Skip to content

Releases: AlmanacCode/Yoke

Yoke 0.1.7

Choose a tag to compare

@rohans0509 rohans0509 released this 11 Jul 19:36

Crash-safe native runtime deployment cleanup. Yoke now tags ephemeral provider directories with their owner process and reclaims only dead-owner directories while preserving active concurrent sessions.

Yoke 0.1.6

Choose a tag to compare

@rohans0509 rohans0509 released this 11 Jul 19:24

Native provider runtime compilation for Codex named agents and Claude skills, with isolated per-session deployment, child skill ownership, model-aware named roles, asynchronous Claude task collection, and verified session cleanup.

Yoke 0.1.5

Choose a tag to compare

@rohans0509 rohans0509 released this 11 Jul 07:49

Adds typed, per-Harness environment variables for embedded runtimes. Claude SDK and Codex app-server merge adapter defaults, Harness.environment, and typed credentials without mutating process-global state. Environment values are excluded from repr and serialization.

Yoke 0.1.4

Choose a tag to compare

@rohans0509 rohans0509 released this 11 Jul 07:29

Codex app-server normalization no longer reports internal user-message, reasoning, or agent-message lifecycle wrappers as tools. Real assistant text and native tool/subagent events remain unchanged.

Yoke 0.1.3

Choose a tag to compare

@rohans0509 rohans0509 released this 11 Jul 07:10

Full Changelog: v0.1.2...v0.1.3

Yoke 0.1.2

Choose a tag to compare

@rohans0509 rohans0509 released this 11 Jul 07:01

Full Changelog: v0.1.1...v0.1.2

Yoke 0.1.1

Choose a tag to compare

@rohans0509 rohans0509 released this 11 Jul 06:51

Full Changelog: v0.1.0...v0.1.1

Yoke 0.1.0

Choose a tag to compare

@rohans0509 rohans0509 released this 11 Jul 05:50

First public alpha release of Yoke, a provider-neutral Python SDK for building agent systems on Claude Code and Codex.\n\nHighlights:\n- Provider surfaces for Claude SDK and Codex CLI, Python SDK, and app-server\n- Portable agents, skills, subagents, workflows, goals, sessions, and event streams\n- Capability discovery, model listing and selection, structured output, authentication, and run persistence\n- Live-tested provider behavior with 440 automated tests