Skip to main content

Unstructured suite (Read & Vision)

The Unstructured suite provides REST APIs for one-off text and image analysis. Use Read for linguistic metrics and sentiment; use Vision for color, texture, gradient, and keypoint features.

These APIs are separate from the Lake medallion flow. For bulk image or text corpora, prefer unstructured zip ingest into bronze, then publish gold for ML.

Products

APIEndpoint baseUse for
Readread.heimdallapp.orgSentiment, word counts, TF-IDF terms, part-of-speech stats
Visionvision.heimdallapp.orgColor histograms, texture, gradients, ORB keypoints

Getting started

  1. Create your account
  2. Generate an unstructured API key in Settings
  3. Make your first API call or jump to text analysis / image analysis

Monitoring usage

Read and Vision call volume appears on Usage (/usage) → Data Intelligence, alongside ML, Forecast, and Loop. There is no separate per-request detail page — use account-level charts to track unstructured API volume.

Lake vs unstructured APIs

ApproachBest when
Read / Vision APIsSingle documents, live user uploads, lightweight enrichment
Lake zip ingestHundreds or thousands of files, labeled folders, training datasets
Gold → MLYou need a deployed classification model on extracted features

FAQ

Do I need Lake for Read/Vision?
No — single API calls only need the unstructured key.

When do I use zip ingest instead?
Hundreds+ of labeled images or text files for ML training → Lake zip ingest.

Where is usage tracked?
UsageData Intelligence tab — no per-request detail page for Read/Vision.

Is this the same key as ML models?
No — ML/Forecast/Loop keys are per-asset on each detail page.

Sidebar shortcut?
Footer link Looking for unstructured? opens API key settings.

Next steps