ACCESS MANAGEMENT

Users and API credentials

Add service users, issue scoped bearer tokens, and revoke access instantly.

Total users—All registered accounts

Active users—Allowed to authenticate

Active tokens—Usable service credentials

Service statusProtectedBearer authentication enabled

Service users

User	Company	Tokens	Status

◎Select a user

Choose a user to manage details and API tokens.

DOCUMENT INTELLIGENCE LAB

Train for your document types

Turn reviewed spreadsheet examples into reusable extraction profiles and reversible LoRA adapters.

Checking trainerLocal GPU environment

Datasets—Uploaded and inspected

Ready examples—Mapped spreadsheet rows

Training runs—Queued, completed, or failed

✦

GUIDED TRAINING WORKFLOW

Teach the extractor your fields—not your descriptions.

Upload corrected examples, map the source and expected values, inspect data quality, and launch a GPU training run. Original model weights remain untouched.

1Upload→2Map→3Validate→4Train→5Activate

CONTINUOUS LEARNING

Document extraction settings

Document type * Display name

Extraction guidance Optional · unlimited length Parameters · array of objects *

nametypedescriptionrequiredaliases[]allowed_values[]

Examples per batch Epochs Learning rate LoRA rank

Collect supervised examplesApply this schema during extraction. Require admin reviewHold feedback in the inbox before training. Train at thresholdCreate a LoRA candidate when ready. Auto-activate candidateAdvanced: use new adapter immediately.

HUMAN REVIEW

Supervised feedback inbox

Loading feedback…

NEW TRAINING DATASET

Dataset builder

Step 1 of 4

DATASET CHECKLIST

What produces the best adapter?

✓Correct values, not field descriptions
✓Examples from different layouts and issuers
✓Real nulls when a field is absent
✓Separate validation documents
✓At least 100 reviewed rows per type

PRIVACY

Keep sensitive data intentional

Training files remain on this server. Use consented data, minimize identifiers, and define a retention policy.

SUPPORTED NOW

Structured extractorLoRA · recommended

Visual OCR modelNeeds image/transcript pairs

DATA LIBRARY

Training datasets

GPU QUEUE

Training runs

DOCUMENT ROUTING

Document profiles and active adapters

Selected automatically from document_type

Users and API credentials

Service users

User

Bearer tokens

Train for your document types

Teach the extractor your fields—not your descriptions.

Document extraction settings

Supervised feedback inbox

Dataset builder

Ready for review

Safe specialization settings

What produces the best adapter?

Keep sensitive data intentional

Training datasets

Training runs

Document profiles and active adapters

Add a user

Generate bearer token

Copy this credential now

Unlock the control room

Service users

User

Bearer tokens

Teach the extractor your fields—not your descriptions.

Document extraction settings

Supervised feedback inbox

Dataset builder

Ready for review

Safe specialization settings

What produces the best adapter?

Keep sensitive data intentional

Training datasets

Training runs

Document profiles and active adapters