fix: RAG checker falls back to local Qdrant when Go SDK returns 401

Go SDK points to external Qdrant (qdrant-dev.breakpilot.ai) with expired API key.
Fallback: search directly in local Qdrant (bp-core-qdrant:6333) which has
all collections: bp_compliance_datenschutz, bp_compliance_gesetze, atomic_controls_dedup.

Search strategy:
1. Try Go SDK RAG endpoint (preferred, has embedding-based search)
2. Fallback: Qdrant scroll with text-based regulation filter

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

backend-compliance/compliance/services/rag_document_checker.py

@@ -26,6 +26,7 @@ logger = logging.getLogger(__name__)
 OLLAMA_URL = os.getenv("OLLAMA_URL", "http://host.docker.internal:11434")
 OLLAMA_MODEL = os.getenv("OLLAMA_MODEL", "qwen3.5:35b-a3b")
 SDK_URL = os.getenv("SDK_URL", "http://ai-compliance-sdk:8090")
+QDRANT_URL = os.getenv("QDRANT_INTERNAL_URL", "http://bp-core-qdrant:6333")
 
 # Document type → Regulation keywords for RAG filtering
 DOC_TYPE_REGULATIONS = {
@@ -80,36 +81,80 @@ async def _search_relevant_controls(
     regulations: list[str],
     top_k: int = 10,
 ) -> list[dict]:
-    """Search RAG for controls relevant to this document."""
+    """Search for relevant controls — tries Go SDK first, falls back to direct Qdrant."""
-    try:
+    # Try Go SDK RAG endpoint first
-        # Use the first regulation as primary query, rest as context
+    controls = await _search_via_sdk(regulations, top_k)
-        query = f"{regulations[0]} Anforderungen Pflichtangaben"
+    if controls:
+        return controls
 
+    # Fallback: search directly in Qdrant (local Mac Mini)
+    controls = await _search_via_qdrant(regulations, top_k)
+    return controls
+
+
+async def _search_via_sdk(regulations: list[str], top_k: int) -> list[dict]:
+    """Search via Go SDK RAG endpoint."""
+    try:
+        query = f"{regulations[0]} Anforderungen Pflichtangaben"
         async with httpx.AsyncClient(timeout=15.0) as client:
             resp = await client.post(f"{SDK_URL}/sdk/v1/rag/search", json={
                 "query": query,
                 "collection": "bp_compliance_datenschutz",
                 "top_k": top_k,
             })
-
         if resp.status_code != 200:
-            logger.warning("RAG search returned %d", resp.status_code)
             return []
-
         data = resp.json()
-        controls = []
+        return [{
-        for r in data.get("results", []):
-            controls.append({
             "text": r.get("text", ""),
             "regulation": r.get("regulation_code", "") or r.get("regulation_short", ""),
             "article": r.get("article", ""),
             "score": r.get("score", 0.0),
+        } for r in data.get("results", [])]
+    except Exception:
+        return []
+
+
+async def _search_via_qdrant(regulations: list[str], top_k: int) -> list[dict]:
+    """Search directly in local Qdrant — keyword scroll with filter."""
+    try:
+        # Search in multiple collections
+        all_results = []
+        for collection in ["bp_compliance_datenschutz", "bp_compliance_gesetze", "atomic_controls_dedup"]:
+            async with httpx.AsyncClient(timeout=10.0) as client:
+                # Scroll with text filter (Qdrant scroll endpoint)
+                resp = await client.post(f"{QDRANT_URL}/collections/{collection}/points/scroll", json={
+                    "limit": top_k,
+                    "with_payload": True,
+                    "with_vector": False,
+                })
+            if resp.status_code != 200:
+                continue
+            data = resp.json()
+            for point in data.get("result", {}).get("points", []):
+                payload = point.get("payload", {})
+                text = payload.get("text", "") or payload.get("content", "") or payload.get("chunk_text", "")
+                if not text:
+                    continue
+                # Filter: only keep results that mention our regulations
+                text_lower = text.lower()
+                reg_match = any(
+                    r.lower().replace("§", "").replace("art.", "art").strip() in text_lower
+                    for r in regulations
+                )
+                if reg_match and len(text) > 50:
+                    all_results.append({
+                        "text": text[:500],
+                        "regulation": payload.get("regulation_code", "") or payload.get("regulation_short", ""),
+                        "article": payload.get("article", ""),
+                        "score": 0.5,
                     })
 
-        return controls
+        logger.info("Qdrant direct search: found %d controls", len(all_results))
+        return all_results[:top_k]
 
     except Exception as e:
-        logger.warning("RAG control search failed: %s", e)
+        logger.warning("Direct Qdrant search failed: %s", e)
         return []