← all repositories
jupediaz/chatgpt-prompt-splitter

When ChatGPT chokes on your wall of text, this tool slices it up

A Flask app that chunks oversized prompts into digestible 15,000-character bites for language models.

564 stars Python LLMOps · EvalOther AI
chatgpt-prompt-splitter
Velocity · 7d
+0.5
★ / day
Trend
steady
star history

What it does

ChatGPT PROMPTs Splitter is a small Flask web app that takes a long text, breaks it into chunks under a character limit you set (default 15,000), and prepends instructions so the AI knows more chunks are coming. You copy each chunk, paste it into ChatGPT, and the model waits for the full payload before responding. That’s the whole trick.

The interesting bit

The actual cleverness isn’t the splitting—it’s the prompt engineering baked into chunk one. The tool tells ChatGPT to “wait for the completion of chunk transmission before processing,” turning a blunt instrument (character limits) into a crude but functional streaming protocol. For a problem that could be solved with a Python one-liner and patience, someone built a deployable web service.

Key highlights

  • Web UI with adjustable chunk size; copy buttons per chunk
  • First chunk auto-includes “wait for all chunks” instructions
  • One-click Vercel deploy; vercel dev for local testing
  • Unit tests included (unittest discover tests)
  • Live demo at chatgpt-prompt-splitter.jjdiaz.dev

Caveats

  • The splitting algorithm is described only as “simple”; no detail on whether it breaks at word boundaries, sentences, or mid-word
  • Default 15,000 characters is a rough heuristic, not tied to actual token limits (ChatGPT uses tokens, not characters)
  • Flask app; no API endpoint mentioned for programmatic use

Verdict

Grab this if you regularly paste novels into ChatGPT and want a friction-free web UI. Skip it if you need token-aware splitting, semantic chunking, or API automation—this is a manual copy-paste tool with training wheels.

heatdrop uses Google Analytics to see which pages get read — nothing else. Your call. How we handle data.