prompt

Template management, token counting, and context windowing for AI agents.

use prompt

Functions

Function	Signature	Description
template	(tmpl, vars) -> String	Fill <<var>> placeholders from a map
token_count	(text) -> Int	Estimate token count (~4 chars/token)
token_truncate	(text, max_tokens) -> String	Truncate text to fit token limit
context_window	(messages, max_tokens) -> [Message]	Fit messages within token budget (keeps newest)
chunk	(text, max_tokens) -> [String]	Split text into token-sized chunks
system_prompt	(role, instructions) -> Message	Format a system message
user_message	(text) -> Message	Format a user message
assistant_message	(text) -> Message	Format an assistant message
format_chat	(messages) -> String	Format message list into chat string

Example

use prompt

let tmpl = "Hello <<name>>, you have <<count>> messages."
let filled = prompt.template(tmpl, {name: "Alice", count: "3"})
# => "Hello Alice, you have 3 messages."

let tokens = prompt.token_count(filled)
# => ~11