fix: isolate DB connection per tool call (backport #56)

[mergify]

LangGraph runs sync tools via asyncio.to_thread, which copies the calling context into the worker thread. Frappe's frappe.local is contextvar-backed, so tool worker threads inherited the same pymysql connection as the agent thread; concurrent tool calls (notably the document-planner subagent) corrupted the shared connection with "Packet sequence number wrong" / InterfaceError(0, '') errors.

Each tool call now opens a private connection, restores the session user, and closes it afterwards. The document-planner subagent tools are now wrapped too (they were previously passed raw).

---

This is an automatic backport of pull request #56 done by Mergify.

[greptile-apps]

Confidence Score: 4/5

The fix is structurally sound and directly addresses the shared-connection corruption; the only open question is that frappe.connect / commit / close block the event loop in the async path, and the async wrapper's lifecycle is not covered by the new tests.

The sync-tool isolation is well-tested and the logic in _isolated_db is correct. The async wrapper's new use of _isolated_db introduces blocking DB calls on the event loop thread and has no corresponding test — both are non-blocking observations on an otherwise clean fix.

The async branch of clear_messages_on_tool_error in toolset.py (lines 1067–1075) and the lack of a matching async test in test_code_tools.py.

Sequence Diagram

%%{init: {'theme': 'neutral'}}%%
sequenceDiagram
    participant AgentThread as Agent Thread
    participant LangGraph as LangGraph Executor
    participant ToolThread1 as Tool Worker Thread 1
    participant ToolThread2 as Tool Worker Thread 2
    participant DB1 as Private DB Conn 1
    participant DB2 as Private DB Conn 2

    AgentThread->>LangGraph: invoke parallel tools
    LangGraph->>ToolThread1: asyncio.to_thread(tool_A)
    LangGraph->>ToolThread2: asyncio.to_thread(tool_B)

    ToolThread1->>DB1: "frappe.connect(set_admin_as_user=False)"
    ToolThread2->>DB2: "frappe.connect(set_admin_as_user=False)"

    ToolThread1->>DB1: execute queries
    ToolThread2->>DB2: execute queries

    ToolThread1->>DB1: commit()
    ToolThread1->>DB1: close()
    ToolThread1->>ToolThread1: restore inherited_db

    ToolThread2->>DB2: commit()
    ToolThread2->>DB2: close()
    ToolThread2->>ToolThread2: restore inherited_db

    LangGraph->>AgentThread: return results

Loading

%%{init: {'theme': 'base', 'themeVariables': {"darkMode": true, "background": "#0d1117", "primaryColor": "#21262d", "primaryTextColor": "#e6edf3", "primaryBorderColor": "#8b949e", "lineColor": "#8b949e", "textColor": "#e6edf3", "edgeLabelBackground": "#161b22", "actorBkg": "#21262d", "actorBorder": "#8b949e", "actorTextColor": "#e6edf3", "actorLineColor": "#8b949e", "signalColor": "#8b949e", "signalTextColor": "#e6edf3", "noteBkgColor": "#373320", "noteBorderColor": "#d4a72c", "noteTextColor": "#f0e6c0", "labelBoxBkgColor": "#21262d", "labelBoxBorderColor": "#8b949e", "labelTextColor": "#e6edf3", "loopTextColor": "#e6edf3", "activationBkgColor": "#30363d", "activationBorderColor": "#8b949e"}}}%%
sequenceDiagram
    participant AgentThread as Agent Thread
    participant LangGraph as LangGraph Executor
    participant ToolThread1 as Tool Worker Thread 1
    participant ToolThread2 as Tool Worker Thread 2
    participant DB1 as Private DB Conn 1
    participant DB2 as Private DB Conn 2

    AgentThread->>LangGraph: invoke parallel tools
    LangGraph->>ToolThread1: asyncio.to_thread(tool_A)
    LangGraph->>ToolThread2: asyncio.to_thread(tool_B)

    ToolThread1->>DB1: "frappe.connect(set_admin_as_user=False)"
    ToolThread2->>DB2: "frappe.connect(set_admin_as_user=False)"

    ToolThread1->>DB1: execute queries
    ToolThread2->>DB2: execute queries

    ToolThread1->>DB1: commit()
    ToolThread1->>DB1: close()
    ToolThread1->>ToolThread1: restore inherited_db

    ToolThread2->>DB2: commit()
    ToolThread2->>DB2: close()
    ToolThread2->>ToolThread2: restore inherited_db

    LangGraph->>AgentThread: return results

Loading

Comments Outside Diff (1)

1. ask_alyf/ask_alyf/test_code_tools.py, line 601-609 (link)

   No test for async wrapper's isolated-DB lifecycle

   The two new tests (test_clear_messages_wrapper_connects_without_changing_user and test_clear_messages_wrapper_propagates_commit_failure) both exercise the sync wrapper path. The async wrapper now also runs inside _isolated_db(), but test_clear_messages_wrapper_preserves_async_tools only checks that the wrapped function is a coroutine and returns the right value — it does not verify that frappe.connect / commit / close are called for async tools. A rogue exception in async commit or close would go entirely undetected by the current suite.

   

_{Reviews (1): Last reviewed commit: "fix: isolate DB connection per tool call..." | Re-trigger Greptile}

[greptile-apps]

Blocking I/O in async_wrapper

frappe.connect(), frappe.db.commit(), and frappe.local.db.close() inside _isolated_db() are all synchronous blocking calls. When async_wrapper enters or exits the context manager, these run directly on the event loop thread — not in a thread pool. For short-lived local DB operations this is low-risk in practice, but it stalls the event loop for the duration of connection setup and teardown on every async tool call.

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!