feat: make diskann runnable

packages/leann-core/src/leann/api.py

@@ -450,7 +450,7 @@ class LeannSearcher:
             prune_ratio=prune_ratio,
             recompute_embeddings=recompute_embeddings,
             pruning_strategy=pruning_strategy,
-            expected_zmq_port=zmq_port,
+            zmq_port=zmq_port,
             **kwargs,
         )
         search_time = time.time() - start_time

packages/leann-core/src/leann/embedding_compute.py

@@ -60,6 +60,9 @@ def compute_embeddings_sentence_transformers(
     """
     Compute embeddings using SentenceTransformer with model caching
     """
+    # Handle empty input
+    if not texts:
+        raise ValueError("Cannot compute embeddings for empty text list")
     logger.info(
         f"Computing embeddings for {len(texts)} texts using SentenceTransformer, model: '{model_name}'"
     )

packages/leann-core/src/leann/interface.py

@@ -64,7 +64,7 @@ class LeannBackendSearcherInterface(ABC):
             prune_ratio: Ratio of neighbors to prune via approximate distance (0.0-1.0)
             recompute_embeddings: Whether to fetch fresh embeddings from server vs use stored PQ codes
             pruning_strategy: PQ candidate selection strategy - "global" (default), "local", or "proportional"
-            zmq_port: ZMQ port for embedding server communication
+            zmq_port: ZMQ port for embedding server communication. Must be provided if recompute_embeddings is True.
             **kwargs: Backend-specific parameters
 
         Returns:

packages/leann-core/src/leann/searcher_base.py

@@ -104,6 +104,10 @@ class BaseSearcher(LeannBackendSearcherInterface, ABC):
         # Try to use embedding server if available and requested
         if use_server_if_available:
             try:
+                # TODO: Maybe we can directly use this port here?
+                # For this internal method, it's ok to assume that the server is running
+                # on that port?
+
                 # Ensure we have a server with passages_file for compatibility
                 passages_source_file = (
                     self.index_dir / f"{self.index_path.name}.meta.json"
@@ -181,7 +185,7 @@ class BaseSearcher(LeannBackendSearcherInterface, ABC):
             prune_ratio: Ratio of neighbors to prune via approximate distance (0.0-1.0)
             recompute_embeddings: Whether to fetch fresh embeddings from server vs use stored PQ codes
             pruning_strategy: PQ candidate selection strategy - "global" (default), "local", or "proportional"
-            zmq_port: ZMQ port for embedding server communication
+            zmq_port: ZMQ port for embedding server communication. Must be provided if recompute_embeddings is True.
             **kwargs: Backend-specific parameters (e.g., batch_size, dedup_node_dis, etc.)
 
         Returns: