Releases: vcoderun/acpkit
Releases · vcoderun/acpkit
v0.9.0_2026-04-21
Highlights
- Added
langchain-acpas a first-class ACP Kit adapter for LangChain and LangGraph runtimes, including session-aware graph factories, provider seams, bridge-based capability wiring, native ACP plan support, event projection, and DeepAgents compatibility helpers. - Expanded
codex-auth-helperso Codex auth state can now back LangChain through a maintainedChatOpenAIfactory, making it possible to build Codex-backed LangChain graphs without hand-rolling token refresh or OpenAI client wiring. - Added
acpremoteas a transport/helper package for exposing any existing ACP server over WebSocket and mirroring remote ACP endpoints back into a local stdio ACP boundary. - Added command-backed remote serving through
acpremote.serve_command(...)andacpremote.serve_stdio_command(...), including metadata, health, auth, and remote working-directory propagation. - Added root
acpkittransport entry points and re-exports for remote usage, includingacpkit serve,acpkit run --addr ...,acpkit.connect_acp(...), andacpkit.serve_acp(...). - Added maintained examples for LangChain graphs, DeepAgents compatibility, and ACP Remote Codex transport flows.
Improvements
- Expanded
langchain-acpprojections to cover search, HTTP request, browser, terminal, community file-management, and finance tool families with truthful ACP rendering. - Added first-class docs, skills, and maintained examples for the Codex-backed LangChain path so the helper, adapter, and example surfaces describe the same integration story.
- Reworked remote-host ownership defaults so mirrored ACP clients no longer leak local host-backed capabilities or local working directories into remote Codex and command-backed ACP servers.
- Rebalanced the public documentation so top-level product messaging is adapter-neutral, with parallel Pydantic and LangChain quickstarts and stronger helper positioning.
Fixed
- Fixed mirrored remote session behavior so remote ACP servers stay authoritative for
cwd, filesystem ownership, and terminal ownership. - Fixed
acpremotecommand environment handling so explicitenv={...}overrides preserve inherited variables such asPATH. - Fixed ACP Remote client and stream handling for large ACP lines by making the reader limit configurable and wiring it through both client and server transport paths.
- Fixed documentation drift between package exports and API docs, including missing public exports and stale generated docs content.
- Fixed release-facing docs examples that previously contained machine-specific absolute paths and stray conversational artifacts.
v0.8.2_2026-04-18
- Fix skill installation command
v0.8.1_2026-04-18
- Improve README file
v0.8.0_2026-04-18
Release Summary — 2026-04-18
This release upgrades pydantic-acp to pydantic-ai-slim==1.83.0, expands the upstream
capability bridge surface, improves ACP-native plan mode behavior, and tightens documentation,
projection coverage, and runtime compatibility seams.
Highlights
- Upgraded
pydantic-acptopydantic-ai-slim==1.83.0. - Hardened hook observation and introspection against upstream private API drift.
- Added a generic bridge capability seam through
CapabilityBridge.build_agent_capabilities(...). - Expanded the built-in capability bridge surface to cover:
ThreadExecutorBridgeSetToolMetadataBridgeIncludeToolReturnSchemasBridgeWebSearchBridgeWebFetchBridgeImageGenerationBridgeMcpCapabilityBridgeToolsetBridgePrefixToolsBridgeOpenAICompactionBridgeAnthropicCompactionBridge
- Re-exported the new bridges from the public
pydantic_acppackage surface.
ACP Projection And UX
- Added
WebToolProjectionMapandBuiltinToolProjectionMap. - Added richer ACP tool-card projection for:
- web search starts and results
- web fetch starts, previews, and binary fetch fallbacks
- image generation starts and summaries
- upstream MCP capability calls and tool-list responses
- Added visible transcript updates for provider-owned compaction, including OpenAI preflight
compaction visibility and AnthropicCompactionPartprojection. - Fixed builtin tool result projection from
ModelResponsehistory. - Expanded the default tool classifier to recognize common web, image, and
mcp_server:*builtin
tool names.
