v0.15.0

AnimaWorks v0.15.0

This release adds background memory review and peer notes, a task board that Animas triage themselves, a skill ledger with rollback, and per-caller internal API tokens. It also shortens prompts and reorganizes the code into single-purpose packages.

Highlights

• Background memory review: after a context compaction, at the end of a task and every 10 chat turns, the background model makes up to 3 knowledge or peer-note updates without blocking the conversation. Peer notes (animas/{name}/peers/) are recalled whenever that person or Anima writes in.
• Better recall on background turns: heartbeat, inbox and cron turns recall recent actions plus a little related knowledge and episodes, and search_memory with scope=activity_log searches the full history.
• Task board: Animas read, claim (with leases), close and annotate tasks with animaworks-tool task .... The owner triages its own tasks, and the delegating Anima is told when a delegated task is closed. The web UI shows a four-column board (Todo, Running, Waiting, Done).
• Skill ledger: every skill change is recorded and can be rolled back; skills must be read before they are edited.
• Security: per-caller internal API tokens with per-endpoint authorization (default log mode), a single fail-closed tool access check, and a confirmation key before the first call_human in a session.
• Shorter prompts: about 20–30% fewer characters per request, and Mode S no longer carries the Skill, SendMessage and ListAgents built-in tools.
• Reliability: daily episode summaries no longer fail on large days or rate limits (with backfill of missed days). Reconsolidation stops rewriting the same knowledge files every night. Crashed Animas keep recovering with backoff. Every engine times out only after 1200 seconds without an engine event.
• First run: setup detects Claude Code and Codex CLI logins, recommends subscription providers, and the leader Anima greets the owner first.
• Also new: a Gemini API text-to-speech provider, and opt-in mid-run compaction of long task sessions.

Compatibility and upgrade notes

• Back up ~/.animaworks before upgrading. Startup migrations rewrite config.json and each Anima's status.json, remove retired keys, and resync runtime prompts from the shipped templates.
• Neo4j graph memory is removed (along with the neo4j extra and the datasketch dependency). Animas that used it return to the file-backed memory with Chroma retrieval.
• Mode B and the process_model setting are gone. Mode B settings become Mode A; phase3 is the only process model.
• Removed: the shared vector worker, the Usage Governor, nightly knowledge mutation (consolidation Phase B), and the CLI commands create-anima, list, gateway, worker, migrate-cron and migrate --resync-db.
• Internal modules moved (core.tools → core.integrations, engines under core/execution/engines/, gateways under server/gateways/, and others). core.tools stays as an alias, and a migration rewrites references in runtime tools and skills.

See CHANGELOG.md for the full list.

Full Changelog: v0.14.0...v0.15.0

日本語

• 会話の圧縮後・タスク終了時・チャット 10 往復ごとに、バックグラウンドで記憶を見直すようになりました。知識と「相手ごとのメモ」を最大 3 件まで更新し、そのメモはその相手から連絡があったときに思い出されます。
• heartbeat・inbox・cron でも、直近の行動と関連する知識・エピソードを少し思い出すようになりました。search_memory の scope=activity_log は全履歴を検索します。
• タスクボードを追加しました。anima が animaworks-tool task ... でタスクの確認・確保（リース付き）・完了・取消・メモができ、委任元にはタスクが閉じられたことが DM で届きます。Web UI は 4 列のボードになりました。
• スキルの変更を台帳に記録し、巻き戻せるようにしました。
• 内部 API に呼び出し元ごとのトークンを導入しました（既定はログのみ）。ツールの許可判定を 1 か所に集約し、call_human はセッション最初の呼び出しで確認キーを求めます。
• 1 回あたりのシステムプロンプトを約 2〜3 割短くしました。
• 日次のエピソード要約が大きな日やレート制限で失敗する問題を直し、抜けた日を後から補うようにしました。知識ファイルが毎晩書き直される問題も直しました。
• 更新前に ~/.animaworks をバックアップしてください。 起動時の移行処理が設定と prompts を書き換えます。
• Neo4j のグラフ記憶・Mode B・process_model 設定・共有 vector worker・Usage Governor・一部の CLI コマンドを削除しました。 設定は移行処理で書き換わります。