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>

.env.example

@@ -9,7 +9,7 @@ SEARXNG_URL=http://192.168.86.149:8080
 REDIS_HOST=192.168.86.149
 PAPERLESS_URL=http://192.168.86.149:8091
 
-OLLAMA_MODEL=mistral-nemo-large:latest
+OLLAMA_LLM_MODEL=gemma4:e2b
 OLLAMA_EMBEDDING_MODEL=nomic-embed-text
 
 # Wiki.js auth

CHANGELOG.md

@@ -5,6 +5,13 @@ All notable changes to Library Desk will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [Unreleased]
+
+### Fixed
+
+- **Ollama generation model env collision** - Renamed the generation-model setting `ollama_model` to `ollama_llm_model` (env: `OLLAMA_LLM_MODEL`, default `gemma4:e2b`). The container env `OLLAMA_MODEL=nomic-embed-text` (meant for embeddings) was shadowing the generation model, breaking Phase 0 keyword extraction and Phase 4 LLM re-ranking on every request. Startup now logs the resolved generation model.
+- **LLM call timeouts** - Phase 0 keyword extraction and Phase 4 re-ranking are wrapped in a 12s `asyncio.wait_for` with graceful fallback, so a hung Ollama call can no longer gate retrieval for the full 120s client timeout.
+
 ## [1.7.3] - 2026-01-07
 
 ### Fixed

src/config.py

@@ -66,7 +66,9 @@ class Settings(BaseSettings):
 
     # Ollama Configuration
     ollama_url: str = Field(default="http://ollama:11434", description="Ollama URL")
-    ollama_model: str = Field(default="mistral-nemo-large:latest", description="Ollama LLM model")
+    # Named ollama_llm_model (env: OLLAMA_LLM_MODEL) to avoid collision with the
+    # OLLAMA_MODEL container env var, which is used for the embedding model.
+    ollama_llm_model: str = Field(default="gemma4:e2b", description="Ollama LLM model for generation (keyword extraction, re-ranking, consolidation)")
     ollama_embedding_model: str = Field(default="nomic-embed-text", description="Ollama embedding model")
 
     # HybridRAG Configuration

src/main.py

@@ -142,7 +142,7 @@ async def health(settings: Settings = Depends(get_settings)) -> HealthResponse:
             },
             "ollama": {
                 "url": settings.ollama_url,
-                "model": settings.ollama_model,
+                "model": settings.ollama_llm_model,
                 "healthy": service_health.get("ollama", False)
             }
         }
@@ -412,6 +412,7 @@ async def startup_event():
     logger.info(f"Wiki.js: {settings.wikijs_url}")
     logger.info(f"SearXNG: {settings.searxng_url}")
     logger.info(f"Ollama: {settings.ollama_url}")
+    logger.info(f"Ollama generation model: {settings.ollama_llm_model} (embedding model: {settings.ollama_embedding_model})")
 
     # Initialize all service clients
     await startup_clients()

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(
+            response = await asyncio.wait_for(
-                prompt=prompt,
+                self.ollama.generate_text(
-                model=self.reranker_model,
+                    prompt=prompt,
-                temperature=0.0  # Deterministic for consistent extraction
+                    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(
+            response = await asyncio.wait_for(
-                prompt=prompt,
+                self.ollama.generate_text(
-                model=self.reranker_model,
+                    prompt=prompt,
-                temperature=0.0  # Deterministic for consistent rankings
+                    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,

tests/test_consolidation.py

@@ -38,7 +38,7 @@ def settings():
     """Get mocked application settings for testing."""
     mock_settings = MagicMock()
     mock_settings.reranker_model = "mistral-nemo"
-    mock_settings.ollama_model = "mistral-nemo"
+    mock_settings.ollama_llm_model = "mistral-nemo"
     return mock_settings