Connecting Hub to Databricks

Hub stores all data in a single PostgreSQL database. You can analyze it in Databricks by connecting Databricks directly to Hub’s Postgres - either through Lakehouse Federation (zero-ETL, queried live) or with a JDBC read from a notebook (load once into a Delta table).

Summary

Step	Action
1	Run Hub, worker, and Postgres (Quick Start); make sure Databricks can reach the Postgres host (Private Link / VPC peering / public endpoint with sslmode=require)
2	Choose a path: Lakehouse Federation (UI, zero-ETL, queries Hub live through Unity Catalog) or JDBC + Delta (notebook code, materialized copy)
3	Federation: create a CONNECTION and foreign CATALOG. JDBC: read with spark.read.format("jdbc") and df.write.format("delta").saveAsTable(...)
4	Run analytics queries using the same dimensions and metrics as in the Quick reference below

Why Hub + Databricks?

• Direct access - Databricks connects to Hub’s Postgres database; no separate ETL pipeline required.
• Two integration paths - Use Lakehouse Federation for live, governed queries through Unity Catalog, or use JDBC + Delta for a materialized lakehouse copy.
• Same data model - Hub’s Data Model uses one row per response in a single table; ready for Spark SQL aggregations.

Step 1: Get Hub running

If Formbricks Hub is not running yet, follow the Quick Start Guide to start Postgres, apply migrations automatically, and start the Hub API and worker.

Note the database connection details you use for Hub (host, port, database name, user, password). You will need them for Databricks. If you use the compose file from the docs, that is typically:

Setting	Example value
Host	A hostname or IP that Databricks can reach (for example, your Postgres public endpoint or a Private Link target)
Port	5432
Database name	hub
Username	formbricks
Password	Your Postgres password (in the Quick Start compose setup this is POSTGRES_PASSWORD in .env; default is formbricks_dev)

Note

Databricks workspaces run in a cloud VPC/VNet, so they cannot reach a Hub Postgres on localhost. For production, prefer AWS PrivateLink, Azure Private Endpoint, or VPC peering; for a quick test, expose Postgres on a public endpoint with sslmode=require and allowlist the Databricks workspace NAT IPs.

Step 2: Choose a connection path

Databricks ships the PostgreSQL JDBC driver (org.postgresql.Driver) on every current Databricks Runtime, so no extra library install is needed. You have two practical ways to query feedback_records:

• Lakehouse Federation (recommended when you have Unity Catalog) - Create a CONNECTION and a foreign CATALOG once, then query Hub live from any notebook, SQL editor, or dashboard. Filters, projections, limits, and aggregates are pushed down to Postgres on DBR 13.3 LTS+ (join pushdown is a separate Public Preview feature that requires DBR 17.2+ or SQL warehouses 2025.30+).
• JDBC + Delta - From a notebook, read feedback_records via JDBC into a Spark DataFrame and optionally write it to a Delta table for repeat queries.

Pick Federation for governed, low-latency live access; pick JDBC + Delta when you want a local materialized copy or are on a workspace without Unity Catalog. Step 3 covers both paths - follow only the one you chose.

Step 3a: Lakehouse Federation (Unity Catalog)

Requires Databricks Runtime 13.3 LTS or later in Standard/Dedicated mode, or a Pro/Serverless SQL warehouse, plus CREATE CONNECTION and CREATE CATALOG privileges on the metastore.

Create the connection

1. In your Databricks workspace, open Catalog Explorer, then click + Add -> Add a connection.
2. Choose PostgreSQL as the connection type.
3. Enter the values from the Step 1 table:
   • Host - Hub’s Postgres host
   • Port - 5432
   • User - formbricks
   • Password - Hub’s Postgres password (Databricks will offer to store it as a Secret)
4. Click Test connection, then Create.

Or use SQL:

CREATE CONNECTION hub_pg
  TYPE postgresql
  OPTIONS (
    host '<hub-postgres-host>',
    port '5432',
    user '<user>',
    password secret('<scope>', '<key>')
  );

Create the foreign catalog

1. Still in Catalog Explorer, click + Add -> Add a catalog.
2. Choose Foreign catalog, name it (for example, hub), select the connection hub_pg, and set the Database to hub.
3. Click Create.

Or use SQL:

CREATE FOREIGN CATALOG hub
  USING CONNECTION hub_pg
  OPTIONS (database 'hub');

You can now query feedback_records live from anywhere in Databricks:

SELECT * FROM hub.public.feedback_records LIMIT 100;

For the full table schema - column names, types, and field semantics (for example, field_type, value_number, value_text) - see the Data Model doc.

