Fix: Auto-build corpora on first launch

app_retrieval_cached.py

@@ -9,6 +9,8 @@ from pathlib import Path
 import pickle
 import hashlib
 import json
+import subprocess
+import sys
 from retriever.index_bm25 import BM25Index
 from retriever.index_dense import DenseIndex
 from retriever.ingest import load_jsonl
@@ -26,6 +28,36 @@ print(" With disk caching for fast startup!")
 print("=" * 70)
 
 
+def _ensure_corpora_exist():
+    """Build corpora files if they don't exist"""
+    data_dir = Path("data/corpora")
+    required_files = [
+        data_dir / "medical_qa.jsonl",
+        data_dir / "miriad_text.jsonl",
+        data_dir / "unidoc_qa.jsonl"
+    ]
+    
+    if all(f.exists() for f in required_files):
+        return  # All files exist
+    
+    print("\n" + "=" * 70)
+    print("⚠️  Corpora files not found. Building them now...")
+    print("    This will take 2-3 minutes on first launch.")
+    print("=" * 70 + "\n")
+    
+    try:
+        # Run build_corpora.py
+        subprocess.run(
+            [sys.executable, "adapters/build_corpora.py"],
+            check=True,
+            capture_output=False
+        )
+        print("\n✓ Corpora files built successfully!\n")
+    except subprocess.CalledProcessError as e:
+        print(f"\n✗ Failed to build corpora: {e}")
+        raise RuntimeError("Could not build corpora files. Please run 'python adapters/build_corpora.py' manually.")
+
+
 def _default_corpora_config() -> Dict[str, dict]:
     return {
         "medical_qa": {"path": "data/corpora/medical_qa.jsonl",
@@ -139,6 +171,10 @@ class CachedRetriever:
         return dense_index
 
 
+
+# Ensure corpora files exist (auto-build if missing)
+_ensure_corpora_exist()
+
 # Initialize cached retriever (fast if cached, slow first time)
 print("\nInitializing retrieval system...")
 cfg = _available(_default_corpora_config())