speckit-memory-md-token-report

name	speckit-memory-md-token-report
description	Compare estimated token usage between full memory reads and optimized synthesis.
compatibility	Requires spec-kit project structure with .specify/ directory
metadata	{"author":"github-spec-kit","source":"memory-md:commands/speckit.memory-md.token-report.md"}

Token Report

Compare estimated token usage between the full durable-memory read and the optimized synthesis flow.

Use this when:

• you want a quick estimate of how much context the optimizer saves
• you are deciding whether the optional SQLite cache is worth enabling for a repo
• you want to compare the baseline markdown-only read against the synthesis path

When the optimizer is enabled and the CLI is available, run:

cd .specify/extensions/memory-md && npx speckit-memory token-report --feature specs/<feature>

Report:

• baseline full durable memory read
• optimized index-and-synthesis flow
• estimated token reduction

Token counts are estimates using @dqbd/tiktoken with the cl100k_base encoding (GPT-4 calibrated). Actual provider billing tokens may differ. For Claude or Gemini, estimates may be 15–30% off.

If the optimizer is disabled or unavailable, report the markdown-only baseline and note that the optimized comparison could not be measured.