`/neuroflow:data`¶

Data intake — locate, inventory, validate, and convert your raw data.

/data takes raw recorded data and gets it ready for preprocessing. It covers locating and inventorying the data, checking BIDS compliance, and running conversion scripts if needed.

When to use it¶

You've finished data collection and need to organize raw files
You want to check if your data is BIDS-compliant
You need to convert from a proprietary format (BrainVision, EDF, etc.) to MNE/BIDS
You're starting fresh with an existing dataset

Before the steps — ethics and data access¶

Ethics gate. For data from your own participants, Claude first reads .neuroflow/ethics/status.md. If the approval is missing, expired, or was not confirmed by a person, intake stops and Claude says why and points to /ethics. Public de-identified datasets and projects marked ethics: not-applicable are outside the gate.
What the model may read. ai_processing in the same file decides whether Claude may open participant-level files. With none, Claude looks at file names, sidecars and column names only, writes the scripts, and works from what they print in aggregate.

Three steps¶

Claude works through these in order:

Step 1 — Locate and inventory¶

Claude uses Glob and file inspection to understand your data:

How many subjects / sessions / runs
File formats present
Whether it looks BIDS-compliant already
Where it lives — a synced folder (OneDrive, Dropbox, iCloud Drive, Google Drive) or a network share gets a warning and a recommendation to work on a local copy
Which folders hold the original recordings — recorded as raw_roots in project_config.md (after asking), and read-only from then on
Whether it is a DataLad dataset — then datalad get / run / unlock / rerun are used instead of a parallel log
Collected vs planned participants (e.g. 18/48) when the preregistration records planned_n
That participants.tsv and file names hold pseudonyms only

Claude: I found 24 subject folders (sub-01 through sub-24), each with:
        - sub-XX_task-oddball_eeg.vhdr
        - sub-XX_task-oddball_eeg.vmrk
        - sub-XX_task-oddball_eeg.eeg

        No dataset_description.json or participants.tsv found.
        This is not yet BIDS-compliant.

        Saving inventory to .neuroflow/data/data-inventory.md

Output: data-inventory.md saved to .neuroflow/data/

Step 2 — Validate BIDS structure¶

Checks for BIDS compliance: - Naming conventions (sub-XX_task-YY_eeg.vhdr) - Required files (dataset_description.json, participants.tsv, events.tsv) - Sidecar JSON completeness

If bids-validator is installed, its --json output is condensed by bids_digest.py into counts per issue code with the first locations — no issue code is dropped, and the full JSON is kept.

If data is not in BIDS format, Claude asks whether you want to convert or proceed as-is.

Step 3 — Convert¶

Raw recordings are never modified: conversion reads from them and writes the BIDS copy elsewhere. BrainVision files are never renamed by hand — .vhdr, .vmrk and .eeg point to each other by name, so a renamed set breaks; MNE-BIDS writes the copy and fixes the pointers.

Runs or writes conversion scripts for common cases:

Source format	Target format
BrainVision (`.vhdr`/`.vmrk`/`.eeg`)	MNE Raw → FIF, or BIDS via MNE-BIDS
EDF / EDF+	MNE Raw → FIF
Raw folder structure	BIDS layout
`.set` (EEGLAB)	MNE Raw

Conversion scripts are saved in your project repository (not inside .neuroflow/).

Output: data-intake.md saved to .neuroflow/data/

Example session¶

/neuroflow:data

You: Data is in data/raw/ — BrainVision format, 24 subjects.

Claude: Found 24 subjects in data/raw/. BrainVision format (.vhdr/.vmrk/.eeg).
        No BIDS structure detected.

        Do you want to convert to BIDS format?
        1. Yes — convert with MNE-BIDS
        2. No — proceed as-is (I'll note the raw paths)

You: Yes — convert to BIDS.

Claude: I'll write a conversion script using mne_bids.write_raw_bids().
        The script will:
        - Create dataset_description.json with your study metadata
        - Create participants.tsv from your subject list
        - Convert each .vhdr file to BIDS naming convention
        - Add sidecar JSONs with EEG metadata

        [generates scripts/convert_to_bids.py]

Files read and written¶

Direction	Files
Reads	`.neuroflow/project_config.md`, `.neuroflow/flow.md`, `.neuroflow/data/flow.md`, `.neuroflow/ethics/status.md`, `.neuroflow/preregistration/status.md`
Writes	`.neuroflow/data/`, `.neuroflow/data/flow.md`, `.neuroflow/sessions/YYYY-MM-DD.md`, `scripts/` (conversion code)

/ethics — the approval and ai_processing this command checks first
/experiment — the recording setup that generated this data
/data-preprocess — the next step after data intake

/neuroflow:data¶