New · SOFI private deployment is ready for enterprise rolloutTalk to us
[ CONNECT ][ 10+ ENGINES ][ FILES ][ ENCRYPTED ]
endpoint · /connect

Connect everything,in minutes

Plug Postgres, Oracle, MongoDB, ClickHouse and ten other engines — plus CSV, JSON and Parquet files — under one governed query layer. Encrypted credentials, pooled connections, schema sync out of the box.

https://private.sofi.local/datasources
10/10 healthy

// connectors · 10 engines

Postgres
encrypted
MySQL
encrypted
SQL Server
encrypted
Oracle
encrypted
MongoDB
encrypted
ClickHouse
encrypted
Cassandra
encrypted
Neo4j
encrypted
Elastic
encrypted
S3 Files
encrypted

// datasource · oracle.billing

engineoracle 19c
hostoracle.acme.local:1521
serviceBILLING
authuser · ORACLE_PWD env
pool2 → 16 connections

// scan

tables

284

rows

62.1M

scanned

8.2 s

freshness

live

[ 01 / 06 ]What you get
// Capabilities //

One client for every source

Databases, files and clouds — all queried, governed and audited the same way.

10+ database engines

Postgres, MySQL, SQL Server, Oracle, MongoDB, ClickHouse, Cassandra, Neo4j, Elastic — and more shipping monthly.

Files first-class

CSV, JSON, Parquet and Excel — local, S3, GCS or Azure. Same query model, same governance.

Encrypted credentials

AES-256-GCM at rest, key rotation supported. Credentials never appear in logs, configs or query plans.

[ 02 / 06 ]How it works
// Flow //

From credentials to a searchable catalog

Four steps — and the source is queryable, governed and lineage-aware.

step · 01

Discover

engine · host · port · sample probe

step · 02

Authenticate

user · key · iam · oauth

step · 03

Sample

schemas · tables · row counts

step · 04

Catalog

indexed · classified · searchable

[ 03 / 06 ]Developer surface
// Register a source //

Same shape for every engine

Datasources are declarative — code, YAML or REST. Add one, scan, query.

# pip install sofi
from sofi import Sofi

sofi = Sofi(api_key="YOUR_KEY")

ds = sofi.datasource.create(
    name="oracle.billing",
    engine="oracle",
    host="oracle.acme.local",
    port=1521,
    service="BILLING",
    auth={"type": "user", "user": "sofi", "password_env": "ORACLE_PWD"},
    pool={"min": 2, "max": 16},
)

# scan and sample
ds.scan(sample=True)
print(ds.tables[:5])
[ 04 / 06 ]Use cases
// What teams build //

Every stack, federated

Hybrid clouds, legacy systems, file lakes, migrations — all under one connector layer.

// pattern

Hybrid stack

Operational Postgres, analytical ClickHouse, document MongoDB, legacy Oracle — federate them under one query layer without picking a winner.

10+engines side-by-side

// pattern

Legacy integration

Plug Oracle, SQL Server and AS/400 into modern apps without ETL pipelines. Read-only by default, write-through where needed.

0ETL pipelines

// pattern

File analytics

Query CSV, JSON and Parquet on S3 like any other table. Schema is inferred, refreshable and cataloged.

<1 minto first query

// pattern

DB migration cutover

Run old and new engines side-by-side under one virtual schema. Rebalance traffic gradually, roll back instantly.

0downtime cutover
[ 05 / 06 ]Performance
// Numbers //

Connectors that behave in production

Pooled, encrypted, monitored — and they ship monthly.

10+

engines supported

Postgres, MySQL, SQL Server, Oracle, MongoDB, ClickHouse, Cassandra, Neo4j, Elastic — file formats included.

AES-256

credential encryption

GCM mode, per-tenant keys, rotation supported. Secrets stay encrypted at rest and in transit.

Pool

managed connections

Connection pools per datasource with health checks, retry budgets and graceful drain on rotation.

Live

schema sync

Schema changes propagate to the catalog within seconds. Views surface deprecation warnings before they break.

[ 06 / 06 ]FAQ
// FAQ //

Questions about connect

What infra teams ask before adopting Connect as the integration layer.

Postgres, MySQL, SQL Server, Oracle, MongoDB, ClickHouse, Cassandra, Neo4j and Elasticsearch. File formats: CSV, JSON, Parquet, Excel. New connectors ship monthly — Snowflake, BigQuery and Databricks are next.

// ready to connect

Plug your sources, query in minutes.

Add a datasource via SDK or YAML, run the scanner, watch the catalog populate. Trial includes all 10 engines.