Run llama directly

Signed-off-by: Fabian von Feilitzsch <fabian@fabianism.us>

Author: fabianvf

roles/tackle/templates/kai/llm-proxy-configmap.yaml.j2

@@ -9,41 +9,9 @@ metadata:
     app.kubernetes.io/component: kai
     app.kubernetes.io/part-of: tackle
 data:
-  lightspeed-stack.yaml: |
-    # Minimal Lightspeed Stack configuration for running ONLY Llama Stack server
-    # This config disables all LCS features and just uses the project to start Llama Stack
-    name: Llama Stack Server Only
-    service:
-      # LCS service won't actually run with this config - we're just using it for Llama Stack
-      host: 0.0.0.0
-      port: 8080
-      auth_enabled: false
-      workers: 1
-    
-    # This is the key - we're NOT using Llama Stack as a library or connecting to external server
-    # Instead, we'll use the project's ability to start Llama Stack as a separate server
-    llama_stack:
-      use_as_library_client: false
-      # Dummy URL since we're not actually connecting to anything
-      url: http://localhost:8321
-      api_key: unused
-    
-    # Disable all LCS features since we only want Llama Stack
-    user_data_collection:
-      feedback_enabled: false
-      transcripts_enabled: false
-    
-    # Use in-memory cache (minimal)
-    conversation_cache:
-      type: "noop"
-    
-    # No authentication
-    authentication:
-      module: "noop"
   run.yaml: |
     version: 2
     image_name: starter
-    external_providers_dir: /tmp/.llama/providers.d
     apis:
     - inference
     providers:

roles/tackle/templates/kai/llm-proxy-deployment.yaml.j2

@@ -28,9 +28,13 @@ spec:
       containers:
         - name: llm-proxy
           image: "{{ kai_llm_proxy_image_fqin }}"
+          command: ["/app-root/.venv/bin/llama"]
           args:
-            - --config
-            - /app/lightspeed-stack.yaml
+            - stack
+            - run
+            - /app/run.yaml
+            - --port
+            - "8321"
           ports:
             - containerPort: 8321
               name: http
@@ -58,14 +62,6 @@ spec:
             successThreshold: 1
             failureThreshold: 3
           env:
-            - name: RUN_CONFIG_PATH
-              value: /app/run.yaml
-            - name: INPUT_CONFIG
-              value: /app/run.yaml
-            - name: LIGHTSPEED_STACK_CONFIG_PATH
-              value: /app/lightspeed-stack.yaml
-            - name: LLAMA_STACK_CONFIG_DIR
-              value: /tmp/.llama
 {% if trusted_ca_enabled | default(false) | bool %}
             # For self-signed certificates
             - name: SSL_CERT_FILE