Configuring Semantic Search

Semantic search allows you to find code based on meaning rather than just keywords. To enable this in kit, you need to configure a vector embedding model and potentially a vector database backend.

Required: Embedding Function

You must provide an embedding function (embed_fn) when first accessing semantic search features via repo.get_vector_searcher() or repo.search_semantic().

This function takes a list of text strings and returns a list of corresponding embedding vectors.

from kit import Repo
# Example using a hypothetical embedding function
from my_embedding_library import get_embeddings

repo = Repo("/path/to/repo")

# Define the embedding function wrapper if necessary
def embed_fn(texts: list[str]) -> list[list[float]]:
    # Adapt this to your specific embedding library/API
    return get_embeddings(texts)

# Pass the function when searching
results = repo.search_semantic("database connection logic", embed_fn=embed_fn)

# Or when getting the searcher explicitly
vector_searcher = repo.get_vector_searcher(embed_fn=embed_fn)

Popular choices include models from OpenAI, Cohere, or open-source models via libraries like Hugging Face’s sentence-transformers.

Backend Configuration

kit’s VectorSearcher uses a pluggable backend system for storing and querying vector embeddings. Currently, the primary supported and default backend is ChromaDB.

ChromaDB (Default)

When you initialize VectorSearcher (typically via repo.get_vector_searcher()) without specifying a backend argument, kit automatically uses an instance of ChromaDBBackend.

Configuration Options:

• persist_dir (Optional[str]): This is the most important configuration option. It specifies the directory where the ChromaDB index will be stored on disk.
  • If you provide a path to repo.get_vector_searcher(persist_dir=...) or directly to the VectorSearcher constructor, that path will be used.
  • If no persist_dir is specified, kit defaults to creating the index in a subdirectory within your repository, typically at YOUR_REPO_PATH/.kit/vector_db/.
  • Persisting the index allows you to reuse it across sessions without needing to re-embed and re-index your codebase every time.

At present, other ChromaDB-specific configurations (like collection names or distance metrics) are managed internally by kit with default settings. Future versions may expose more fine-grained control.

# Example: Initialize with default ChromaDB backend and specify a persist directory
vector_searcher = repo.get_vector_searcher(
    embed_fn=my_embedding_function,
    persist_dir="./my_custom_kit_vector_index" # Index will be saved here
)

# Building the index (first time or to update)
vector_searcher.build_index()

# Later, to reuse the persisted index:
# Ensure you use the same embed_fn and persist_dir
vector_searcher_reloaded = repo.get_vector_searcher(
    embed_fn=my_embedding_function,
    persist_dir="./my_custom_kit_vector_index"
)
results = vector_searcher_reloaded.search("my query")

Other Backends

While the VectorDBBackend interface is designed to support other vector databases, ChromaDB is the primary focus for now. If you have a need for other backends like Faiss (especially for purely in-memory, non-persisted use cases) or others, please raise an issue on the kit GitHub repository.

Choosing an Embedding Model

Popular choices include models from OpenAI, Cohere, or open-source models via libraries like Hugging Face’s sentence-transformers.