Total token budget for the conversation (history + system prompt + file context).
A sliding window automatically drops the oldest messages to stay within this limit.
Default: 4096  |  WebLLM 0.5B: 4096  |  GPT-4o / Claude: 32768–131072  |  Ollama models vary.

Clears the conversation history sent to the model after each reply. Prevents models from repeating themselves or leaking their system prompt when the context fills up.

Never sent anywhere except api.github.com over HTTPS

Blank = all text files up to 50 files. Large repos may take a moment.