Skip to content

Installation

From PyPI

pip install tdfs4ds

From source

git clone https://github.com/denismolin/tdfs4ds.git
cd tdfs4ds
pip install -e .

Dependencies

tdfs4ds declares the following runtime dependencies in setup.py:

Library Purpose
teradataml >= 17.20 Teradata client used for every query
pandas, numpy DataFrame primitives
plotly, tqdm, networkx Lineage Sankey + progress bars + graph algebra
sqlparse Lightweight SQL parsing for the lineage walker
langchain_openai, langchain_aws, langchain_core, langchain_chroma, langchain_teradata LLM and embedding adapters
langgraph StateGraph orchestrating the consumer agent and query optimizer
chromadb Persistent vector store for the consumer agent
pydantic Structured output for intent classification
gradio Chatbot UI
nbformat Notebook utilities

LangChain / LangGraph / Gradio dependencies are lazy-loaded — they are imported only when the genai or agent sub-modules are first accessed, so import tdfs4ds stays fast.

Verifying the install

import tdfs4ds
print(tdfs4ds.__version__)        # 0.2.9.9 or newer
print(tdfs4ds.list_skills())      # 22 bundled agent skills

Installing the agent skills

The package ships with 22 discoverable SKILL.md files organized in four families that teach Claude Code (and compatible agents) how to drive the feature store end-to-end:

  • fs-* (10 skills) — Feature store operations (upload, document, analyze, etc.)
  • qo-skill-* (3 skills) — Query optimization with nested prompts
  • td-* (1 skill) — Teradata utilities (EXPLAIN visualization)
  • ca-* (9 skills) — Consumer agent intents (discovered dynamically when using the agent)

Python API

import tdfs4ds

# Copy all skills to ~/.claude/skills/ (Claude Code user-level, all projects)
tdfs4ds.install_skills()

# Or into the current project only
tdfs4ds.install_skills(target_dir='.claude/skills')

# Only add skills that are not present yet (safe for shared project dirs)
tdfs4ds.export_skills('.claude/skills')        # overwrite=False by default

# Install a specific subset
tdfs4ds.install_skills(skills=['fs-setup', 'fs-upload'])

CLI (with the venv active)

# User-level install — picked up in every project
tdfs4ds-install-skills

# Project-level install
tdfs4ds-install-skills --target .claude/skills

# Skip skills already present
tdfs4ds-install-skills --target .claude/skills --no-overwrite

# List bundled skill names
tdfs4ds-install-skills --list

Sharing with teammates

If the project repository already contains .claude/skills/, teammates get the skills automatically on git pull — no install step needed.

If each person installs tdfs4ds into their own virtual environment, they run tdfs4ds.install_skills() (or the CLI equivalent) once after activation.

See Agent skills for the full catalogue, workflow chains, and authoring guide.

Next step

Open a Teradata context and provision the feature store catalogs in Connect and configure.