OPENAI-AGENTS-JS V0.7.2 PATCHES HOSTED SHELL REPLAY AND CLEANS AGENT HISTORY
OpenAI’s openai-agents-js v0.7.2 fixes orphan hosted shell call replay and cleans public history in multi-turn agent sessions. The patch release [v0.7.2](https...
OpenAI’s openai-agents-js v0.7.2 fixes orphan hosted shell call replay and cleans public history in multi-turn agent sessions.
The patch release v0.7.2 stops replaying orphan hosted shell calls across turns and prunes those orphans from the public history in agents-core. That reduces accidental duplicate executions and confusing traces during multi-turn runs.
No API changes here—just stability fixes, docs updates, and version bumps. If you use the hosted shell tool or parse agent histories, upgrade looks worthwhile.
Prevents duplicate shell executions across turns that can cause unexpected side effects and noisy logs.
Produces cleaner, more reliable run histories, improving debugging and observability for multi-turn agents.
-
terminal
Run a multi-turn agent that uses the hosted shell; confirm no commands replay across turns and history shows only intended calls.
-
terminal
Validate any log parsers or analytics that consume agent public history to ensure pruning does not break assumptions.
Legacy codebase integration strategies...
- 01.
Safe patch upgrade from v0.7.1; regression test workflows that shell out or depend on public history artifacts.
- 02.
If you relied on orphan entries for debugging, update your tooling to accommodate pruned histories.
Fresh architecture paradigms...
- 01.
Start new agent projects on v0.7.2 for cleaner traces and fewer accidental retries in shell-based tools.
- 02.
Adopt hosted shell for controlled command execution with improved history hygiene.