Gpt Bat

GPT:BAT is deliberately small, and that is the point. The entire product is one page with seven controls: a separator mode (newline, with aggregation), a System prompt, a User prompt, Max Tokens, Temperature, a model selector, and an OpenAI/API2D key field. Below that you either choose a file or paste text, then hit start. That narrowness means there is almost nothing to learn — the page itself links a tutorial and the source code, so you can read exactly what it does before you trust it with a folder of documents. The parts that justify using it over a five-line Python script are the boring ones. The token estimate before the run tells you roughly what a large batch will cost on your own key, before you commit. The progress counter lets you see a 500-block job advancing instead of guessing whether it hung. Automatic retries on OpenAI rate limits mean a long run does not die at block 320. And results can be read individually in the page or exported wholesale as JSON or CSV, which is the format most people actually want to hand to someone else. The constraints are real and worth planning around. Input splits on newline, so any document whose records span multiple lines will be shredded into the wrong units — this tool is for clean line-delimited text, not for arbitrary PDFs or nested records. Text is bounded by what the browser will store: the page states 10MB for Edge and Chrome and 5MB for Safari, so a large corpus has to be run in chunks. Everything bills against your own OpenAI or API2D key, and there is no cost cap beyond the estimate shown up front. GPT4o-mini is the model named in the interface and OpenAI and API2D are the only key sources the page mentions. Because the run happens in the tab, closing the window kills the job — there is no queue, no scheduling, and no way to monitor it from anywhere else. Results leave as JSON or CSV; nothing pushes into Notion, Airtable or a database. Where it fits: a developer or researcher with a few hundred line-separated records and one prompt they want applied uniformly, who wants output in a spreadsheet today and does not want to write and babysit a loop. Where it does not: recurring jobs, multi-user workflows that need retention policies or audit trails, and any dataset where one record spans several lines.