From 9812ff46f32de52132de98c0f432d270c18cbdde Mon Sep 17 00:00:00 2001
From: Benjamin Admin <benjaminadmin@MacBookPro.fritz.box>
Date: Fri, 13 Mar 2026 18:42:40 +0100
Subject: [PATCH] feat: add 7-stage control generator pipeline with 3 license
 rules
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- control_generator.py: RAG→License→Structure/Reform→Harmonize→Anchor→Store→Mark pipeline
  with Anthropic Claude API (primary) + Ollama fallback for LLM reformulation
- anchor_finder.py: RAG-based + DuckDuckGo anchor search for open references
- control_generator_routes.py: REST API for generate, job status, review queue, processed stats
- 046_control_generator.sql: job tracking, chunk tracking, blocked sources tables;
  extends canonical_controls with license_rule, source_original_text, source_citation

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---
 .../compliance/services/control_generator.py  | 63 ++++++++++++++++---
 1 file changed, 55 insertions(+), 8 deletions(-)

diff --git a/backend-compliance/compliance/services/control_generator.py b/backend-compliance/compliance/services/control_generator.py
index be9b34e..9fffc0a 100644
--- a/backend-compliance/compliance/services/control_generator.py
+++ b/backend-compliance/compliance/services/control_generator.py
@@ -42,9 +42,11 @@ logger = logging.getLogger(__name__)
 # ---------------------------------------------------------------------------
 
 SDK_URL = os.getenv("SDK_URL", "http://ai-compliance-sdk:8090")
-LLM_CHAT_URL = f"{SDK_URL}/sdk/v1/llm/chat"
 EMBEDDING_URL = os.getenv("EMBEDDING_URL", "http://embedding-service:8087")
-LLM_MODEL = os.getenv("CONTROL_GEN_LLM_MODEL", "qwen3:30b-a3b")
+ANTHROPIC_API_KEY = os.getenv("ANTHROPIC_API_KEY", "")
+ANTHROPIC_MODEL = os.getenv("CONTROL_GEN_ANTHROPIC_MODEL", "claude-sonnet-4-6")
+OLLAMA_URL = os.getenv("OLLAMA_URL", "http://host.docker.internal:11434")
+OLLAMA_MODEL = os.getenv("CONTROL_GEN_OLLAMA_MODEL", "qwen3:30b-a3b")
 LLM_TIMEOUT = float(os.getenv("CONTROL_GEN_LLM_TIMEOUT", "120"))
 
 HARMONIZATION_THRESHOLD = 0.85  # Cosine similarity above this = duplicate
@@ -218,32 +220,77 @@ class GeneratorResult:
 # ---------------------------------------------------------------------------
 
 async def _llm_chat(prompt: str, system_prompt: Optional[str] = None) -> str:
-    """Call the Go SDK LLM chat endpoint."""
+    """Call LLM — Anthropic Claude (primary) or Ollama (fallback)."""
+    if ANTHROPIC_API_KEY:
+        result = await _llm_anthropic(prompt, system_prompt)
+        if result:
+            return result
+        logger.warning("Anthropic failed, falling back to Ollama")
+
+    return await _llm_ollama(prompt, system_prompt)
+
+
+async def _llm_anthropic(prompt: str, system_prompt: Optional[str] = None) -> str:
+    """Call Anthropic Messages API."""
+    headers = {
+        "x-api-key": ANTHROPIC_API_KEY,
+        "anthropic-version": "2023-06-01",
+        "content-type": "application/json",
+    }
+    payload = {
+        "model": ANTHROPIC_MODEL,
+        "max_tokens": 4096,
+        "messages": [{"role": "user", "content": prompt}],
+    }
+    if system_prompt:
+        payload["system"] = system_prompt
+
+    try:
+        async with httpx.AsyncClient(timeout=LLM_TIMEOUT) as client:
+            resp = await client.post(
+                "https://api.anthropic.com/v1/messages",
+                headers=headers,
+                json=payload,
+            )
+            if resp.status_code != 200:
+                logger.error("Anthropic API %d: %s", resp.status_code, resp.text[:300])
+                return ""
+            data = resp.json()
+            content = data.get("content", [])
+            if content and isinstance(content, list):
+                return content[0].get("text", "")
+            return ""
+    except Exception as e:
+        logger.error("Anthropic request failed: %s", e)
+        return ""
+
+
+async def _llm_ollama(prompt: str, system_prompt: Optional[str] = None) -> str:
+    """Call Ollama chat API (fallback)."""
     messages = []
     if system_prompt:
         messages.append({"role": "system", "content": system_prompt})
     messages.append({"role": "user", "content": prompt})
 
     payload = {
-        "model": LLM_MODEL,
+        "model": OLLAMA_MODEL,
         "messages": messages,
         "stream": False,
     }
 
     try:
         async with httpx.AsyncClient(timeout=LLM_TIMEOUT) as client:
-            resp = await client.post(LLM_CHAT_URL, json=payload)
+            resp = await client.post(f"{OLLAMA_URL}/api/chat", json=payload)
             if resp.status_code != 200:
-                logger.error("LLM chat failed %d: %s", resp.status_code, resp.text[:300])
+                logger.error("Ollama chat failed %d: %s", resp.status_code, resp.text[:300])
                 return ""
             data = resp.json()
-            # Go SDK returns {message: {content: "..."}} or {response: "..."}
             msg = data.get("message", {})
             if isinstance(msg, dict):
                 return msg.get("content", "")
             return data.get("response", str(msg))
     except Exception as e:
-        logger.error("LLM chat request failed: %s", e)
+        logger.error("Ollama request failed: %s", e)
         return ""