Gentleman-Programming/gentle-ai v1.36.1

on GitHub

gentle-ai v1.36.1 — Codex native parity + native SDD status engine

The complete release: Codex native parity (from v1.36.0) and the native SDD status dispatcher (#750, first shipped in v1.35.0), reconciled into one line. Recommended for everyone.

Codex native parity

• Context7 MCP injected as [mcp_servers.context7] into ~/.codex/config.toml.
• Codex 0.137.0 (latest stable).
• SDD model-selection profiles — sdd-strong / sdd-mid / sdd-cheap via model_reasoning_effort, usable with codex --profile <name>.
• Multi-agent SDD (default-on) — features.multi_agent = true with agents.max_threads/max_depth; delegates phases via spawn_agent/wait_agent using reasoning_effort + fork_turns: "none", with graceful solo fallback.
• TUI per-phase model picker — three plan-tied presets (Low-cost / Recommended / Powerful), shown at install and in the Configure-models menu.

Native SDD status engine (#750)

• gentle-ai sdd-status [change] --json --instructions and gentle-ai sdd-continue [change] for authoritative, machine-readable phase routing.
• Native SDD Dispatcher Guard across all agent orchestrators: route by nextRecommended and dependency states, stop on blockedReasons.

Install

brew upgrade gentle-ai