Merge pull request '[loop-cycle-7] feat: add OLLAMA_NUM_CTX config (#83)' (#98) from fix/num-ctx-remaining into main

Reviewed-on: http://localhost:3000/rockachopa/Timmy-time-dashboard/pulls/98

.env.example

@@ -17,6 +17,11 @@
 # LLM model to use via Ollama (default: qwen3.5:latest)
 # OLLAMA_MODEL=qwen3.5:latest
 
+# Ollama context window size (default: 4096 tokens)
+# Set higher for more context, lower to save RAM. 0 = model default.
+# qwen3:30b + 4096 ctx ≈ 19GB VRAM; default ctx ≈ 45GB.
+# OLLAMA_NUM_CTX=4096
+
 # Enable FastAPI interactive docs at /docs and /redoc (default: false)
 # DEBUG=true
 

config/providers.yaml

@@ -28,6 +28,7 @@ providers:
       - name: qwen3.5:latest
         default: true
         context_window: 128000
+        # Note: actual context is capped by OLLAMA_NUM_CTX (default 4096) to save RAM
         capabilities: [text, tools, json, streaming]
       - name: llama3.1:8b-instruct
         context_window: 128000

src/config.py

@@ -22,6 +22,11 @@ class Settings(BaseSettings):
     # llama3.2 (3B) hallucinated tool output consistently in testing.
     ollama_model: str = "qwen3.5:latest"
 
+    # Context window size for Ollama inference — override with OLLAMA_NUM_CTX
+    # qwen3:30b with default context eats 45GB on a 39GB Mac.
+    # 4096 keeps memory at ~19GB. Set to 0 to use model defaults.
+    ollama_num_ctx: int = 4096
+
     # Fallback model chains — override with FALLBACK_MODELS / VISION_FALLBACK_MODELS
     # as comma-separated strings, e.g. FALLBACK_MODELS="qwen3.5:latest,llama3.1"
     # Or edit config/providers.yaml → fallback_chains for the canonical source.

src/timmy/agent.py

@@ -306,9 +306,12 @@ def create_timmy(
         logger.warning("Failed to load memory context: %s", exc)
         full_prompt = base_prompt
 
+    model_kwargs = {}
+    if settings.ollama_num_ctx > 0:
+        model_kwargs["options"] = {"num_ctx": settings.ollama_num_ctx}
     agent = Agent(
         name="Agent",
-        model=Ollama(id=model_name, host=settings.ollama_url, timeout=300),
+        model=Ollama(id=model_name, host=settings.ollama_url, timeout=300, **model_kwargs),
         db=SqliteDb(db_file=db_file),
         description=full_prompt,
         add_history_to_context=True,

src/timmy/agents/base.py

@@ -73,9 +73,12 @@ class BaseAgent(ABC):
                 if handler:
                     tool_instances.append(handler)
 
+        ollama_kwargs = {}
+        if settings.ollama_num_ctx > 0:
+            ollama_kwargs["options"] = {"num_ctx": settings.ollama_num_ctx}
         return Agent(
             name=self.name,
-            model=Ollama(id=self.model, host=settings.ollama_url, timeout=300),
+            model=Ollama(id=self.model, host=settings.ollama_url, timeout=300, **ollama_kwargs),
             description=system_prompt,
             tools=tool_instances if tool_instances else None,
             add_history_to_context=True,

tests/timmy/test_agent.py

@@ -67,6 +67,7 @@ def test_create_timmy_respects_custom_ollama_url():
     ):
         mock_settings.ollama_model = "llama3.2"
         mock_settings.ollama_url = custom_url
+        mock_settings.ollama_num_ctx = 4096
         mock_settings.timmy_model_backend = "ollama"
         mock_settings.airllm_model_size = "70b"
 
@@ -237,6 +238,7 @@ def test_create_timmy_includes_tools_for_large_model():
     ):
         mock_settings.ollama_model = "llama3.1"
         mock_settings.ollama_url = "http://localhost:11434"
+        mock_settings.ollama_num_ctx = 4096
         mock_settings.timmy_model_backend = "ollama"
         mock_settings.airllm_model_size = "70b"
         mock_settings.telemetry_enabled = False