Knowledge catalogs in Kontext are backed by the unified project knowledge corpus. It combines indexed project documents with curated runtime knowledge records.
Current support on this page reflects the document extractor, document indexing job, and curated knowledge indexing available in the current runtime.
Knowledge Sources
| Source | Scope | Notes |
|---|
| Project document uploads | document | Uploaded project files indexed into document chunks |
| Curated runtime knowledge | knowledge | Runtime-owned curated records indexed into the same corpus |
| Upload type | Indexed for knowledge search | Notes |
|---|
Markdown (.md, .markdown) | Yes | Parsed as UTF-8 text |
PDF (.pdf) | Yes | Extracted through markitdown |
Text (.txt) | Yes | Parsed as UTF-8 text |
SQL (.sql) | Yes | Indexed as text content |
YAML (.yaml, .yml) | Yes | Indexed as text content |
Data lake files (CSV, TSV, Parquet, JSON, JSONL, Excel) | No | These belong to the data lake, not document indexing |
| Images, Draw.io, notebooks, charts, data tables | No | Stored as files or artifacts but not indexed as knowledge documents |
Available In
| Surface | Usage |
|---|
| Document indexing job | Rebuild or refresh project document chunks in the knowledge corpus |
| Data Agent / RAG tools | Search project documents for notes, documentation, and textual context |
| Unified knowledge index | Store both document and knowledge scopes in the same corpus |
