T

q792602257 031b929067 fix(multi-agent-collaboration): compose TaskRunner bookkeeping instead of forking it

CollaborativeTaskRunner hand-rolled its own step loop instead of
composing runtime/task.py's TaskRunner as design.md D3 requires,
silently dropping Timeline recording, WorldModel wiring,
TaskMetadataStore sync, and the on_task_succeeded/skill-synthesis
hook. Now calls TaskRunner's shared _start_world_view/
_record_step_result helpers for that bookkeeping. Also fixes
pre_observation being reused stale across steps in a multi-step plan
instead of refreshing to the prior step's post-observation.

openspec: multi-agent-collaboration capability, archived change multi-agent-runtime

2026-07-07 08:30:57 +08:00

agents

fix(multi-agent-collaboration): compose TaskRunner bookkeeping instead of forking it

2026-07-07 08:30:57 +08:00

api

feat: checkpoint device agent runtime milestones

2026-07-06 17:24:03 +08:00

cloud

fix(device-pool): scope primary key by host and thread through capability_tags

2026-07-07 08:30:47 +08:00

console

feat: checkpoint device agent runtime milestones

2026-07-06 17:24:03 +08:00

core

fix(device-pool): scope primary key by host and thread through capability_tags

2026-07-07 08:30:47 +08:00

device

fix(device-pool): scope primary key by host and thread through capability_tags

2026-07-07 08:30:47 +08:00

docs

feat: checkpoint device agent runtime milestones

2026-07-06 17:24:03 +08:00

driver

feat: checkpoint device agent runtime milestones

2026-07-06 17:24:03 +08:00

openspec

spec

2026-07-06 23:52:53 +08:00

perception

feat: checkpoint device agent runtime milestones

2026-07-06 17:24:03 +08:00

research

feat: checkpoint device agent runtime milestones

2026-07-06 17:24:03 +08:00

runtime

fix(semantic-scene): move describe_screen_semantic out of the tools/ layer

2026-07-07 08:30:49 +08:00

semantic

feat: checkpoint device agent runtime milestones

2026-07-06 17:24:03 +08:00

skills_learning

feat: add skill learning runtime

2026-07-06 18:07:53 +08:00

storage

feat: checkpoint device agent runtime milestones

2026-07-06 17:24:03 +08:00

tests

fix(multi-agent-collaboration): compose TaskRunner bookkeeping instead of forking it

2026-07-07 08:30:57 +08:00

tools

fix(semantic-scene): move describe_screen_semantic out of the tools/ layer

2026-07-07 08:30:49 +08:00

workflow

fix(workflow-orchestration): correct resume and failure-branch routing in WorkflowRunner

2026-07-07 08:30:43 +08:00

world

fix(world-model): scope WorldModel state per-task to prevent cross-task contamination

2026-07-07 08:30:36 +08:00

.gitignore

feat: checkpoint device agent runtime milestones

2026-07-06 17:24:03 +08:00

pyproject.toml

cloud

2026-07-06 23:44:18 +08:00

README.md

feat: checkpoint device agent runtime milestones

2026-07-06 17:24:03 +08:00

README.md

Device Agent Runtime

Device Agent Runtime is a device-agnostic runtime for LLM-driven automation. It gives agents a stable way to observe, decide, and act against real devices through a small set of domain models, driver contracts, tools, perception providers, and runtime orchestration.

iPhone automation through WebDriverAgent/Appium is the first driver, not the platform boundary. Future drivers can target Android, browsers, desktop environments, or other device surfaces without changing the runtime's core contracts.

Current Shape

core/: shared domain models and runtime errors.
driver/: the Driver contract, concrete driver adapters, and driver-type registry.
device/: device lifecycle and active driver management.
tools/: device capabilities exposed to runtime and API layers.
perception/: screen-to-Scene perception behind PerceptionProvider.
runtime/: planning and execution orchestration.
api/: REST/MCP transport adapters.
storage/: timeline, task, and device configuration persistence.

Project Direction

The durable roadmap is in docs/ROADMAP.md. The architecture invariants future changes must preserve are in docs/CONSTITUTION.md.