Gentleman-Programming/gentle-ai v1.36.0 on GitHub

gentle-ai v1.36.0 — Codex native parity

Brings Codex up to parity with the other agents and makes its configuration first-class.

Context7 MCP for Codex — now injected as [mcp_servers.context7] into ~/.codex/config.toml (previously skipped for TOML-based agents).
Codex bumped to 0.137.0 (latest stable).
SDD model-selection profiles — ~/.codex/sdd-strong.config.toml, sdd-mid, and sdd-cheap, model-agnostic via model_reasoning_effort, usable with codex --profile <name>.
Multi-agent SDD (default-on) — writes features.multi_agent = true plus agents.max_threads/max_depth. The Codex orchestrator delegates SDD phases via spawn_agent/wait_agent (passing reasoning_effort with fork_turns: "none") and falls back to solo execution when the tools are unavailable.
TUI per-phase model picker — pick a Codex SDD preset (Low-cost / Recommended / Powerful, tied to ChatGPT Plus $20 / Pro $100 / Pro $200) at install time and from the Configure-models menu.

brew upgrade gentle-ai

v1.36.1 additionally includes the native SDD status dispatcher (#750). Prefer v1.36.1 for the complete set.