Release 0.6.0

Release 0.6.0

Date: 2026-09-16

Token Terminator v0.6.0 adds request component attribution and SkillGate.

Highlights

• Provider-bound requests are now attributed across instructions, skill catalogs, tool schemas, tool results, current-user content, history, other fields, and request framing.
• SkillGate reduces large Hermes <available_skills> indexes before provider dispatch while preserving skills_list / skill_view discovery.
• There is no default maximum number of retained skills. Every skill above the relevance threshold is kept; an explicit max_skills remains available only as an opt-in host constraint.
• Routing touches only trusted system/developer instruction fields and leaves quoted skill tags in user/tool/history content unchanged.
• The current lexical-IDF scorer is deterministic and dependency-free; Runtime.set_skill_scorer(...) is the extension point for a future tiny trained/local relevance model.
• Raw/final component metrics are content-free and use exact tokenizer measurements when available, with the existing labelled fallback otherwise.

Compatibility

This is an in-place upgrade from v0.5.2. Existing content-addressed artifacts remain valid. The new request-attribution telemetry is additive. Python support remains 3.10–3.13. The Rust interoperability crate is released as token-terminator 0.6.0.

The v0.6.0 Git tag and GitHub release are cut from the same tested source tree used for the Python package and Rust crate.

Install

python -m pip install 'git+https://github.com/AronAxe/Token-Terminator.git@v0.6.0'

Rust:

[dependencies]
token-terminator = "0.6.0"