Configuration

Full reference for agent configuration — every setting, its type, and its effect on agent behavior.

Overview

Configuration file

The agent is configured through a YAML configuration file provided during setup. Your bootstrap package includes a pre-filled template with defaults — update the credential references and environment settings to match your setup.

Unknown keys are rejected. The config validator rejects any key not in the approved schema. If you add an unrecognized field, the agent will refuse to start with a clear error identifying the offending key.

database

The database block

database.vendorstringrequired

Database engine. Currently only "snowflake" is supported. Additional vendors are planned.

database.auth.methodstringrequired

Authentication method. Currently only "keypair" (RSA private key authentication) is supported.

database.auth.cache_ttl_secinteger

How long (in seconds) to cache resolved credentials in memory. Default: 43200 (12 hours). The agent re-resolves credentials from the source on expiry.

database.db_users.readerobjectrequired

Credentials for the read-only database user. Used by Discovery, Profile, and Certifier tasks.

database.db_users.pipelineobjectrequired

Credentials for the DML database user. Used by Operate tasks to execute certified INSERT/UPDATE/DELETE/MERGE statements.

database.db_users.ddl_adminobjectrequired

Credentials for the DDL database user. Used by DDL Deploy tasks in development environments to execute CREATE/ALTER/DROP statements.

db_users.*.usernamestringrequired

Database username for this user type. Must match the user object in your database exactly.

db_users.*.secret_refstringrequired

Credential source reference. Points to your secret manager or a local credential file. See the Credentials guide for supported formats and setup instructions.

Settings

Operational settings

All agent behavior is configured through operational settings. These fields are validated and merged with platform defaults.

environmentstringrequired

Target environment. One of: "dev", "lab", "int", "prd". Controls the execution path, logging format, and credential restrictions.

db_vendorstringrequired

Database vendor. Must match database.vendor. Currently only "snowflake".

warehousestringrequired

Database warehouse used for all agent-initiated queries. The warehouse must exist and be accessible by all three database users.

scratch_databasestring

Database used for profiling scratch tables.

scratch_schemastring

Schema within scratch_database for profiling work.

scratch_schema_prefixstring

Prefix for profiling schemas created within scratch_database.

max_scratch_rowsstring

Maximum rows per profiling temp table. Accepts K, M suffixes (e.g. 100M). Default: 100M.

max_scratch_bytesstring

Maximum bytes per profiling temp table. Accepts KB, MB, GB suffixes (e.g. 5GB). Default: 5GB.

log_levelstring

Minimum log level. One of: "DEBUG", "INFO", "WARNING", "ERROR". Default: "INFO".

log_formatstring

"text" (human-readable, default for dev/lab) or "json" (structured, required for production). JSON format is suitable for CloudWatch, Splunk, Datadog.

Environments

Environment values and their behavior

"dev"PECO/D

•PECO/D path: Plan, Enrich, Certify, DDL Deploy, re-Certify, Operate
•DDL Deploy enabled
•File-based credentials allowed
•Text log format default

"lab"PECO/D

•PECO/D path: DDL Deploy enabled for fast iteration
•Lab sandbox — isolated from dev and int
•File-based credentials allowed

"int"PECO

•PECO path: no DDL Deploy
•Schema changes via promotion only
•JSON log format recommended

"prd"PECO

•PECO path: Plan, Enrich, Certify, Operate
•File-based credentials blocked — secret manager required
•JSON logging enforced

Need help?

Email us at support@boltpipeline.ai