feat: examples/ demos + fix record_turn

[SonicBotMan]

3 个端到端 demo + 修复 record_turn 参数顺序

• examples/01_basic_workflow.py — DAG 创建、手动推进、状态查询
• examples/02_scheduler_and_memory.py — 自动调度、三层记忆系统
• examples/03_error_handling.py — 超时、取消、重试、环检测
• Fix: memory/__init__.py record_turn 缺少 workflow_id 参数导致参数错位

---

Note

Low Risk
Low risk: adds standalone example scripts and makes a small, backward-compatible fix to MemorySystem.record_turn to avoid mis-ordered args when calling episodic recording.

Overview
Adds three runnable demo scripts under examples/ covering (1) DAG creation + manual step advancement + status/progress checks, (2) scheduler-driven parallel execution with a mock executor plus working/episodic/semantic memory usage, and (3) timeout, retry, cancellation, and cycle-detection scenarios.

Fixes hermes-plugin/memory/__init__.py MemorySystem.record_turn to explicitly pass workflow_id to EpisodicMemory.record, preventing parameter misalignment and ensuring turn events are recorded with the intended execution_id/metadata.

^{Reviewed by Cursor Bugbot for commit 85bf31b. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{Bugbot Autofix is ON, but it could not run because on-demand usage is turned off. To enable Bugbot Autofix, turn on on-demand usage and set a spend limit in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 85bf31b. Configure here.}

[cursor]

Setting scheduler causes background execution race with manual advances

Medium Severity

Calling ws.set_scheduler(...) causes start_workflow to automatically fire asyncio.create_task(self._scheduler.execute_workflow(workflow_id)) in the background using the default executor. The rest of the example then performs manual advance_step calls on the same steps, creating a race between the background scheduler and the manual advances. Example 02 correctly avoids this by not calling set_scheduler.

Additional Locations (1)

• examples/01_basic_workflow.py#L80-L81

 

^{Reviewed by Cursor Bugbot for commit 85bf31b. Configure here.}

[cursor]

Layer count displays step count instead of layers

Low Severity

wf.get('layers', {}) is a dict mapping each step_id to its layer index. len() of this dict returns the number of steps (6), not the number of unique layers (5: layers 0 through 4). The display "DAG 层级: 6 层" is incorrect — it needs to count unique values, e.g. len(set(wf.get('layers', {}).values())).

 

^{Reviewed by Cursor Bugbot for commit 85bf31b. Configure here.}