Conversation Simulator

A planning tool that estimates how token count and cost grow across a multi-turn LLM chat. You pick one of eight bundled models, optionally add a system prompt, then insert user and assistant messages one by one; the tool shows cumulative input/output token counts, remaining context window, total estimated cost, and a color-coded context-usage bar.

1. Choose a model from the Select dropdown (each option shows its context window size in K). 2. Optionally type a System Prompt. 3. Click Add User or Add Assistant to append message rows; type content into each Textarea. 4. Watch the Summary card update input tokens, output tokens, remaining tokens, and estimated cost, plus the usage bar (green under 50%, amber 50-80%, red above 80%).

Tokens are approximated as ceil(characters / 3.5) per message. The system prompt counts only toward input. Each user message adds to cumulative input tokens and each assistant message to cumulative output tokens. Cost for the whole conversation is computed as (cumulativeInput / 1_000_000 * model.inputPrice) + (cumulativeOutput / 1_000_000 * model.outputPrice), so every added turn rebills the full prior history as input — illustrating why chat costs grow super-linearly.