heavy-file-ingestion-claude-code

To read large PDF files without exceeding token limits, convert the heavyweight documents into lightweight markdown or CSV artifacts first. This approach builds a lightweight index, ensuring only compressed text is sent to the model for extraction instead of raw binaries.

The best way to normalize large CSV or TSV files for structured analysis is running a deterministic conversion script to map the raw data into structured CSV artifacts. This index-driven workflow reads a generated index before processing the compressed data.

Claude Code does not directly ingest raw DOCX and XLSX binaries efficiently. Instead, it uses a bundled converter script to map these heavyweight files into markdown or CSV, creating lightweight summaries that prevent context slowdowns.

Loading heavy spreadsheets slows down document processing because raw binaries consume excessive model context. Converting these files into markdown or CSV first prevents wasted context and maintains processing speed during extraction tasks.

To handle missing dependencies when converting heavy documents, the conversion workflow implements a smart escalation strategy that optionally installs missing deterministic converter dependencies or retries with cheaper approaches based on the index.

You should avoid loading raw binaries into model context when processing large PDFs, DOCX, PPTX, XLSX, or CSV files. Converting them to markdown or CSV first ensures token-efficient workflows and prevents wasted context.