This release adds 2 notable features for engineering teams evaluating rollout.
✓ No known CVEs patched in this version
Topics
Summary
AI summaryUpdates fix, Documentation & Other Changes, and feat across a mixed release.
Changes in this release
| Type | Severity | Summary | CVE |
|---|---|---|---|
| Feature | Low |
Make task and turn tracing spans configurable. Make task and turn tracing spans configurable. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Feature | Low |
Track response usage in realtime session context. Track response usage in realtime session context. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Bugfix | Medium |
Serialize OpenAI conversation session initialization correctly. Serialize OpenAI conversation session initialization correctly. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Bugfix | Medium |
Raise UserError for conflicting provider arguments instead of assert. Raise UserError for conflicting provider arguments instead of assert. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Bugfix | Medium |
Preserve user messages containing history wrappers during handoffs. Preserve user messages containing history wrappers during handoffs. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Bugfix | Medium |
Isolate provider instances across concurrent runs in computer component. Isolate provider instances across concurrent runs in computer component. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Bugfix | Medium |
Avoid redundant E2B workspace root creation in sandbox. Avoid redundant E2B workspace root creation in sandbox. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Bugfix | Medium |
Preserve streamed session input across model retries. Preserve streamed session input across model retries. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Bugfix | Medium |
Prevent stale prepared‑item identity reuse. Prevent stale prepared‑item identity reuse. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Bugfix | Medium |
Clear session and pop_item metadata leaks in AdvancedSQLiteSession. Clear session and pop_item metadata leaks in AdvancedSQLiteSession. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Bugfix | Low |
Redact non-tool trace error details. Redact non-tool trace error details. Source: granite4.1:30b@2026-07-17-audit Confidence: low |
— |
| Bugfix | Low |
Parse Google docstrings missing the blank line before a section. Parse Google docstrings missing the blank line before a section. Source: granite4.1:30b@2026-07-17-audit Confidence: low |
— |
| Bugfix | Low |
Bound $ref expansion in strict JSON schema conversion. Bound $ref expansion in strict JSON schema conversion. Source: granite4.1:30b@2026-07-17-audit Confidence: low |
— |
Full changelog
What's Changed
- feat: make task and turn tracing spans configurable by @seratch in https://github.com/openai/openai-agents-python/pull/3834
- feat(realtime): track response usage in session context by @seratch in https://github.com/openai/openai-agents-python/pull/3812
- fix(memory): serialize OpenAI conversation session initialization by @russeell in https://github.com/openai/openai-agents-python/pull/3821
- fix(models): raise UserError for conflicting provider args instead of assert by @Dhruva-Das in https://github.com/openai/openai-agents-python/pull/3809
- fix(handoffs): preserve user messages containing history wrappers by @russeell in https://github.com/openai/openai-agents-python/pull/3815
- fix(computer): isolate provider instances across concurrent runs by @russeell in https://github.com/openai/openai-agents-python/pull/3843
- fix(sandbox): avoid redundant E2B workspace root creation by @seratch in https://github.com/openai/openai-agents-python/pull/3845
- fix: preserve streamed session input across model retries by @seratch in https://github.com/openai/openai-agents-python/pull/3857
- fix: prevent stale prepared-item identity reuse by @seratch in https://github.com/openai/openai-agents-python/pull/3858
- fix: AdvancedSQLiteSession clear_session and pop_item metadata leaks by @okaditya84 in https://github.com/openai/openai-agents-python/pull/3755
- fix: parse Google docstrings missing the blank line before a section by @sohumt123 in https://github.com/openai/openai-agents-python/pull/3832
- fix: bound $ref expansion in strict JSON schema conversion by @ioleksiuk in https://github.com/openai/openai-agents-python/pull/3838
- fix: redact non-tool trace error details by @seratch in https://github.com/openai/openai-agents-python/pull/3848
Documentation & Other Changes
- docs: updates for hosted multi-agent support by @seratch in https://github.com/openai/openai-agents-python/pull/3789
- docs: make UserContext example runnable in Agents documentation by @MuhammadHamidRaza in https://github.com/openai/openai-agents-python/pull/3851
- chore: enable flake8-async for runtime source by @seratch in https://github.com/openai/openai-agents-python/pull/3813
- Release 0.18.3 by @github-actions[bot] in https://github.com/openai/openai-agents-python/pull/3859
New Contributors
- @Dhruva-Das made their first contribution in https://github.com/openai/openai-agents-python/pull/3809
- @russeell made their first contribution in https://github.com/openai/openai-agents-python/pull/3815
- @okaditya84 made their first contribution in https://github.com/openai/openai-agents-python/pull/3755
- @sohumt123 made their first contribution in https://github.com/openai/openai-agents-python/pull/3832
Full Changelog: https://github.com/openai/openai-agents-python/compare/v0.18.2...v0.18.3
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About openai-agents-python
A lightweight, powerful framework for multi-agent workflows
Related context
Related tools
Beta — feedback welcome: [email protected]