Concepts
How ashlr's token-saving primitives work — snipCompact, genome RAG, the savings ledger, and the live status line.
These pages explain how ashlr saves tokens, not what it does. Reach for them when you want to understand why a particular tool returns the shape it does, or when you're tuning configuration.
Pages in this section
| Status line | Every glyph in the live ledger — what each number, color, and animation state means. |
| Savings math | How tokensSaved is computed and why every value is reproducible byte-for-byte. |
| Stats schema | The on-disk JSON shape of the per-session ledger. |
| Animation | The reduced-motion-friendly status line animation contract. |
| Confidence badges | What genome-retrieved 3 sections and similar annotations on grep output mean. |
| Genome | The on-disk knowledge graph that powers grep RAG and cross-machine sync. |
| Automatic hooks | The two lifecycle hooks (SubagentStop, Stop) — what each does and the env toggles that control them. |
Where these connect
- Status line + Savings math are the user-facing surface. If the number you see doesn't match what you expect, start in Savings math. The
(API-measured)vs(est.)label in Savings math explains whether your lifetime counter is heuristic or API-verified. - Genome + Confidence badges govern grep RAG quality. If grep results feel sparse, check Genome.
- Stats schema is the source of truth for any tooling that reads
~/.ashlr/stats.json. - Automatic hooks explains the background hooks that write to the session log and roll up subagent savings.