fix utf-8 encoding issue, add embedding_model configuration

nemoguardrails/actions/llm/generation.py

@@ -71,6 +71,14 @@ def __init__(self, config: RailsConfig, llm: BaseLLM, verbose: bool = False):
         self.kb = None
         self._init_kb()
 
+        # If we have a customized embedding model, we'll use it.
+        self.embedding_model = "all-MiniLM-L6-v2"
+        for model in self.config.models:
+            if 'embedding' in model.type:
+                self.embedding_model = model.model
+                assert model.engine == "SentenceTransformer"
+                break
+
     def _init_user_message_index(self):
         """Initializes the index of user messages."""
 
@@ -86,7 +94,7 @@ def _init_user_message_index(self):
         if len(items) == 0:
             return
 
-        self.user_message_index = BasicEmbeddingsIndex()
+        self.user_message_index = BasicEmbeddingsIndex(self.embedding_model)
         self.user_message_index.add_items(items)
 
         # NOTE: this should be very fast, otherwise needs to be moved to separate thread.
@@ -107,7 +115,7 @@ def _init_bot_message_index(self):
         if len(items) == 0:
             return
 
-        self.bot_message_index = BasicEmbeddingsIndex()
+        self.bot_message_index = BasicEmbeddingsIndex(self.embedding_model)
         self.bot_message_index.add_items(items)
 
         # NOTE: this should be very fast, otherwise needs to be moved to separate thread.
@@ -141,7 +149,7 @@ def _init_flows_index(self):
         if len(items) == 0:
             return
 
-        self.flows_index = BasicEmbeddingsIndex()
+        self.flows_index = BasicEmbeddingsIndex(self.embedding_model)
         self.flows_index.add_items(items)
 
         # NOTE: this should be very fast, otherwise needs to be moved to separate thread.
@@ -154,7 +162,7 @@ def _init_kb(self):
             return
 
         documents = [doc.content for doc in self.config.docs]
-        self.kb = KnowledgeBase(documents=documents)
+        self.kb = KnowledgeBase(documents=documents, embedding_model=self.embedding_model)
         self.kb.init()
         self.kb.build()
 

nemoguardrails/kb/basic.py

@@ -28,10 +28,11 @@ class BasicEmbeddingsIndex(EmbeddingsIndex):
     It uses Annoy to perform the search.
     """
 
-    def __init__(self, index=None):
+    def __init__(self, embedding_model=None, index=None):
         self._model = None
         self._items = []
         self._embeddings = []
+        self.embedding_model = embedding_model
 
         # When the index is provided, it means it's from the cache.
         self._index = index
@@ -42,7 +43,7 @@ def embeddings_index(self):
 
     def _init_model(self):
         """Initialize the model used for computing the embeddings."""
-        self._model = SentenceTransformer("all-MiniLM-L6-v2")
+        self._model = SentenceTransformer(self.embedding_model)
 
     def _get_embeddings(self, texts: List[str]) -> List[List[float]]:
         """Compute embeddings for a list of texts."""

nemoguardrails/kb/kb.py

@@ -33,10 +33,11 @@
 class KnowledgeBase:
     """Basic implementation of a knowledge base."""
 
-    def __init__(self, documents: List[str]):
+    def __init__(self, documents: List[str], embedding_model: str):
         self.documents = documents
         self.chunks = []
         self.index = None
+        self.embedding_model = embedding_model
 
     def init(self):
         """Initialize the knowledge base.
@@ -79,10 +80,10 @@ def build(self):
             ann_index = AnnoyIndex(embedding_size, "angular")
             ann_index.load(cache_file)
 
-            self.index = BasicEmbeddingsIndex(index=ann_index)
+            self.index = BasicEmbeddingsIndex(embedding_model=self.embedding_model, index=ann_index)
             self.index.add_items(index_items)
         else:
-            self.index = BasicEmbeddingsIndex()
+            self.index = BasicEmbeddingsIndex(self.embedding_model)
             self.index.add_items(index_items)
             self.index.build()
 

nemoguardrails/rails/llm/config.py

@@ -210,11 +210,11 @@ def from_path(config_path: str):
                                 )
 
                     elif file.endswith(".yml") or file.endswith(".yaml"):
-                        with open(full_path) as f:
+                        with open(full_path, 'r', encoding='utf-8') as f:
                             _raw_config = yaml.safe_load(f.read())
 
                     elif file.endswith(".co"):
-                        with open(full_path) as f:
+                        with open(full_path, 'r', encoding='utf-8') as f:
                             _raw_config = parse_colang_file(file, content=f.read())
 
                     _join_config(raw_config, _raw_config)