fix: rename generation model setting to ollama_llm_model to avoid OLLAMA_MODEL env collision

The deployed container sets OLLAMA_MODEL=nomic-embed-text for embeddings,
which shadowed the generation-model setting and broke Phase 0 keyword
extraction and Phase 4 LLM re-ranking on every request. The setting is now
ollama_llm_model (env: OLLAMA_LLM_MODEL, default gemma4:e2b), startup logs
the resolved generation model, and Phase 0/Phase 4 LLM calls are wrapped in
a 12s asyncio.wait_for with graceful fallback so a hung call cannot gate
retrieval for the full 120s client timeout.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

src/services/consolidation_service.py

@@ -505,7 +505,7 @@ JSON:"""
             # Call Ollama for analysis (temperature=0.0 for consistent classification)
             response = await self.ollama.generate_text(
                 prompt=prompt,
-                model=self.settings.ollama_model,
+                model=self.settings.ollama_llm_model,
                 stream=False,
                 temperature=0.0
             )
@@ -1079,7 +1079,7 @@ JSON:"""
         try:
             response = await self.ollama.generate_text(
                 prompt=prompt,
-                model=self.settings.ollama_model,
+                model=self.settings.ollama_llm_model,
                 stream=False,
                 temperature=0.0
             )

src/services/hybrid_rag_service.py

@@ -34,6 +34,10 @@ from src.core.multi_tenancy import get_neo4j_user_base_label, get_neo4j_user_lab
 
 logger = logging.getLogger(__name__)
 
+# Timeout for auxiliary LLM calls (keyword extraction, re-ranking).
+# A hung Ollama call must not gate retrieval for the full client timeout.
+LLM_CALL_TIMEOUT_SECONDS = 12.0
+
 
 class HybridRAGService:
     """
@@ -69,7 +73,7 @@ class HybridRAGService:
         self.content_extractor = content_extractor
         self.settings = settings
         self.volatile = volatile_service
-        self.reranker_model = settings.ollama_model
+        self.reranker_model = settings.ollama_llm_model
 
     async def search(
         self,
@@ -224,10 +228,13 @@ Return format:
 JSON:"""
 
         try:
-            response = await self.ollama.generate_text(
-                prompt=prompt,
-                model=self.reranker_model,
-                temperature=0.0  # Deterministic for consistent extraction
+            response = await asyncio.wait_for(
+                self.ollama.generate_text(
+                    prompt=prompt,
+                    model=self.reranker_model,
+                    temperature=0.0  # Deterministic for consistent extraction
+                ),
+                timeout=LLM_CALL_TIMEOUT_SECONDS
             )
 
             # Parse JSON response (handle potential extra text)
@@ -261,6 +268,16 @@ JSON:"""
                 "synonyms": {},
                 "expansions": {}
             }
+        except asyncio.TimeoutError:
+            logger.warning(
+                f"Keyword extraction timed out after {LLM_CALL_TIMEOUT_SECONDS}s, using fallback"
+            )
+            return {
+                "core_keywords": query.split(),
+                "entities": [],
+                "synonyms": {},
+                "expansions": {}
+            }
         except Exception as e:
             logger.error(f"Keyword extraction failed: {e}", exc_info=True)
             return {
@@ -772,10 +789,13 @@ Example output: 3,1,5,2,4
 
 Ranking:"""
 
-            response = await self.ollama.generate_text(
-                prompt=prompt,
-                model=self.reranker_model,
-                temperature=0.0  # Deterministic for consistent rankings
+            response = await asyncio.wait_for(
+                self.ollama.generate_text(
+                    prompt=prompt,
+                    model=self.reranker_model,
+                    temperature=0.0  # Deterministic for consistent rankings
+                ),
+                timeout=LLM_CALL_TIMEOUT_SECONDS
             )
 
             # Parse response: "3,1,5,2,4" → [2, 0, 4, 1, 3] (0-indexed)
@@ -796,6 +816,11 @@ Ranking:"""
             logger.info(f"LLM re-ranking: reordered {len(reranked)} results")
             return reranked
 
+        except asyncio.TimeoutError:
+            logger.warning(
+                f"LLM re-ranking timed out after {LLM_CALL_TIMEOUT_SECONDS}s, using RRF order"
+            )
+            return results  # Fallback to RRF order
         except Exception as e:
             logger.warning(f"LLM re-ranking failed: {e}, using RRF order")
             return results  # Fallback to RRF order

src/services/wiki_page_writer.py

@@ -34,7 +34,7 @@ class WikiPageWriter:
             settings: Application settings
         """
         self.ollama = ollama_client
-        self.model = settings.ollama_model
+        self.model = settings.ollama_llm_model
 
     async def create_page(
         self,