-
Low-level helpers, like for: string, json or file management.
-
To higher level AI tools (which will be in a separate library in the future), including:
-
LLM querying tools (including: output parsing, fallbacks, paralellization and batching) powered by Langchain,
-
RAG toolbox, including:
-
A complete and modulable injection pipeline with: metadata handling, chunking, embedding, vector database creation, and querying.
-
A complete and modulable inference pipeline with: pre-treatment (query translation, multi-querying,metadata extraction & pre-filtering), hybrid search (BM25 & vector search), and post-treatment. This tool is itself based on:
-
A generic Workflow Executor capable of interpreting a scripted workflow defined into a YAML file which specify the workflow steps and structure, as well as parallele steps execution (as separate thread or async methods).
-
-
Agents & tools.
- To be completed.
-
The package supports flexible installation profiles to minimize dependencies based on your needs.
Basic installation (core features only - includes ChromaDB as default vector database):
pip install -e <CommonToolsPath>Installation with optional features:
# Install with Pinecone vector database support (requires C++ redistributable for pinecone-text)
pip install -e <CommonToolsPath>[pinecone]
# Install with Qdrant vector database support
pip install -e <CommonToolsPath>[qdrant]
# Install with database support (SQLite + PostgreSQL)
pip install -e <CommonToolsPath>[database]
# Install with ML/scientific computing dependencies (scikit-learn, scipy, pandas)
pip install -e <CommonToolsPath>[ml]
# Install with advanced features (langgraph, langsmith, ragas)
pip install -e <CommonToolsPath>[advanced]
# Install with multiple optional features
pip install -e <CommonToolsPath>[pinecone,qdrant,database,ml,advanced]
# Install everything (all optional dependencies)
pip install -e <CommonToolsPath>[full]Environment variable-based installation (for CI/CD):
# Windows
set COMMON_TOOLS_INSTALL_MODE=full
pip install -e <CommonToolsPath>
# Linux/Mac
COMMON_TOOLS_INSTALL_MODE=full pip install -e <CommonToolsPath>
# Install specific profiles (comma-separated)
set COMMON_TOOLS_INSTALL_MODE=pinecone,database,ml
pip install -e <CommonToolsPath>Available installation profiles:
pinecone- Pinecone vector database (requires C++ redistributable)qdrant- Qdrant vector databasedatabase- Database support (SQLite + PostgreSQL)ml- ML/scientific computing (scikit-learn, scipy, pandas)advanced- Advanced AI features (langgraph, langsmith, ragas)vectordb- Both Pinecone and Qdrantfull- All optional dependencies
Tips:
- Look into
setup.pyfile, the dependencies are organized into core and optional extras. - Replace
<CommonToolsPath>with the actual path of your local "common tools" project root folder. - To build the package 'common_tools', simply run the command: "libs_build.bat", from within the "common tools" root folder. The built package will be found in the "dist" folder with an auto-incremented version. Copy the '*.whl' file where you want this library to be referenced.