Knowledge Enrichment
Knowledge Enrichment is a two-layer API platform that transforms raw, unstructured enterprise content into structured, AI-ready assets. The Data Curation API (DC API) extracts and normalizes text from any file type; the Context API adds semantic meaning — summaries, classifications, entities, embeddings — on top of that normalized text. Together they form a preprocessing pipeline consumed by retrieval-augmented generation (RAG) pipelines, AI agents, and search indexes.
New here? Start with the Quickstart.
Two APIs, one pipeline
A consumer can use the DC API alone, or chain both APIs together.
Data Curation API (DC API)
Accepts raw files and produces clean, structured text. Key capabilities include:
- Format identification — detects file type before processing
- Text extraction — digital documents use Document Filters (direct text extraction, not OCR), image-based content uses OCR-based extraction, audio/video uses speech transcription
- Normalization — consistent Unicode, quote and dash normalization
- Semantic chunking — configurable chunk size and overlap with positional metadata
- Embeddings — optional vector generation per chunk
- PII detection and redaction — detect and mask sensitive data before storage
Output is stored in the Content Lake with per-tenant isolation. Processing is asynchronous — the consumer submits a job and polls for completion. The output is structured JSON containing text, metadata, chunk positions, and optionally embeddings.
Context API
Applies enrichment operations (called actions) to already-curated content. The Context API is read-only — it returns structured metadata without modifying source content. Actions include:
- Summarization
- Classification
- Named entity recognition (NER)
- Image description
- Embeddings
- Contextual chunking
- PII detection
Not every document needs every action — actions are selective and policy-driven.
The Context API supports two input flows:
- Direct upload — The consumer uploads a file and requests enrichment immediately. The file is stored temporarily and processed on demand.
- Event-driven (Content Lake) — When the DC API finishes ingesting a document, it automatically notifies the Context API via an internal event. The Context API retrieves the document from the Content Lake and applies enrichment without any additional request from the consumer.
In both cases, processing is asynchronous and results are retrieved by polling.
Navigation
| Section | Description |
|---|---|
| Quickstart | Step-by-step guide to your first API call |
| Overview | Platform architecture, key concepts, guardrails |
| Guides | Task-oriented walkthroughs: authentication, curating documents, enriching content, PII, testing |
| Context API Reference | All actions, endpoints, OpenAPI spec, samples |
| Data Curation API Reference | Endpoints, processing options, PII handling, OpenAPI spec, samples |
| Errors | HTTP errors, action errors, status values for both APIs |
| Releases | Release notes and changelogs |