status_code,
+                    "time_seconds": elapsed,
+                    "filesize": filesize,
+                    "data": data,
+                })
+            else:
+                print(f"HTTP {resp.status_code} ({elapsed}s)")
+                results.append({
+                    "filename": filename,
+                    "status": resp.status_code,
+                    "time_seconds": elapsed,
+                    "filesize": filesize,
+                    "error": resp.text[:200],
+                })
+        except Exception as e:
+            elapsed = round(time.time() - start, 2)
+            print(f"FAILED ({elapsed}s) - {e}")
+            results.append({
+                "filename": filename,
+                "status": -1,
+                "time_seconds": elapsed,
+                "filesize": filesize,
+                "error": str(e),
+            })
+
+    with open(OUTPUT_FILE, "w", encoding="utf-8") as f:
+        json.dump(results, f, ensure_ascii=False, indent=2)
+    print(f"\nResults saved to {OUTPUT_FILE}")
+
+    success = sum(1 for r in results if r["status"] == 200)
+    failed = len(results) - success
+    print(f"Total: {len(results)}, Success: {success}, Failed: {failed}")
+
+
+if __name__ == "__main__":
+    main()