Repository navigation
Conversation
|
Naming: "run" already means something else upstream, so a playbook instance needs a different noun. Upstream uses "run" for one execution of a turn inside a thread: Two meanings of a core word in one server is a reading tax on every file that touches both, and the tables and ids are the part that gets expensive to rename after data exists. Please pick another noun for a playbook instance and apply it to the contract type, tables, ids, route and copy in one pass. The UI already leans toward just "playbook" ("1 playbook", "No playbooks in this scope"), so that or "execution" would both read fine; Jackson's call. Filed from the stack review by Claude Fable 5.1 (Claude Code). |
|
Heads-up: this PR touches Squadron code that is being removed. What changed on
Still to come. The server migration re-keys the ledger from Squadrons to projects, removes For this PR. It uses Squadron ids in about 16 changed lines in When you next rebase: use the thread's or workspace's Posted by an AI agent on Jackson's behalf. |
Playbook execution needs durable state and predictable behavior across retries and restarts.
Add the event-backed run store, pure transitions, typed contracts, migrations, and worker reconciliation. Preserve completed evidence and require explicit recovery of interrupted work.
Stack 2/9. Depends on #156. Merge in order, retargeting to
j5/mainas predecessors land. Full-stack testing branch:codex/playbooks-09-aurora-loops.Validation at the integrated stack tip: 179 server/shared tests and 104 web tests passed; server and web typechecks passed. Each regenerated runtime manifest was checked. The adapter slice independently passed 92 tests, including the parser regressions.
Model: GPT-5 · Harness: Codex