Get started with PowerCenter conversion

Scan PowerCenter XML exports to assess automation potential before committing to a conversion.

[analyze]
path         = "./exports/"
output       = "./analysis.pdf"
summary_only = false
no_recursion = false

Submit analysis_report.pdf to benza@cepheus.in. The PDF embeds cryptographic fingerprints of your exports. Cepheus replies with a license file.

[verify_license]
path    = "./exports/"
license = "./license.pdf"

VALID      repo_export.xml      (sha256: a1b2c3...)
VALID      archive_export.xml   (sha256: f6e5d4...)

MISMATCH = file modified since analysis (re-run analyze). UNLICENSED = not in this license.

[convert]
path       = "./exports/"
license    = "./license.pdf"
output     = "./output/"
target     = "dbt"          # dbt | pyspark | snowpark | glue | adf | dataform
dialect    = "snowflake"    # snowflake | bigquery | redshift | databricks | synapse | generic
orchestrator = "airflow"    # python | airflow | dagster | prefect (dbt only)

Target platforms

Platform-specific settings

[convert]
target         = "pyspark"
workspace_path = "/Workspace/benza"

[convert]
target    = "snowpark"
warehouse = "COMPUTE_WH"

[convert]
target        = "glue"
glue_database = "benza_db"

Custom translation rules

[convert]
rules_dir = "./my_custom_rules"

See Custom Rules for how to write rule classes.

Structure varies by target. All produce a reports/ directory with signed governance artifacts.

dbt

output/
├── dbt_project.yml
├── models/           <model>.sql + sources.yml + .yml docs
├── macros/           <mapplet>.sql
├── snapshots/        SCD Type 2 snapshot configs
├── orchestration/    Airflow / Dagster / Prefect / Python
├── validation/       Reconciliation queries
└── reports/          audit_trail.json + .pdf | equivalence + .pdf | risk_scores.json

PySpark

output/
├── jobs/             PySpark job scripts
├── notebooks/        Databricks notebook cells
├── orchestration/    Databricks Workflow JSON
└── reports/

Snowpark

output/
├── procedures/       Snowpark stored procedures
├── orchestration/    Snowflake Task DAG (CREATE TASK)
└── reports/

Glue

output/
├── jobs/             Glue PySpark + GlueContext
├── orchestration/    Step Functions | Glue Workflow | Airflow
└── reports/

ADF

output/
├── dataflows/        Mapping Data Flow JSON
├── pipelines/        ADF Pipeline JSON
├── datasets/ + linkedServices/ + arm_template.json
└── reports/

Dataform

output/
├── definitions/      SQLX models
├── includes/         JavaScript macros
├── workflow_settings.yaml
└── reports/

Confidence levels

Local Flask dashboard for reviewing conversion results.

[serve]
output_dir = "./output/"
port       = 8080
host       = "127.0.0.1"

From the Workbench, each mapping can be Approved or Flagged (with notes). Decisions are recorded in reports/approvals.jsonl — append-only, immutable, timestamped. Part of the governance evidence trail.

Optional. Runs validation queries against source and target databases, compares results.

[reconcile]
output_dir    = "./output/"
source_conn   = "postgresql://user:pass@source-host:5432/db"
target_conn   = "snowflake://user:pass@account/db/schema?warehouse=wh"
source_schema = "operational"
target_schema = "analytics"
tolerance     = 0.0
timeout       = 120

Database drivers

Result statuses

Produced on every conversion run. Cannot be disabled.

Audit trail commands

[audit_report]
output_dir = "./output/"
export     = ""               # leave blank for all, or set to a mapping name
printable  = false

All share the same translation engine, expression grammar, and SCD detection. Governance artifacts are identical across targets.

dbt target

Other targets — built-in

from cepheus_benza.core.rules.base import TranslationRule

class MyLookupRule(TranslationRule):
    priority = 50

    def applies_to(self, node) -> bool:
        return node.transformation_type == "Lookup" and "CUSTOM_" in node.name

    def translate(self, node, context):
        return node_output

Place rule files in a directory and set rules_dir in [convert]. Higher priority takes precedence over built-in rules.

Override spec.toml values. Useful for credentials you don't want in config files.

"No licensed exports found"

Checksums don't match. File was modified after analysis or wrong license. Re-run analyze and request an updated license.

"LicenseNotFoundError"

Check the license path in spec.toml. Ensure the file exists and is readable.

Parse errors

Not a valid PowerCenter export. Verify: exported from Repository Manager, UTF-8 or Latin-1 encoding, file not truncated.

Missing output models

Some types produce no output without a connected Target. Check audit trail for stub confidence or disconnected node warnings.

Reconciliation: sqlalchemy not installed

pip install sqlalchemy <driver-for-your-warehouse>

Reconciliation: connection errors

• Verify connection string format matches your driver
• Check network access to both databases
• Use timeout in spec.toml to increase per-query limit

Workbench shows no models

Verify output_dir in [serve] matches output in [convert], and reports/audit_trail.json exists.