Running jobs with inline blueprints
Instead of creating a project, you can pass blueprints directly on a job with the
Blueprints parameter of StartDocumentInsights. This is useful for
one-off jobs or quick tests, where defining a reusable project is unnecessary. A job
provides exactly one of ProjectArn or Blueprints — not both.
Inline jobs return custom extraction results only
A job started with inline blueprints returns custom extraction
results only. It does not produce extraction results: there is no
extraction/ folder, only the custom/ folder and
job_metadata.json. If you need extraction as well, use a project with an
ExtractionConfiguration; see Working with projects.
Start a job with inline blueprints
Pass the blueprints as a JSON array to --blueprints. Each entry carries a
BlueprintArn and an optional BlueprintVersion. The call returns
a JobArn.
aws textract start-document-insights \ --input-configuration '{ "S3Uri": "s3://amzn-s3-demo-bucket/w2.pdf" }' \ --output-configuration '{ "S3Uri": "s3://amzn-s3-demo-destination-bucket/output/" }' \ --blueprints '[{"BlueprintArn":"arn:aws:textract:us-east-1:aws:blueprint/public-w2-form"}]' \ --region us-east-1
Then poll GetDocumentInsights with the returned job identifier until
Status is SUCCEEDED or FAILED. See Running and monitoring Document Insights jobs.
Output layout for an inline job
On success, the output location holds only the custom extraction result and the job
metadata — no extraction/ folder:
<prefix>/<job-id>/job_metadata.json
<prefix>/<job-id>/0/custom/0/result.json
The job_metadata.json for an inline job lists only custom_path
entries in segment_metadata (there is no extraction_path), with
the bucket name shown here as a placeholder:
{ "job_id": "b949fd79-3eee-4167-87bb-e35bac1d2d46", "job_status": "PROCESSED", "output_metadata": [ { "asset_id": 0, "segment_metadata": [ { "custom_path": "s3://amzn-s3-demo-destination-bucket/output/<job-id>/0/custom/0/result.json", "custom_status": "MATCH" } ] } ] }
The custom result.json has the same shape as the project-based custom
extraction results, including matched_blueprint, document_class,
split_document, inference_result, and
explainability_info. See Custom extraction with blueprints for the full field reference.
When to use a project instead
Use inline blueprints for one-off or exploratory jobs. Use a project when you want to:
-
produce extraction results alongside custom extraction results,
-
reuse the same configuration across many jobs,
-
enable splitting, generative output, sensitive-data handling, or a fallback blueprint.
See Working with projects for the configuration blocks a project supports. For request parameters and errors, see the Amazon Textract Insights API Reference.