llm, handle max_tokens a little more dynamically, likely only affects the smaller models.

tasks/workstation/linux/software/llm.yml

@@ -832,11 +832,13 @@
           {% if llm.model in llm_models.stdout %}
           {% if "chat" in llm.purpose %}
 
-          [providers.models.ollama.{{ llm.provider }}]
+          [providers.models.ollama.{{ llm.provider }}_{{ 'CODE' if "code" in llm.purpose else 'CHAT'
+            }}{{ '' if llm.model in llm_models.stdout else '-MISSING'
+            }}]
           model = "{{ llm.model }}"
           timeout_secs = 3600 # 1 hour, default is 120 (2 minutes), issue happens between 5 and 6 minutes though.
           context_window = {{ [llm.context] | min | int }}
-          max_tokens = {{ [(llm.context / 8)] | min | int }}
+          max_tokens = {{ [(llm.context / 8), 16000] | min | int }}
 
           {% endif %}
           {% endif %}
@@ -865,19 +867,22 @@
             {% if llm.model in llm_models.stdout %}
             {% if "code" in llm.purpose %}
 
-            - name: {{ llm.provider }}_CHAT
+            - name: {{ llm.provider }}_{{ 'CODE' if "code" in llm.purpose else 'CHAT'
+                }}{{ '' if llm.model in llm_models.stdout else '-MISSING'
+                }}
               provider: ollama
               model: {{ llm.model }}
               defaultCompletionOptions:
                 contextLength: {{ [llm.context] | min | int }}
-                maxTokens: {{ [16000] | min | int }}
+                maxTokens: {{ [(llm.context / 8), 16000] | min  | int }}
 
-            - name: {{ llm.provider }}_AUTO
+            - name: {{ llm.provider }}_AUTO{{ '' if llm.model in llm_models.stdout else '-MISSING'
+                    }}
               provider: ollama
               model: {{ llm.model }}
               defaultCompletionOptions:
                 contextLength: {{ [llm.context] | min | int }}
-                maxTokens: {{ [4000] | min | int }}
+                maxTokens: {{ [(llm.context / 8), 4000] | min | int }}
               roles:
                 - autocomplete
 
@@ -918,10 +923,10 @@
                     },
                     "timeout": 600000,
                     "streamIdleTimeoutMs": 600000,
-                    "contextWindowSize": {{ llm.context }},
+                    "contextWindowSize": {{ [llm.context] | min | int }},
                     "maxRetries": 3,
                     "samplingParams": {
-                      "max_tokens": 8000
+                      "max_tokens": {{ [(llm.context / 8), 16000] | min | int }}
                       ,"temperature": 0.7
                     }
                   }