Tale v0.5.6

0.5.6 is a fix release on the 0.5 line: automation agent nodes list the connectors you have connected, keep the model pin you pick in New automation, and keep running past the first drain window — including after a worker restart.

Highlights

Automation equipment, wizard model pin, and drive chain (#3118)

Three stacked defects on the automation agent lane, plus the recovery backstop the third one exposed:

• Equipment lists connected connectors. The agent-node Equipment menu and the project-agent dialog always said "No connectors added yet," even when the organization had an active GitHub (or other) credential. They now list every connector slug that holds an active credential — disabled and needs-reauth stay out. The grant is the connector slug, the same vocabulary the sandbox bridge and dispatch already use.
• The New automation wizard keeps the provider pin. Picking a subscription model (for example Anthropic-subscription Fable 5) used to save a pinless node. The run's two-pass walk then served OpenRouter first and stole the pick. The wizard now stores {model, modelProvider} like the other pickers, filters subscription entries to the host default harness, and labels them "· Subscription".
• Turns keep running past the first window. The 0.4 driveWorkflowAgentTurn self-chainer was dropped in the Postgres cutover, so every multi-window agent turn failed after 4 attempts while the agent kept working in the sandbox. The drive job is back (automation.agent_drive, no retry, 12h expiry — a second drive would replay the ring buffer twice).
• A dead drive chain re-attaches. A worker restart that lost the drive chain used to strand the run until the 12h deadline. watchdog.automation_agents (every two minutes) now re-attaches waiting runs whose op lease went silent. Ask-parks (awaiting_human) are left alone — re-attaching one would settle a question mid-wait.

Failed runs that burned through the auto-retry ladder before this fix need a rerun after the deploy; the equipment and pin repairs apply to new nodes and new wizard saves.

Migration notes

No database migrations in this release. No operator action is needed beyond the upgrade itself.

Upgrading

• On the 0.5 line (0.5.0 – 0.5.5):

  tale update
  tale deploy

• New install:

  curl -fsSL https://raw.githubusercontent.com/tale-project/tale/main/scripts/install-cli.sh | bash
  mkdir tale-05 && cd tale-05
  tale init
  tale deploy

• Coming from 0.4.x: 0.5 remains a breaking minor with no in-place upgrade in either direction — read the v0.5.0 release notes and Upgrades → 0.4 → 0.5: breaking cutover before touching anything.

What's Changed

• fix(platform): automation equipment, wizard model pin, and drive chain by @larryro in #3118

Full Changelog: v0.5.5...v0.5.6