Training¶
A guided tour through tdfs4ds. Follow the thirteen modules in order to go
from a fresh install to running the full stack in production. Each module is
short — most are a single screen of prose plus links into the reference docs.
Audience
These modules assume you have completed Getting started › Installation and have an active Teradata Vantage connection. If not, start there first.
The journey¶
| # | Module | What you'll learn |
|---|---|---|
| 1 | First feature | Ingest your first feature end-to-end |
| 2 | Documentation | Let an LLM write the business dictionary |
| 3 | Query optimization | Auto-rewrite slow process views |
| 4 | Consumer agent | Chat with your feature store |
| 5 | Vector store backends | Chroma vs Teradata — pick where the chatbot's memory lives |
| 6 | FilterManager | Segmented ingestion for large or partitioned sources |
| 7 | TimeManager | Replay history at business time |
| 8 | Partitioning | Why and how to partition feature tables |
| 9 | NULLs in entities | The entity_null_substitute story |
| 10 | Catalogs and datasets | The three catalogs and why datasets matter |
| 11 | Logs | The follow-up table and how to recover from failures |
| 12 | Lineage | Dependency graphs and Sankey visualisations |
| 13 | Migration manifest | Replay a feature store in another environment |
How to read it¶
Modules 1–4 are the day-one path: ingest, document, optimize, chat. Module 5 is the architectural decision that unlocks production deployments. Modules 6–9 are the operational toolbox you reach for when ingestion has to scale or recover. Modules 10–13 cover governance and lifecycle — catalogs, audit logs, lineage, and cross-environment replay.
Every module ends with a Learn more → block linking to the authoritative reference pages, which is where the deep parameter tables live. The training section's job is to give you the shape of the picture and the order of the moves; the reference docs answer "what does this flag do?".
After training¶
Once you have walked through the thirteen modules, the Concepts section is the right next stop for the deeper architectural model (temporal data, hidden-tables pattern, data domains).