From c0dad7503b79369a5ccdfc3b2a578e7b4d076fac Mon Sep 17 00:00:00 2001
From: "codeflash-ai[bot]"
 <148906541+codeflash-ai[bot]@users.noreply.github.com>
Date: Sun, 26 Oct 2025 07:24:08 +0000
Subject: [PATCH] Optimize
 SparseVectorIndexConfig.validate_embedding_function_field

The optimization introduces **signature caching** to eliminate redundant computation during validation. The key changes are:

1. **Cached protocol signature**: The protocol signature `signature(SparseEmbeddingFunction.__call__).parameters.keys()` is computed once and cached globally, rather than being recomputed on every validation call.

2. **Tuple conversion for faster comparison**: Both signatures are converted to tuples instead of comparing `dict_keys` objects directly, which provides faster equality comparison in Python.

3. **Lazy initialization**: The protocol signature is computed only when first needed via `_get_protocol_signature()`, avoiding any import-time overhead.

**Why this leads to speedup**: The `inspect.signature()` function performs introspection on the method, which involves parsing the function's metadata. This is computationally expensive when done repeatedly. By caching the protocol signature (which never changes), we eliminate this repeated work. The tuple conversion also optimizes the comparison operation itself.

**Test case performance patterns**: The optimization shows consistent 30-40% speedups across all test cases that involve signature validation (e.g., `test_valid_sparse_embedding_function`: 38.9% faster, `test_multiple_instances_large_scale`: 90.3% faster). The most dramatic improvements occur in scenarios with multiple validations, where the caching benefit compounds. Simple cases like `test_none_embedding_function` show minimal impact since they bypass signature validation entirely.
---
 chromadb/api/types.py | 19 +++++++++++++++----
 1 file changed, 15 insertions(+), 4 deletions(-)

diff --git a/chromadb/api/types.py b/chromadb/api/types.py
index 9c45112809f..fa26915c0a3 100644
--- a/chromadb/api/types.py
+++ b/chromadb/api/types.py
@@ -53,6 +53,8 @@
 import struct
 import math
 
+_cached_protocol_signature = None
+
 # Re-export types from chromadb.types
 __all__ = [
     "Metadata",
@@ -1480,10 +1482,10 @@ def validate_sparse_embedding_function(
     sparse_vector_function: SparseEmbeddingFunction[Embeddable],
 ) -> None:
     """Validate that a sparse vector function conforms to the SparseEmbeddingFunction protocol."""
-    function_signature = signature(
-        sparse_vector_function.__class__.__call__
-    ).parameters.keys()
-    protocol_signature = signature(SparseEmbeddingFunction.__call__).parameters.keys()
+    function_signature = tuple(
+        signature(sparse_vector_function.__class__.__call__).parameters.keys()
+    )
+    protocol_signature = _get_protocol_signature()
 
     if not function_signature == protocol_signature:
         raise ValueError(
@@ -1492,6 +1494,15 @@ def validate_sparse_embedding_function(
         )
 
 
+def _get_protocol_signature():
+    global _cached_protocol_signature
+    if _cached_protocol_signature is None:
+        _cached_protocol_signature = tuple(
+            signature(SparseEmbeddingFunction.__call__).parameters.keys()
+        )
+    return _cached_protocol_signature
+
+
 # Index Configuration Types for Collection Schema
 class FtsIndexConfig(BaseModel):
     """Configuration for Full-Text Search index. No parameters required."""