Handle Empty Titles (#1891)

backend/danswer/indexing/embedder.py

@@ -100,15 +100,16 @@ class DefaultIndexingEmbedder(IndexingEmbedder):
         # Drop any None or empty strings
         chunk_titles_list = [title for title in chunk_titles if title]
 
-        title_embeddings = self.embedding_model.encode(
-            chunk_titles_list, text_type=EmbedTextType.PASSAGE
-        )
-        title_embed_dict.update(
-            {
-                title: vector
-                for title, vector in zip(chunk_titles_list, title_embeddings)
-            }
-        )
+        if chunk_titles_list:
+            title_embeddings = self.embedding_model.encode(
+                chunk_titles_list, text_type=EmbedTextType.PASSAGE
+            )
+            title_embed_dict.update(
+                {
+                    title: vector
+                    for title, vector in zip(chunk_titles_list, title_embeddings)
+                }
+            )
 
         # Mapping embeddings to chunks
         embedding_ind_start = 0

backend/danswer/search/search_nlp_models.py

@@ -112,6 +112,10 @@ class EmbeddingModel:
         text_type: EmbedTextType,
         batch_size: int = BATCH_SIZE_ENCODE_CHUNKS,
     ) -> list[list[float]]:
+        if not texts:
+            logger.warning("No texts to be embedded")
+            return []
+
         if self.provider_type:
             embed_request = EmbedRequest(
                 model_name=self.model_name,

backend/model_server/encoders.py

@@ -284,6 +284,9 @@ def calc_sim_scores(query: str, docs: list[str]) -> list[list[float]]:
 async def process_embed_request(
     embed_request: EmbedRequest,
 ) -> EmbedResponse:
+    if not embed_request.texts:
+        raise HTTPException(status_code=400, detail="No texts to be embedded")
+
     try:
         if embed_request.text_type == EmbedTextType.QUERY:
             prefix = embed_request.manual_query_prefix
@@ -315,6 +318,11 @@ async def process_rerank_request(embed_request: RerankRequest) -> RerankResponse
     if INDEXING_ONLY:
         raise RuntimeError("Indexing model server should not call intent endpoint")
 
+    if not embed_request.documents or not embed_request.query:
+        raise HTTPException(
+            status_code=400, detail="No documents or query to be reranked"
+        )
+
     try:
         sim_scores = calc_sim_scores(
             query=embed_request.query, docs=embed_request.documents