Power BI Scanner AI Parser¶
Alation Cloud Service Applies to Alation Cloud Service instances of Alation
Alation uses Generative AI via the Amazon Bedrock integration to overcome the challenge of extracting the connection information from Power BI’s M query expressions. This feature helps surface lineage and metadata from complex M queries that cannot be parsed, which results in more accurate and detailed data lineage for Alation users who use the Power BI Scanner OCF connector. This feature is available from Alation version 2025.1.2 and Power BI Scanner connector version 2.9.0.
Important
This feature is only available on Alation Cloud Service instances.
How It Works¶
Alation integrates with Anthropic’s Claude model hosted on Amazon Bedrock to extract the connection information from M query for Power BI. The AI deconstructs complex M queries, identifying and extracting key metadata elements like source connections (host, port, database), object names (schema and table names), and specifics of query transformations. This granular extraction builds accurate lineage charts that are available within Alation’s Lineage Graph.
Enable Power BI AI Parser¶
Important
The Power BI AI parser feature isn’t available by default and must be enabled by a source admin when configuring the BI source settings.
To enable this feature:
On the Settings page of BI Server source, go to the Metadata Extraction tab.
Under the AI-based expression parsing section, turn on Enable extraction of metadata from complex M Query expressions using AI toggle to extract some or all of the connection information from M query expressions of the BI source using Generative AI. Once you turn on this toggle, a pop-up message will appear. Read the message and click Continue to enable the feature.
Note
Enabling this feature may increase the metadata extraction time.
Once you turn on the toggle, the M queries will be parsed during the next MDE, which will include a call to the LLM in the metadata extraction process.
Security¶
Content Privacy¶
Amazon Bedrock ensures the following protections for your content:
Content is not utilized to improve underlying AI models.
Query content is not stored or reused by the model provider. No inputs are retained for training or inspection..
All data remains encrypted at rest and in transit.
Data travels exclusively over AWS’s private network infrastructure, not over the public internet. See AWS Documentation for more information.
Region |
Private Link |
|---|---|
us-east-1 (US East, N. Virginia) |
Yes |
us-west-2 (US West, Oregon) |
Yes |
ap-northeast-1 (Asia Pacific, Tokyo) |
Yes |
ap-southeast-1 (Asia Pacific, Singapore) |
Yes |
ap-southeast-2 (Asia Pacific, Sydney) |
Yes |
ap-south-1 (Asia Pacific, Mumbai) |
Yes |
ca-central-1 (Canada, Central) |
Yes |
eu-central-1 (Europe, Frankfurt) |
Yes |
eu-west-1 (Europe, Ireland) |
Yes |
Safety¶
Amazon Bedrock implements automated abuse detection mechanisms to identify and mitigate potential violations of AWS’s Acceptable Use Policy (AUP), Responsible AI Policy, or a third-party model provider’s AUP.
Abuse detection mechanisms are fully automated, so user inputs and model outputs are not reviewed or accessed by humans. See Amazon Bedrock abuse detection in the AWS documentation for more information.
Geographical Availability¶
Amazon Bedrock-backed AI features are currently not universally available across all Alation-supported regions. To extend AI functionality, Alation securely routes traffic cross-region through AWS private infrastructure using TLS 1.2 encryption. See the following table for region-specific routing details:
Origin Region |
Target Region |
|---|---|
us-east-1 (US East, N. Virginia) |
us-east-1 (US East, N. Virginia) |
us-west-2 (US West, Oregon) |
us-west-2 (US West, Oregon) |
ap-northeast-1 (Asia Pacific, Tokyo) |
ap-northeast-1 (Asia Pacific, Tokyo) |
ap-southeast-1 (Asia Pacific, Singapore) |
ap-southeast-1 (Asia Pacific, Singapore) |
ap-southeast-2 (Asia Pacific, Sydney)* |
ap-northeast-1 (Asia Pacific, Tokyo) ap-northeast-2 (Asia Pacific, Seoul) ap-south-1 (Asia Pacific, Mumbai) ap-southeast-1 (Asia Pacific, Singapore) ap-southeast-2 (Asia Pacific, Sydney) |
ap-south-1 (Asia Pacific, Mumbai)* |
ap-northeast-1 (Asia Pacific, Tokyo) ap-northeast-2 (Asia Pacific, Seoul) ap-south-1 (Asia Pacific, Mumbai) ap-southeast-1 (Asia Pacific, Singapore) ap-southeast-2 (Asia Pacific, Sydney) |
ca-central-1 (Canada, Central) |
ca-central-1 (Canada, Central) |
eu-central-1 (Europe, Frankfurt) |
eu-central-1 (Europe, Frankfurt) |
eu-west-1 (Europe, Ireland)* |
eu-central-1 (Europe, Frankfurt) eu-west-1 (Europe, Ireland) eu-west-3 (London) |
* Cross-region calls are allowed from the origin region to any of the mentioned target regions. The underlying model is hosted in specific AWS regions. When a customer’s deployment region doesn’t support the model, the request is routed securely to the nearest supported region using AWS PrivateLink. See Supported foundation models in Amazon Bedrock for more information.
Frequently Asked Questions¶
Initially, metadata extraction may take longer due to generative AI processing each complex M query individually. However, subsequent MDE runs will significantly improve speed, as Alation caches responses from the AI for unchanged M queries. Only the response and a MD5 hash of the expression is cached and not the actual expression.
Only M query expressions that cannot be parsed by Alation’s built-in parser are sent to the AI model via Amazon Bedrock. The AI model helps extract connection metadata such as hostname, database, and port that is used to populate the catalog and support lineage generation. No other data is sent.
Identifies data sources utilized in Power BI reports.
Extracts the connection information from the M query.
Detects the host, port, table name, query, and other information which helps to construct Lineage.
Helps surface source-level metadata and basic transformation logic where possible.
Generative AI parsing is exclusive to Alation’s Power BI Scanner OCF connector and is explicitly employed for complex M queries that Regex-based parsing cannot reliably handle.
The Claude Sonnet model is created by Anthropic. Alation accesses this model securely via its integration with Amazon Bedrock.
Alation does not directly train or modify the model. It uses it for interpreting complex Power BI M query expressions only.
No. File-based inputs like excel, workbook, sharepoint files, tables from rows, and CSV documents are skipped, as they do not contain connection-level lineage metadata.
Parameters can be used in M queries to dynamically insert values (like project names, dataset names, and others). The AI Parser does not yet include that in scope as of today. In such cases, the AI parser will fail.
These expressions do not contain structured connection metadata like host, port, or schema, and typically do not contribute to source-level lineage.
If parsing fails or exceeds limits, no lineage is created for that expression. The failure is recorded in job history logs