Source DAT

The Source DAT operator converts DAT content into a standardized document table that RAG Index can consume directly. Point it at any Table DAT or Text DAT, choose how to split the data, and pulse Process to produce a table of documents with unique IDs, content, and metadata — ready to wire into a RAG Index operator.

Input/Output

Inputs

• Input 1 (optional): Any DAT (table or text). Used as the source when no explicit Input DAT is set on the parameter panel.

Outputs

• Output 1: The index_table DAT containing processed documents with columns: doc_id, filename, source_path, content, metadata, timestamp.

Usage Examples

Feeding a Table into RAG Index

1. Create a Source DAT operator and a RAG Index operator in your network.
2. Wire your data Table DAT into the Source DAT’s first input (or set it via the Input DAT parameter on the Source page).
3. Set Mode to control how rows are processed:
   • Each Row — every row becomes its own document, with all columns formatted as key: value pairs.
   • Single Column — each row becomes a document using only the column specified in Content Column.
   • Whole Table — the entire table is serialized as a single markdown-formatted document.
4. Pulse Process. Check the Status field to confirm completion.
5. Wire the Source DAT’s output into the RAG Index’s first input.
6. On the RAG Index, set Input Mode to “Doc Table” and pulse Create Index.

Indexing a Text DAT

1. Set Input DAT to a Text DAT (e.g., a script, notes, or any freeform text).
2. Pulse Process. The entire text content becomes a single document.
3. Wire the output into a RAG Index operator to make the text searchable.

Best Practices

• Each Row mode works well for structured data like CSV imports or database exports where each row is a distinct record.
• Single Column mode is useful when your table has a dedicated content column alongside metadata columns you do not need indexed.
• Whole Table mode is best for compact reference tables where context across rows matters (e.g., lookup tables, configuration data).
• Column matching for Single Column mode is case-insensitive, so Content, content, and CONTENT all work.

Troubleshooting

• “No input DAT specified or connected to in1”: Either set the Input DAT parameter or wire a DAT into the operator’s first input.
• “Column not found”: In Single Column mode, verify the Content Column value matches a header in your table. The match is case-insensitive.
• “Table has no data rows”: Your table DAT has a header row but no data beneath it. Add data rows and re-process.

Parameters

Source

Input DAT (Inputdat) op('source_dat').par.Inputdat OP

DAT to process as source content

Default:

"" (Empty String)

Mode (Mode) op('source_dat').par.Mode Menu

Each Row: all columns per row as one document. Single Column: one column as content. Whole Table: entire table as one document.

Default:

each_row

Options:

each_row, single_column, whole_table

Content Column (Contentcolumn) op('source_dat').par.Contentcolumn Str

Column name for content (Single Column mode only)

Default:

"" (Empty String)

Process (Process) op('source_dat').par.Process Pulse

Default:

False

Clear (Clear) op('source_dat').par.Clear Pulse

Default:

False

Status (Status) op('source_dat').par.Status Str

Default:

"" (Empty String)

Changelog

v0.2.02026-03-01

• Complete rewrite of SourceDATEXT for RAG indexing - 3 modes: Each Row, Single Column, Whole Table - Removed redundant chunking (handled by rag_index) - Simplified from 13 parameters to 6 - Falls back to in1 input when no Input DAT set - Outputs standardized 6-column table format
• Initial commit

Related Operators

• Rag Index
• Rag Retriever
• Agent