Note

Federation does not copy data. Each query is rewritten to a pushdown-aware PostgreSQL query, so make sure Hub’s Postgres has enough capacity for your analytics workload, or schedule materialized copies with the JDBC pattern below.

Step 3b: JDBC + Delta (notebook)

If you do not have Unity Catalog or you want a materialized copy of feedback_records in Delta, read it from a notebook.

Store Postgres credentials as a secret

databricks secrets create-scope hub
databricks secrets put-secret hub pg_user --string-value formbricks
databricks secrets put-secret hub pg_password --string-value <your-password>

Read feedback_records via JDBC

jdbc_url = "jdbc:postgresql://<hub-postgres-host>:5432/hub?sslmode=require"

reader = (
    spark.read.format("jdbc")
    .option("url", jdbc_url)
    .option("driver", "org.postgresql.Driver")
    .option("user", dbutils.secrets.get("hub", "pg_user"))
    .option("password", dbutils.secrets.get("hub", "pg_password"))
)

# Whole table
df = reader.option("dbtable", "public.feedback_records").load()

# Or push a query down (faster for incremental loads)
df_recent = (
    reader
    .option(
        "query",
        """
        SELECT id, tenant_id, submission_id, source_type, source_id,
               field_id, field_type, value_text, value_number,
               value_boolean, value_date, metadata, collected_at,
               created_at, updated_at
        FROM public.feedback_records
        WHERE updated_at > now() - interval '1 day'
        """,
    )
    .load()
)

Note

Databricks Runtime 11.3 LTS and above also ship a named PostgreSQL connector (spark.read.format("postgresql")) that accepts separate host, port, and database options instead of a JDBC url. Either form works; this guide uses format("jdbc") so the same code runs on older runtimes too.

Note

dbtable and query are mutually exclusive. Use dbtable when you also want parallel reads with partitionColumn, lowerBound, upperBound, and numPartitions. Use query to push filters and projections down to Postgres.

Write to a Delta table

(
    df.write
    .mode("overwrite")  # or .mode("append") for incremental loads
    .format("delta")
    .saveAsTable("main.hub.feedback_records")
)

You can now query the Delta table from Databricks SQL or any notebook:

SELECT source_type, count(*) FROM main.hub.feedback_records GROUP BY source_type;

Step 4: Example queries

Whether you query the foreign catalog (hub.public.feedback_records) or a Delta table (main.hub.feedback_records), the analytics are identical.

NPS trend over time

Average NPS per week, filtered to NPS questions:

SELECT
  date_trunc('week', collected_at) AS week,
  ROUND(AVG(value_number), 2)      AS avg_nps,
  COUNT(*)                         AS responses
FROM hub.public.feedback_records
WHERE field_type = 'nps' AND value_number IS NOT NULL
GROUP BY 1
ORDER BY 1;

Feedback count by source

SELECT source_type, COUNT(*) AS count
FROM hub.public.feedback_records
GROUP BY source_type
ORDER BY count DESC;

NPS promoters / passives / detractors

SELECT
  CASE
    WHEN value_number <= 6 THEN 'Detractor (0-6)'
    WHEN value_number <= 8 THEN 'Passive (7-8)'
    ELSE                        'Promoter (9-10)'
  END                AS nps_bucket,
  COUNT(*)           AS count
FROM hub.public.feedback_records
WHERE field_type = 'nps' AND value_number IS NOT NULL
GROUP BY 1
ORDER BY MIN(value_number);

Quick reference: dimensions and metrics

For the complete list of columns and field types, see the Data Model.

Use case	Dimension / filter	Metric
NPS/CSAT over time	collected_at; filter field_type = nps	AVG(value_number) or custom NPS %
Count by source	source_type	COUNT(*)
Score by question	field_id or field_label	AVG(value_number)
Categorical breakdown	value_text (for example, device, segment)	COUNT(*)
Multi-tenant	Filter tenant_id	any

Once Databricks can reach Hub’s Postgres, your feedback data is ready for SQL, notebooks, and dashboards alongside everything else in your lakehouse.

References

Hub

• Quick Start - Run Hub, worker, and Postgres
• Data Model - Table schema, columns, and field types for feedback_records
• Connecting Hub to Power BI - Same data, with Microsoft Power BI
• Connecting Hub to Superset - Same data, with Apache Superset
• Connecting Hub to Airbyte - Schedule periodic syncs into Snowflake, Databricks, BigQuery, or Redshift

Databricks

• Query PostgreSQL with Databricks - JDBC reads and writes
• Run federated queries on PostgreSQL - Lakehouse Federation setup
• What is Lakehouse Federation? - Overview
• Secrets management - Store database credentials safely