Guide

Find your way around the three modules.

Each module below is a self-contained tool with its own actions. Jump into whichever one matches what you're trying to do, or use the sub-commands as a quick reference for what each page can do.

[1] data-pre-inference
$cat clean --help
Upload a CSV (or import a Jupyter notebook's rendered DataFrame output) and CAT profiles columns, standardizes text, coerces numeric types, imputes missing values, removes outliers, and drops duplicates.
You get back a cleaned file, a quality report, and a row-level audit trail.
$cat clean upload <file.csv>
Drag a .csv onto the intake window, or browse to select one. Runs the full cleaning pipeline and redirects to the job's report.
$cat clean import --notebook <file.ipynb> [--table <name>]
Import a .ipynb that was run and saved with its outputs visible. CAT reads the rendered DataFrame only — notebook code never executes. Leave --table blank to use the first DataFrame found, or name a specific cell output.
$cat clean status <job_id>
Poll a job's progress while it's running — the report page uses this automatically, so you just watch it update.
$cat clean download <job_id>
Download the cleaned CSV once a job finishes.
$cat clean audit <job_id>
Download the row-level audit log — every cell CAT changed, and why.
open data-pre-inference →
[2] llm-payload-optimisation
$seedfmt --help
Re-encodes a JSON payload into a token-lean format before you send it to an LLM, so you pay for fewer tokens per request.
$seedfmt encode <payload.json>
Paste a JSON object or array of records. SeedFmt extracts a shared "seed" and encodes each record as a diff against it, instead of repeating the full structure every time.
Returns the encoded string plus a token count comparison against the raw JSON baseline, so you can see exactly how much you saved.
open llm-payload-optimisation →
[3] repocat
$repocat --help
Point RepoCat at a public GitHub repository and it indexes the codebase so you can ask grounded questions about it. Answers are sourced from the indexed repo, not a general guess.
$repocat add <https://github.com/owner/repository>
Clone and index a public repository, building its file tree, symbol index, and directory/language breakdown.
$repocat sync <repo>
Re-index an already-added repository on demand, to pick up new commits.
$repocat ask "<question>" --repo <repo>
Ask a question about an indexed repository — architecture overviews, where to start reading, or how specific pieces fit together.
open repocat →
workspace
$cat status --pre-inference
Total jobs, completed jobs, rows processed, and rows saved — shown on your command-centre dashboard.
$seedfmt stats
Payloads converted, records encoded, and tokens saved across all your SeedFmt runs.
$repocat stats
Repos indexed and questions answered across your RepoCat workspaces.
open command-centre →