Converts monthly support case comment CSV exports into NotebookLM-compatible text chunks (under 200 MB and 500,000 words per file).

No naming convention required. Drop any CSV export from Salesforce into the project root directory (the same directory as chunk_csv_for_notebooklm.py and run-podman.sh) — the script picks up every *.csv file there, regardless of name. A descriptive name (e.g. 2026-06-germany-case-comments.csv) is still handy for your own organization, since it's used as the base name for the generated output files.

Files missing the expected Salesforce columns (Account Name: Account Name, Case Comment Number, Case Number, Case Comment CreatedBy Location, Comment Body) are skipped with a warning instead of causing an error.

When exporting from Salesforce, export only one month at a time. Monthly exports keep file sizes manageable and avoid timeouts or memory issues during processing.

Use these Salesforce report filters to scope each export to a single month:

Adjust the From/To dates for each month, then export and name the resulting CSV accordingly (e.g., 2026-01-germany-case-comments.csv for the example above).

The CSV inputs and generated notebooklm_chunks/ output can contain personally identifiable information (PII) and other sensitive support data, including:

• Customer and account names
• Case and comment identifiers
• Email addresses and personal names in comment bodies
• Internal support engineer names and locations

These data files are listed in .gitignore and are not committed to this repository. Only the conversion tooling is tracked in git. Do not commit CSV or chunk files, and treat any generated output as confidential.

No host Python installation required — run everything in a container:

./run-podman.sh

This will:

1. Build a Podman image with Python and pandas
2. Run the container to process matching CSV files in the current directory
3. Save output to notebooklm_chunks/ on your host
4. Clean up the container automatically

podman build -t csv-to-notebooklm .

podman run --rm \
  -v "$(pwd):/app/data:Z" \
  -w /app/data \
  csv-to-notebooklm

The -v flag mounts your current directory into the container, so:

• CSV files are read from your host directory
• Output files are written back to your host notebooklm_chunks/ directory

Nothing on your host. Everything runs in the container:

• Python 3.11
• pandas
• The conversion script

All dependencies are isolated in the container image.

After running, you'll find:

• notebooklm_chunks/*.txt — Text files ready for NotebookLM
• notebooklm_chunks/metadata.json — Chunk statistics
• notebooklm_chunks/README.md — Upload instructions

All files are created on your host system in the notebooklm_chunks/ directory.

If you encounter permission issues with the :Z flag on SELinux systems, you can remove it:

podman run --rm -v "$(pwd):/app/data" -w /app/data csv-to-notebooklm

If you modify the script or requirements, rebuild:

podman build -t csv-to-notebooklm .

The script output is displayed directly. If you need to debug:

podman run --rm -it -v "$(pwd):/app/data:Z" -w /app/data csv-to-notebooklm /bin/bash