v2.5.1

npm install -g yuva-ai@2.5.1

Full Changelog: v2.5.0...v2.5.1

v2.5.0

npm install -g yuva-ai@2.5.0

Full Changelog: v2.4.0...v2.5.0

v2.4.0

npm install -g yuva-ai@2.4.0

Full Changelog: v2.3.0...v2.4.0

Release v2.3.0

What's Changed

feat: add token optimization commands and implement token budgeting logic; enhance CLI help and work package structure

Installation

npx yuva-ai init

Full changelog: https://github.com/yuvamind/yuva-ai/commits/v2.3.0

Release v1.6.0

What's Changed

feat: consolidate AI state into .yuva/, protect orchestration files, always boot AI in spawned terminals

• Sessions now live in .yuva/session/ (legacy .session/ dirs migrate
  automatically); orchestrate/status/doctor read both locations
• PROTECTED FILES rule injected into every work package, worker boot prompt,
  generated config, and AGENTS.md: workers must never delete/move .yuva/,
  .session/, .aiautomations/, AGENTS.md, CLAUDE.md or any AI config —
  "cleanup" never includes them
• yuva swarm clear now requires --force
• yuva swarm spawn always boots an AI CLI (--cli > configured tool > claude
  default) so worker terminals never sit idle; all windows open in the SAME
  project directory sharing one codebase and one bus

Co-Authored-By: Claude Fable 5 noreply@anthropic.com

Installation

npx yuva-ai init

Full changelog: https://github.com/yuvamind/yuva-ai/commits/v1.6.0

Release v1.5.0

What's Changed

feat: swarm is now the default flow — auto-spawn worker terminals in the same project dir

• yuva swarm spawn: auto-opens worker terminal windows (Windows/macOS/Linux),
  every window rooted at THIS project directory sharing one codebase and one
  .yuva/ bus — never a copy or different path
• yuva worker boot: runs inside spawned terminals — launches the configured
  AI CLI (claude/gemini/codex/agy/...) with a strict worker-loop prompt
  (claim → follow work package → task done → repeat; never leave the project
  directory); --headless runs the unattended loop instead
• orchestrate now reports swarm state (tasks/workers) when a bus exists, and
  otherwise instructs the DEFAULT multi-terminal flow; opt out with
  "yuva config set mode solo"
• Generated configs (all 20 platforms) + AGENTS.md: swarm flow is mandatory
  for multi-step tasks; solo only for trivial changes or on user request
• 9 new tests (terminal spawn specs, boot prompt safety, orchestrate swarm)

Co-Authored-By: Claude Fable 5 noreply@anthropic.com

Installation

npx yuva-ai init

Full changelog: https://github.com/yuvamind/yuva-ai/commits/v1.5.0

Release v1.4.1

What's Changed

fix: yuva upgrade rejected projects that only have .aiautomations/

Fresh clones often lack AGENTS.md/CLAUDE.md because generated master files
are gitignored — upgrade now accepts a bare .aiautomations/ dir and restores
AGENTS.md from the template.

Co-Authored-By: Claude Fable 5 noreply@anthropic.com

Installation

npx yuva-ai init

Full changelog: https://github.com/yuvamind/yuva-ai/commits/v1.4.1

Release v1.4.0

What's Changed

feat: auto-detect project on greeting + questions with recommended selectable options

• Claude Code: generate SessionStart hook that runs "yuva agent orchestrate"
  so project context is injected the moment a session opens — typing "hey"
  now gets a project status + suggested next actions
• All platforms (base template + AGENTS.md for codex/opencode/antigravity/
  aider): greeting rule — first message triggers orchestrate + status reply
  with numbered next actions
• All platforms: questions must offer 2-4 options with one "(Recommended)"
  and a reason, answerable by number or "go with your recommendations"
• Requirements agent: STEP 2 rewritten to mandatory options format, max 5-6
  questions per round, recommendations based on orchestrate output

Co-Authored-By: Claude Fable 5 noreply@anthropic.com

Installation

npx yuva-ai init

Full changelog: https://github.com/yuvamind/yuva-ai/commits/v1.4.0

Release v1.3.1

What's Changed

fix: yuva update failed on legacy projects with "Not initialized"

update required AGENTS.md, but projects initialized before the
CLAUDE.md→AGENTS.md migration only have CLAUDE.md — every legacy user was
told to run init. Now update accepts AGENTS.md, legacy CLAUDE.md, or a bare
.aiautomations/ dir, and auto-runs the upgrade migration for legacy installs
before regenerating configs.

Co-Authored-By: Claude Fable 5 noreply@anthropic.com

Installation

npx yuva-ai init

Full changelog: https://github.com/yuvamind/yuva-ai/commits/v1.3.1

Release v1.3.0

What's Changed

feat: swarm mode — multi-terminal orchestrator/workers with enforced quality gates

• yuva gate: runs real lint/typecheck/test/build commands, exits non-zero on
  failure; auto-detected from package.json (also Cargo/Go/Python), overridable
  via .aiautomations/config.json "gates"
• Task bus (.yuva/): zero-dependency, crash-resumable file bus — atomic task
  claims, dependencies, priorities, worker heartbeats, stale-claim release,
  event log
• yuva swarm: orchestrator terminal — init, plan brief, live dashboard with
  automatic gate verification, rejected tasks return to queue with feedback
• yuva worker: hybrid workers — interactive (worker next prints a full work
  package: agent prompt + checklists + standards + gate protocol) or headless
  (worker start --auto --cli "claude -p" pipes packages into any LLM CLI)
• yuva task: add/list/show/done/fail — task done enforces gates before
  accepting completion
• Native configs now include gate/swarm rules; .yuva/ auto-gitignored
• README: gates + swarm docs and FAQ (multi-vendor swarms, connection model)
• 36 new tests (task-bus, gate-runner, work-package, arg-utils)

Co-Authored-By: Claude Fable 5 noreply@anthropic.com

Installation

npx yuva-ai init

Full changelog: https://github.com/yuvamind/yuva-ai/commits/v1.3.0