Behavior - large files

Large files and batches.

How FileDigest handles large inputs, multi-file jobs, and failure boundaries when a source exceeds the current processing envelope.

How inputs are handled

FileDigest processes each uploaded source as an ordered document and combines all converted sources into one Markdown digest. Multi-file jobs, ZIP batches, and folder uploads preserve source order and source headers so downstream LLMs can distinguish where each section came from.

Very large individual files are accepted up to the per-job limits shown in the Limits page. If a source cannot be processed within the current envelope, the job reports that condition instead of silently truncating content.

How output is assembled

Converted sources are stitched into a single Markdown digest in upload order. Source headers, tables, section headings, and metadata are preserved so the artifact can be pasted directly into an LLM or passed into an agent workflow.

Current behavior

The dashboard currently exposes one production output format: a source-tagged Markdown digest. It does not expose a manual split toggle. FileDigest handles supported inputs within the current per-plan envelope and returns a clear job status when an input falls outside it. See Limits for the current per-plan allocations.

The invariant is simple: output should be complete or explicitly marked failed. FileDigest should never make you guess whether a digest was silently clipped.