documentation_chunks.parquet•790 kB
PAR1���ʆ]L� ��qh� ---
sidebar_position: 3�Llabel: "Feature Development Workflow"
---
import Tabs from '@theme/Tabs'; im Item ># ';�5 # nn 4
In Tecton, f �s are d �hed and tested in a notebook $hen
produc�iz1(s code with2 a` re%ory (]op;<ally, a
GitOps w-P to enable an integra�CI/CD'�).
This gives the benefit of fast iteration spee:� T, while preserving
Dev�Hbest practices for .� R like vers_control,) reviews,
��,.
A typical1hE�e$building a59B!�it%�
train4data set looks �<this:
1. Create!�ewPdefinE�L5�
2. Run%FM pipeline%�%vely)� sA�correct1�$
3. Fetch �0of registered*M� a�spac!�d cJ�
�
4. GeneA 6to%"�.� E�0model
5. CopyB# 2to your�E�
6. Apply changAlo a livA=q�viakmCLI or a�E5Z0
:::note
Ifc do!� nA�.� 1�� ,c n5Dwould skip steps 3Ez4 above.fm,paga�8ll walk throughKse
=$in detail.�have�\already done so, installA
�
o�r loA�machAHA� i
��Senvirona. Alsa�E^that�r
,$has access% h!�4levant computeeH%DZ Va�.
## . �1W2 a3��
Any1�objA�can b�.fin>!. We c- s
%eos "s
Fs". SimAZwrite!Ab� cel!�$```python
�� t�� �8�Entity, BatchSource, FileConfig, batch_
� _��(, Attribute.R .typesX$Field, Str�"ydatetime#0delta
user =
�(name="4", join_keys=[T (_id"^)]) B(_sign_ups =.� (aB U'", � c� =� 7$uri="s3://)<�.ai.public/tutorials/fraud_demo/customers/data.pq
iLfile_format="parquet!stamp_f!A ="signup_
,),
)
@ � f�%� � s%�s=[!F� ] � e!� i
���nda6 schedule=�%�(days=1) N��
�s=[%vUHH("credit_card_issue!�I>
h ]�def �F0 (!0A):f2 [Nf ] =B&