Connect a warehouse
The context tells Cassis what your business means; the warehouse supplies the current data. Connect them so Cassis can turn a business question into SQL, run it, and return an answer with real results.
Why Cassis connects to the warehouse
When someone asks a question, Cassis uses the context to choose the right definitions, tables, joins, and metrics. It then sends the generated SQL to your warehouse and turns the returned rows into an answer. Your warehouse remains the source of the data; Cassis is the context and query layer in front of it.
- Understand the structure
- Cassis reads schema metadata such as tables and columns.
- Answer from current data
- Cassis runs the SQL it generates, so the answer includes result rows rather than SQL alone.
- Stay aligned
- Cassis detects schema changes that may require a context update.
Use a live connection
When you create the project, or later under Settings → Project settings → Connection, paste the connection string and use Test. Cassis supports PostgreSQL, Snowflake, and BigQuery, detects the SQL dialect from the connection, and stores the credentials encrypted.
Cassis connects only to a warehouse with a public address, and only with the forms below. Any other string is refused with “Connection to this host is not allowed” when you save it and whenever Cassis connects.
- PostgreSQL
postgresql://user:password@host:5432/database: one host, and a user. Options:sslmode,sslnegotiation,ssl_min_protocol_version,ssl_max_protocol_version,target_session_attrs,application_name,search_path.sslmode=verify-caandverify-fullcheck the server certificate against publicly trusted certificate authorities.- Snowflake
- Your account identifier or
*.snowflakecomputing.comhost, port 443 if any. Sign in with a password, a key pair (authenticator=SNOWFLAKE_JWTwithprivate_key_b64), a programmatic access token, or OAuth (authenticator=oauthwithtoken). Options:warehouse,role,database,schema,application,client_session_keep_alive,timezone,cache_column_metadata,case_sensitive_identifiers,enable_decfloat. Cassis refusesPUTandGETin any SQL it runs, gold SQL included. - BigQuery
- A Google service-account key in
credentials_base64. Options:location,arraysize,list_tables_page_size,maximum_bytes_billed,priority,use_query_cache. - Special characters
- Percent-encode
/,?,#, and@in a password. An@in an email-style username is encoded for you.
When the warehouse schema changes, use Sync from warehouse on the Schema page. Cassis computes a plan, the schema diff and every context change it implies, for you to review and apply. Nothing is rewritten silently; see Review schema updates.
When Cassis must not reach the warehouse
Upload the schema as a DDL file instead. Cassis then generates grounded SQL without running it, and your own tools can run it. See Run without a warehouse connection.
Check the connection
Open the Schema page and confirm the expected tables and columns appear. Once a first context is published, ask a question: the answer should include rows and the SQL that produced them.
Give the connection’s user read-only access to the schemas Cassis should answer from, and nothing more.