| 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.