release: v0.16.4 — parallel execution safety + bug fixes

CHANGELOG.md

@@ -5,6 +5,22 @@ All notable changes to selectools will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.16.4] - 2026-03-15
+
+### Fixed
+
+- **Parallel tool execution missing coherence checks**: Both `_execute_tools_parallel()` and `_aexecute_tools_parallel()` now run coherence checks before executing tools, matching sequential execution behavior.
+- **Parallel tool execution missing output screening**: Both parallel execution helpers now call `_screen_tool_result()` after tool execution, closing a security gap where parallel tool calls could bypass prompt injection detection.
+- **astream() usage tracking**: Fixed inline provider calls in astream() not feeding into `self.usage`, which caused `tool_usage` and `tool_tokens` dicts to remain empty during streaming.
+- **Input guardrails mutating caller's Message objects**: `_prepare_run()` now copies messages before applying input guardrails, preventing caller-side mutation when guardrails rewrite content.
+- **`ask()`/`aask()` missing `parent_run_id`**: Both convenience methods now accept and pass through `parent_run_id` for trace linking, matching `run()`/`arun()`/`astream()`.
+
+### Tests
+
+- **15 new tests** (total: 1477): Regression tests for all 5 fixes above, including parallel coherence/screening (sync + async), astream usage tracking, guardrail mutation prevention, and ask/aask parent_run_id.
+
+---
+
 ## [0.16.3] - 2026-03-14
 
 ### Added

CLAUDE.md

@@ -69,7 +69,7 @@ src/selectools/
 ├── types.py                 # Core types (Message, Role, ToolCall, AgentResult)
 └── env.py                   # Environment variable helpers
 
-tests/                       # 1462 tests (unit, integration, regression, E2E)
+tests/                       # 1477 tests (unit, integration, regression, E2E)
 ├── agent/                   # Agent core tests
 ├── providers/               # Provider-specific tests
 ├── rag/                     # RAG pipeline tests
@@ -281,5 +281,6 @@ Every `AgentTrace` contains `TraceStep` entries with one of these types:
 - **v0.16.1** ✅ Consolidation (6 bug fixes, thread safety, 68 new tests, mypy 0 errors)
 - **v0.16.2** ✅ astream() prompt leak fix + documentation updates
 - **v0.16.3** ✅ Agent refactoring + astream() full parity (14+ bug fixes, 29 new tests, ~800 lines dedup)
+- **v0.16.4** ✅ Parallel execution safety + 5 bug fixes
 - **v0.17.0** 🟡 Multi-Agent Orchestration — see `MULTI_AGENT_PLAN.md`
 - **Backlog**: Connector Expansion, Ecosystem Parity, Polish & Community

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "selectools"
-version = "0.16.3"
+version = "0.16.4"
 description = "Production-ready AI agents with tool calling, structured output, execution traces, and RAG. Provider-agnostic (OpenAI, Anthropic, Gemini, Ollama) with fallback chains, batch processing, tool policies, streaming, caching, and cost tracking."
 readme = "README.md"
 requires-python = ">=3.9"

src/selectools/__init__.py

@@ -1,6 +1,6 @@
 """Public exports for the selectools package."""
 
-__version__ = "0.16.3"
+__version__ = "0.16.4"
 
 # Import submodules (lazy loading for optional dependencies)
 from . import embeddings, guardrails, models, rag, toolbox