🚧 THIS IS A WORK IN PROGRESS 🚧

Unified RAG dbt Package (Docs)

What does this dbt package do?

The main focus of this dbt package is to generate an end model and Cortex Search Service (for Snowflake destinations only) which contains the below relevant unstructured document data to be used for Retrieval Augmented Generation (RAG) applications leveraging Large Language Models (LLMs):

• HubSpot: Deals
• Jira: Issues
• Zendesk: Tickets

The following table provides a detailed list of all models materialized within this package by default.

TIP: See more details about these models in the package's dbt docs site.

Table	Description
rag__unified_document	Each record represents a chunk of text prepared for vectorization and additional fields for use in LLM workflows.

Additionally, for Snowflake destinations, a Cortex Search Service will be generated as a result of this data model. The Cortex Search Service uses the results of the rag__unified_document and enables Snowflake users to take advantage of low-latency, high quality "fuzzy" search over their data for use in RAG applications leveraging LLMs. See the below table for details.

Snowflake Cortex Search Service	Description
rag__unified_embedding	Generates a Snowflake Cortex Search service via the embedding_generation macro as a post-hook for Snowflake destinations. This Cortex Search Service is currently configured with a target lag of 1 day. Please be aware that this search service will refresh automatically once a day.

How do I use the dbt package?

Step 1: Prerequisites

To use this dbt package, you must have the following:

• At least one of the below support Fivetran connectors syncing data into your destination.
  • HubSpot (specifically deals)
  • Jira
  • Zendesk Support
• A Snowflake, BigQuery, Redshift, Databricks, or PostgreSQL destination.
  • Please note, the Cortex Search Service will only be generated for Snowflake destinations.

Step 2: Install the package

Include the following package_display_name package version in your packages.yml file:

TIP: Check dbt Hub for the latest installation instructions or read the dbt docs for more information on installing packages.

packages:
  - package: fivetran/unified_rag
    version: [">=0.1.0", "<0.2.0"] # we recommend using ranges to capture non-breaking changes automatically

Step 3: Define database and schema variables

Single connector

By default, this package looks for your ad platform data in your target database. If this is not where your app platform data is stored, add the relevant <connector>_database variables to your dbt_project.yml file (see below).

# dbt_project.yml

vars:
    rag_hubspot_schema: hubspot
    rag_hubspot_database: your_database_name

    rag_jira_schema: jira
    rag_jira_database: your_database_name

    rag_zendesk_schema: zendesk
    rag_zendesk_database: your_database_name

Union multiple connectors

If you have multiple supported connectors in Fivetran and would like to use this package on all of them simultaneously, we have provided functionality to do so. The package will union all of the data together and pass the unioned table into the transformations. You will be able to see which source it came from in the source_relation column of each model. To use this functionality, you will need to set either the <package_name>_union_schemas OR <package_name>_union_databases variables (cannot do both) in your root dbt_project.yml file. Below are the variables and examples for each connector:

# dbt_project.yml

vars:
    rag_hubspot_union_schemas: ['hubspot_rag_test_one', 'hubspot_rag_test_two']
    rag_hubspot_union_databases: ['hubspot_rag_test_one', 'hubspot_rag_test_two']

    rag_jira_union_schemas: ['jira_rag_test_one', 'jira_rag_test_two']
    rag_jira_union_databases: ['jira_rag_test_one', 'jira_rag_test_two']

    rag_zendesk_union_schemas: ['zendesk_rag_test_one', 'zendesk_rag_test_two']
    rag_zendesk_union_databases: ['zendesk_rag_test_one', 'zendesk_rag_test_two']

The native source.yml connection set up in the package will not function when the union schema/database feature is utilized. Although the data will be correctly combined, you will not observe the sources linked to the package models in the Directed Acyclic Graph (DAG). This happens because the package includes only one defined source.yml.

To connect your multiple schema/database sources to the package models, follow the steps outlined in the Union Data Defined Sources Configuration section of the Fivetran Utils documentation for the union_data macro. This will ensure a proper configuration and correct visualization of connections in the DAG.

Step 4: Enabling/Disabling Models

This package takes into consideration that not every account will have leverage every supported connector type. If you do not leverage all of the supported connector types, you are able to disable the respective dependent models using the below variables in your dbt_project.yml.

vars:
    rag__using_hubspot: False # by default this is assumed to be True
    rag__using_jira: False # by default this is assumed to be True
    rag__using_zendesk: False # by default this is assumed to be True

(Optional) Step 5: Additional configurations

Changing the Build Schema

By default this package will build the Unified RAG staging models within a schema titled (<target_schema> + _stg_unified_rag) and the Unified RAG final models within a schema titled (<target_schema> + unified_rag) in your target database. If this is not where you want your modeled Unified RAG data to be written to, add the following configuration to your dbt_project.yml file:

models:
    unified_rag:
        +schema: my_new_schema_name # leave blank for just the target_schema
        staging:
            +schema: my_new_schema_name # leave blank for just the target_schema

Change the source table references

If an individual source table has a different name than the package expects, add the table name as it appears in your destination to the respective variable:

IMPORTANT: See this project's dbt_project.yml variable declarations to see the expected names.

# dbt_project.yml

vars:
    rag_<default_source_table_name>_identifier: your_table_name 

(Optional) Step 6: Orchestrate your models with Fivetran Transformations for dbt Core™

Expand for details

Fivetran offers the ability for you to orchestrate your dbt project through Fivetran Transformations for dbt Core™. Learn how to set up your project for orchestration through Fivetran in our Transformations for dbt Core setup guides.

Does this package have dependencies?

This dbt package is dependent on the following dbt packages. These dependencies are installed by default within this package. For more information on the following packages, refer to the dbt hub site.

IMPORTANT: If you have any of these dependent packages in your own packages.yml file, we highly recommend that you remove them from your root packages.yml to avoid package version conflicts.

packages:
    - package: fivetran/fivetran_utils
      version: [">=0.4.0", "<0.5.0"]

    - package: dbt-labs/dbt_utils
      version: [">=1.0.0", "<2.0.0"]

How is this package maintained and can I contribute?

Package Maintenance

The Fivetran team maintaining this package only maintains the latest version of the package. We highly recommend you stay consistent with the latest version of the package and refer to the CHANGELOG and release notes for more information on changes across versions.

Contributions

A small team of analytics engineers at Fivetran develops these dbt packages. However, the packages are made better by community contributions.

We highly encourage and welcome contributions to this package. Check out this dbt Discourse article on the best workflow for contributing to a package.

Are there any resources available?

• If you have questions or want to reach out for help, refer to the GitHub Issue section to find the right avenue of support for you.
• If you want to provide feedback to the dbt package team at Fivetran or want to request a new dbt package, fill out our Feedback Form.
• Have questions or want to be part of the community discourse? Create a post in the Fivetran community and our team along with the community can join in on the discussion.