Native Plan Mode
- Added
TaskPlanas the structured native-plan output surface for ACP plan mode. - Added a session-local
plan_generation_typeconfig option forPrepareToolsBridgewith
structuredandtoolsmodes. - Made
structuredthe default plan-mode behavior so native plan mode can require structured plan
generation without exposingacp_set_plan. - Restored tool-based native plan recording behind
plan_generation_type="tools". - Kept plan progress tools (
acp_update_plan_entry,acp_mark_plan_done) tied to
plan_tools=Trueexecution flows.
Runtime And Maintainability
- Moved remaining private and semi-private hook access behind a dedicated compatibility layer.
- Preserved hook ordering when wrapping
Hooks(...)capabilities. - Split the large
runtime/adapter.pydelegation surface into dedicated runtime mixins. - Added focused inline docstrings on the public ACP-facing adapter methods and runtime mixins.
Docs And Examples
- Extended the maintained bridges example to cover the new bridge and projection surfaces.
- Updated API and bridge docs to reflect the generic capability seam and the current bridge set.
- Clarified CLI ownership boundaries between
acpkit run ...,run_acp(...), and
create_acp_agent(...). - Expanded the factory example docs with explicit session-derived input handling and fallback
guidance. - Documented native plan persistence wiring and clarified how ACP-native plan rendering already
works throughAgentPlanUpdateplusacp_get_plan. - Clarified
FileSessionStorerecovery guarantees versus host-owned recovery metrics.
Verification
- Added regression coverage for the new capability bridges, hook compatibility paths, projection
updates, native plan generation modes, and adapter runtime mixin delegation paths. - Added helper-path tests to keep coverage above the repository thresholds.
v0.7.0_2026-04-15
What's Changed
- Bump codecov/codecov-action from 5 to 6 by @dependabot[bot] in #6
Full Changelog: v0.6.0_2026-04-11...v0.7.0_2026-04-15
v0.6.0_2026-04-11
Full Changelog: v0.5.0_2026-04-10...v0.6.0_2026-04-11
v0.5.0_2026-04-10
Changelog since v0.4.2
SDK / runtime
- Added native ACP thinking capability with session-local effort selection
- Added slash commands for dynamic mode switching and thinking effort control
- Added dynamic mode-derived slash commands instead of hard-coded ask/plan/agent commands
- Added reserved mode-id validation to prevent collisions with built-in slash commands
- Added real ACP run cancellation with preserved transcript/state on stop
- Added native plan persistence hooks and automatic plan-state syncing
- Separated native plan mode from agent-mode plan progress tooling
- Improved runtime internals with dedicated agent/session/prompt runtime modules
- Added bridge model-settings support and a new ThinkingBridge
- Added HookBridge hide-all mode for quiet integrations
Pydantic ACP examples
- Simplified strong agent examples and reduced unnecessary hook/projection noise
- Fixed workspace path handling so session cwd is authoritative
- Removed agent-managed plan file writing from example tool surfaces
- Hid .acpkit plan storage from normal repo search/read flows
- Restricted plan mode so shell/write surfaces are not exposed
Docs
- Reworked docs into a fuller production-grade docs site with better structure and styling
- Added getting started, adapter config, runtime controls, plans/thinking/approvals, providers, bridges, host backends, helpers, testing, and example guides
- Added API reference pages and maintained example walkthroughs
- Added/updated llms.txt and llms-full.txt with published docs URLs
- Added a Pydantic ACP overview page and fixed strict MkDocs nav/link issues
- Improved README and skill guidance around ACP Kit’s role as an SDK for exposing existing agent surfaces through ACP
Skills
- Added packaged .agents/skills/acpkit-sdk skill structure
- Split lightweight orchestration skill from longform root SKILL.md reference
- Added skill resources, references, helper scripts, and published llms links
Tooling / quality
- Added save-coverage and check-coverage flows plus COVERAGE summary generation
- Added coverage threshold enforcement (line and branch)
- Reworked pre-commit so heavy gates run only for major changes
- Fixed docs workflow/build setup and tightened MkDocs config
- Removed stale mypy-specific config and cleaned pyright/type-ignore handling
v0.4.2_2026-04-07
Full Changelog: v0.4.1_2026-04-07...v0.4.2_2026-04-07
v0.4.1_2026-04-07
Full Changelog: v0.4.0_2026-04-07...v0.4.1_2026-04-07
v0.4.0_2026-04-07
Full Changelog: v0.3.0_2026-04-06...v0.4.0_2026-04-07