One handler per format.
Read whole, never chunked.
PARALLAX RC® reads a corpus through format handlers — plugins that know what a given file actually is. Each handler extracts text, structure and metadata in place, and each extraction carries the digest of the bytes it came from.
A format is a plugin, not a feature request.
Formats change, and every domain brings its own. Rather than one parser that half-reads everything, PARALLAX RC® loads a handler per family — so a new format is an added plugin, not a rebuilt product.
Whole files, no chunking
A handler presents the file entire. Nothing is cut into fragments before it is read, so a passage keeps the context it sits in.
Structure preserved
Headings, tables, revision marks, sheet metadata: what the format carries is kept, because it is often what proves the point.
Digest at extraction
Each extraction records a SHA-512 of the source bytes, so a citation can be replayed against the original file later.
Evolutive by design
A handler is versioned and hot-loaded. A corpus already indexed can be re-read by a newer handler without being re-ingested.
The families PARALLAX RC® reads.
Documents and correspondence
Word processing and page formats, with headings, tables and tracked revisions intact.
Read the handler page →Spreadsheets
Cell values, formulas and sheet structure, so a figure traces to its cell.
Read the handler page →Presentations
Slides, speaker notes and embedded tables, read as the dated deck they are.
Read the handler page →Mail archives
Containers opened to the single message, with headers, thread and attachments kept together.
Read the handler page →Drawings and models
Geometry, layers, title block, revision clouds and every annotation on the sheet.
Read the handler page →Programmes
Scheduling exports read as dated revisions, activity by activity, logic preserved.
Read the handler page →Images and scans
Layout-aware recognition, with the page image kept alongside the text.
Read the handler page →Markup and structured text
Web and interchange formats read with their tree, not flattened to a string.
Read the handler page →Archives and containers
Opened to their members; each member is read by its own handler.
Read the handler page →Logs and configuration
Machine-written records read line by line, with their timestamps as fields.
Read the handler page →Databases and extracts
Tables and rows read as records, each traceable to its source file.
Read the handler page →This is a handler catalogue, not a claim to read every format in existence. Thousands of formats exist and most are irrelevant to a file; where one matters and is not listed, it is built as a plugin against the same interface.
What the drawings plugin actually reads.
A drawing is not a picture. The handler reads the sheet as the record it is: the geometry, the layers it sits on, the title block that dates it, and every annotation written on it.
Sheet and title block
Sheet number, revision, scale and issue date are read as fields, not as pixels.
Geometry and layers
Entities are read with the layer they belong to, so a change can be attributed to a discipline.
Annotations and dimensions
Text, leaders, dimensions and revision clouds are extracted as citable passages.
Revision comparison
Two revisions of one sheet are compared entity by entity: what was added, moved or removed.
DWG · DXF · DWF · IFC · vector PDF