Add skip_prefilter option to control generator

Local LLM prefilter (llama3.2 3B) was incorrectly skipping annex chunks
that contain concrete requirements. Added skip_prefilter flag to bypass
the local pre-filter and send all chunks directly to Anthropic API.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

backend-compliance/compliance/api/control_generator_routes.py

@@ -54,6 +54,7 @@ class GenerateRequest(BaseModel):
     skip_web_search: bool = False
     dry_run: bool = False
     regulation_filter: Optional[List[str]] = None  # Only process these regulation_code prefixes
+    skip_prefilter: bool = False  # Skip local LLM pre-filter, send all chunks to API
 
 
 class GenerateResponse(BaseModel):
@@ -146,6 +147,7 @@ async def start_generation(req: GenerateRequest):
         skip_web_search=req.skip_web_search,
         dry_run=req.dry_run,
         regulation_filter=req.regulation_filter,
+        skip_prefilter=req.skip_prefilter,
     )
 
     if req.dry_run:

backend-compliance/compliance/services/control_generator.py

@@ -385,6 +385,7 @@ class GeneratorConfig(BaseModel):
     dry_run: bool = False
     existing_job_id: Optional[str] = None  # If set, reuse this job instead of creating a new one
     regulation_filter: Optional[List[str]] = None  # Only process chunks matching these regulation_code prefixes
+    skip_prefilter: bool = False  # If True, skip local LLM pre-filter (send all chunks to API)
 
 
 @dataclass
@@ -1886,7 +1887,7 @@ Kategorien: {CATEGORY_LIST_STR}"""
                         self._update_job(job_id, result)
 
                     # Stage 1.5: Local LLM pre-filter — skip chunks without requirements
-                    if not config.dry_run:
+                    if not config.dry_run and not config.skip_prefilter:
                         is_relevant, prefilter_reason = await _prefilter_chunk(chunk.text)
                         if not is_relevant:
                             chunks_skipped_prefilter += 1