Add logic to read from COPILOT_EMBEDDING_CONFIG_DIR, if it exists.

Author: cl-gavan

packages/cloudera-ai-inference-package/cloudera_ai_inference_package/cloudera_ai_embedding_provider.py

@@ -22,6 +22,15 @@ class ClouderaAIInferenceEmbeddingModelProvider(BaseEmbeddingsProvider, Embeddin
         os.getenv("COPILOT_CONFIG_DIR", ""), model_type="embedding"
     )
 
+    # Read from both config files, as embedding models could still be in the old config file for an older CML version.
+    embedding_ai_inference_models, embedding_models = getCopilotModels(
+        os.getenv("COPILOT_EMBEDDING_CONFIG_DIR", ""), model_type="embedding"
+    )
+
+    # Merge lists, removing duplicates.
+    ai_inference_models = ai_inference_models + embedding_ai_inference_models
+    models = models + embedding_models
+
     def __init__(self, **kwargs):
         super().__init__(**kwargs)
         self.model_endpoint = self._get_inference_endpoint()

packages/jupyter-ai/jupyter_ai/handlers.py

@@ -455,6 +455,13 @@ def getConfiguredThirdPartyModels(self):
         if copilot_config and "thirdPartyModels" in copilot_config and copilot_config["thirdPartyModels"]:
             third_party_models = copilot_config["thirdPartyModels"]
 
+        copilot_embedding_config_dir = os.getenv("COPILOT_EMBEDDING_CONFIG_DIR")
+        if copilot_embedding_config_dir and os.path.exists(copilot_embedding_config_dir):
+            f = open(copilot_embedding_config_dir)
+            copilot_embedding_config = json.load(f)
+            if copilot_embedding_config and "thirdPartyModels" in copilot_embedding_config and copilot_embedding_config["thirdPartyModels"]:
+                third_party_models += copilot_embedding_config["thirdPartyModels"]
+
         # Fill in provider_id if it is missing.
         for third_party_model in third_party_models:
             if len(third_party_model["provider_id"]) == 0: