stephenschoettler/hermes-lcm v0.15.0

on GitHub

Highlights

• Added the benchmark-backed codex_spark_context preset for GPT-5.3 Codex Spark and near-128k Codex-style routes.
• /lcm preset suggest now separates 128k Spark-style windows from the 272k codex_gpt_long_context preset.
• Preset docs and tests now cover Spark benchmark provenance and dry-run application; runtime config remains inspect-only with no auto-apply behavior.

Release: https://github.com/stephenschoettler/hermes-lcm/releases/tag/v0.15.0

Bounded context, unbounded memory. Nothing is ever lost.

Contributors

• @stephenschoettler