tinyjuice 0.2.1

Pluggable token compression for OpenHuman.
Documentation
# Reference Specs

These documents describe candidate TinyJuice strategies for reducing model
context growth. They are design references, not implemented behavior or
benchmark claims.

- [Ranked search and selective read]ranked-search-read-spec.md
- [AST stub reads]ast-stub-read-spec.md
- [Batched edit, fuzzy matching, and validation]batched-edit-validation-spec.md
- [SQL introspection reduction]sql-introspection-reduction-spec.md
- [Shell output interception]shell-output-intercept-spec.md
- [Web extract truncate-store]web-extract-truncate-store-spec.md
- [Subagent summaries]subagent-summary-spec.md
- [Savings accounting]savings-accounting-spec.md
- [TurboQuant-style vector compression]turboquant-vector-spec.md
- [TinyJuice integration]tinyjuice-integration-spec.md
- [TokenJuice improvement ingestion]tokenjuice-improvement-spec.md
- [Headroom algorithms and strategies]headroom-algorithms-strategies-spec.md
- [Headroom improvement ingestion]headroom-improvement-ingestion-spec.md
- [Hermes compression algorithms and techniques]hermes-compression-algorithms-spec.md