feat: Add VectorXDB Vector Search Tool

[Mithun748]

Summary

This PR adds a new VectorX Search Tool integration to the CrewAI tools ecosystem, enabling seamless semantic and hybrid vector search using the VectorX vector database.

The integration leverages Google Gemini embeddings (gemini-embedding-001) by default to generate dense vector representations and optionally supports SPLADE for sparse lexical embeddings to facilitate hybrid retrieval.

Features

• Implements the VectorXVectorSearchTool with full CrewAI BaseTool compliance.
• Supports dense vector search powered by Google Gemini embeddings.
• Optional hybrid search by integrating SPLADE sparse embeddings.
• Provides flexible embedding customization via an injectable embedding function.
• Securely manages API keys and encryption keys through environment variables.
• Robust error handling and logging during indexing and querying.
• Includes sample usage scripts demonstrating integration with CrewAI agents and tasks.
• Adds comprehensive unit tests covering upsert, query, and error scenarios.
• Updates documentation with setup, usage, and configuration instructions.

Dependencies

• vecx for VectorX database interactions.
• google-genai for Gemini embeddings and LLM support.
• Optional: transformers and torch for SPLADE support.

Testing

• Unit tests are included to mock VectorX client behavior and validate embedding workflows.
• Tested locally against a VectorX instance with real embeddings and queries.

Impact

• Provides CrewAI users access to VectorX as a powerful vector search backend.
• Enhances CrewAI’s multi-vector store architecture and extensibility.

---

Thank you for reviewing this PR! I look forward to your feedback and suggestions.

---

Note

Adds VectorX vector search tool using Gemini embeddings with optional SPLADE hybrid search, plus docs, tests, and an optional vecx dependency group.

• Tools:
  • VectorX Vector Search: Introduces VectorXVectorSearchTool with VectorXSearchArgs and SpladeSparseEmbedder for dense (Gemini) and optional SPLADE-based hybrid retrieval; supports upsert/query, error handling, and deferred embedding dimension detection. Exposed via crewai_tools and tools __init__.
• Docs:
  • Adds tools/vectorx_vector_search_tool/README.md with install, usage, and hybrid configuration guidance.
• Tests:
  • Adds tests/tools/test_vectorx_search_tool.py covering dense and hybrid flows with dummy client and SPLADE.
• Packaging:
  • Introduces optional dependency group vectorx with vecx>=0.33.1b5 in pyproject.toml.

^{Written by Cursor Bugbot for commit 6edf726. This will update automatically on new commits. Configure here.}

[Mithun748]

Hi @lucasgomide @lorenzejay, @tonykipkemboi

I’ve submitted this contribution on behalf of the VectorX DB team. Please feel free to let us know if there’s anything we should improve or adjust.

Looking forward to your feedback—thanks!