Datasource Onboarding: 33 Engines in One Playbook

AskTable connects 33 classes of sources—relational databases, warehouses, NoSQL, files, SaaS tables, and streaming endpoints. This guide explains the happy path and the operational checklist behind a healthy connection.

---

1. What you can connect

• •Relational: MySQL/MariaDB/TiDB, PostgreSQL/CockroachDB/Greenplum, Oracle, SQL Server, SQLite, DB2
• •Warehouses: Snowflake, BigQuery, Redshift, Synapse, ClickHouse, Doris, StarRocks, Databend, …
• •NoSQL / specialized: MongoDB, InfluxDB, TimescaleDB, …
• •Files: Excel, CSV, JSON, Parquet
• •Online sheets: Feishu Bitable, Google Sheets
• •APIs / streams: REST, GraphQL, Kafka, Flink (where product edition supports)

Exact engine list evolves—check the in-product picker for the live matrix.

---

2. Standard onboarding flow

Steps

1. •Choose engine in Datasources → Add.
2. •Connection: host, port, database, user, password, SSL mode, and any engine-specific fields.
3. •Test—confirm latency and error messages look sane.
4. •Sync metadata—select schemas/tables you want the model to see (skip temp/test tables when possible).
5. •Policies—row filters, column masks, roles.
6. •Validate—run a few natural-language questions and confirm row-level security.

---

3. Configuration sketches

MySQL

{
  "host": "mysql.example.com",
  "port": 3306,
  "database": "mydb",
  "username": "readonly_user",
  "password": "********"
}

Prefer read-only credentials, TLS, explicit timeouts, and pool sizing on the AskTable side.

CREATE USER 'readonly_user'@'%' IDENTIFIED BY 'password';
GRANT SELECT ON mydb.* TO 'readonly_user'@'%';
FLUSH PRIVILEGES;

PostgreSQL

{
  "host": "postgres.example.com",
  "port": 5432,
  "database": "mydb",
  "username": "readonly_user",
  "password": "********",
  "schema": "public"
}

Grant CONNECT, USAGE, and SELECT, and set default privileges for future tables if analysts add objects often.

ClickHouse

{
  "host": "clickhouse.example.com",
  "port": 8123,
  "database": "default",
  "username": "readonly_user",
  "password": "********",
  "protocol": "http"
}

Tune max_execution_time, compression, and TLS to match your cluster.

Excel upload

1. •Choose Excel as the engine.
2. •Upload .xlsx / .xls / .csv.
3. •Pick sheet, preview types, import.

Typical limits: ~100 MB files and 100k rows—see product docs for current caps.

Feishu Bitable

{
  "app_id": "cli_xxxxx",
  "app_secret": "********",
  "app_token": "bascnxxxxx"
}

Create a tenant app on the Feishu Open Platform, enable Bitable scopes, and paste the app token for the base you need.

---

4. Networking patterns

Always combine least privilege DB users, IP allowlists, and TLS.

---

5. Metadata hygiene

• •Sync only tables analysts actually need—smaller catalogs mean faster linking and cheaper prompts.
• •Maintain table and column comments—they feed the semantic layer and SQL quality.
• •Use regex include/exclude rules for large estates (^(fact|dim)_ vs .*_staging).

---

6. Performance and reliability

• •Configure pool sizes, statement timeouts, and max rows per query.
• •Add indexes on frequent join/filter columns in the source DB—not in AskTable.
• •Use metadata caching and optional result caching where product settings allow.

---

7. When things break

---

8. Security checklist

• •Read-only accounts for analytics
• •No secrets in front-end bundles
• •Audit datasource changes and high-cost queries
• •Rotate credentials on a schedule

---

Next steps

• •Connect your first datasource
• •Permission model
• •MCP deployment

Site: asktable.com