Use model groups in config

README.md

@@ -9,7 +9,7 @@ This demo showcases a chatbot built with Rasa's LLM-native approach: [CALM](http
 > CALM's current stage of development.
 
 > [!NOTE]
-> This demo bot is currently compatible with `3.11.0rc2`.
+> This demo bot is currently compatible with `3.11.2`.
 
 ## Terms of Use
 

config/config.yml

@@ -28,10 +28,7 @@ pipeline:
 - name: NLUCommandAdapter
 - name: SingleStepLLMCommandGenerator
   llm:
-    model: gpt-4
-    timeout: 7
-    temperature: 0.0
-    top_p: 0.0
+    model_group: openai-direct-gpt-4
 
 policies:
 - name: RulePolicy

config/multistep-config.yml

@@ -28,10 +28,7 @@ pipeline:
 - name: NLUCommandAdapter
 - name: MultiStepLLMCommandGenerator
   llm:
-    model: gpt-3.5-turbo-0125
-    timeout: 7
-    temperature: 0.0
-    top_p: 0.0
+    model_group: openai-direct-gpt-3.5-turbo
 
 policies:
 - name: RulePolicy

config/qdrant-config.yml

@@ -28,10 +28,7 @@ pipeline:
 - name: NLUCommandAdapter
 - name: SingleStepLLMCommandGenerator
   llm:
-    model: gpt-4
-    timeout: 7
-    temperature: 0.0
-    top_p: 0.0
+    model_group: openai-direct-gpt-4
 
 policies:
 - name: RulePolicy

endpoints.yml

@@ -71,4 +71,20 @@ vector_store:
   host: 0.0.0.0
   port: 6334
   content_payload_key: page_content
-  metadata_payload_key: metadata
+  metadata_payload_key: metadata
+
+model_groups:
+  - id: openai-direct-gpt-4
+    models:
+      - provider: openai
+        model: gpt-4
+        timeout: 7
+        temperature: 0.0
+        top_p: 0.0
+  - id: openai-direct-gpt-3.5-turbo
+    models:
+      - provider: openai
+        model: gpt-3.5-turbo-0125
+        timeout: 7
+        temperature: 0.0
+        top_p: 0.0