Summary

• Add NOT_GIVEN to the _types import in src/agentex/resources/agents/agents.py to fix NameError: name 'NOT_GIVEN' is not defined raised by the custom create_task/cancel_task helpers (which surfaced in tests/test_client.py::test_retrying_timeout_errors_doesnt_leak).
• Convert from tests.utils import assert_matches_type to relative imports in the new test files for checkpoints, deployments, and schedules so pyright can resolve them (matches the convention used by every other generated test).

Test plan

• rye run pytest tests/test_client.py tests/api_resources/test_checkpoints.py tests/api_resources/agents/ — 145 passed, 392 skipped
• rye run lint — All checks passed
• rye run pyright tests/api_resources/test_checkpoints.py tests/api_resources/agents/test_deployments.py tests/api_resources/agents/test_schedules.py — 0 errors

🤖 Generated with Claude Code

Greptile Summary

This PR resolves two categories of lint/test failures introduced by recently added custom RPC helper endpoints. It adds the missing NOT_GIVEN sentinel import and expands the agent_rpc_response import block in agents.py, and converts three test files from absolute tests.utils imports to relative imports matching the project convention.

• NOT_GIVEN import fix: The custom helpers (create_task, cancel_task, send_message, etc.) already used NOT_GIVEN as a default value, but it was never imported — causing a NameError at module load time. Adding it to the _types import line resolves this.
• agent_rpc_response expansion: The six specific response type aliases (CreateTaskResponse, CancelTaskResponse, etc.) are now imported by name and added to __all__ so they are properly accessible and re-exportable.
• Test relative imports: test_checkpoints.py, test_deployments.py, and test_schedules.py are each updated to use the correct relative import path that matches every other generated test in the repo and satisfies pyright's module resolution.

Confidence Score: 5/5

All changes are targeted import fixes with no logic changes; the custom RPC helper methods themselves are untouched and tests pass.

The diff is entirely mechanical: one missing sentinel is added to an import line, six names are added to an all list, and three test files switch from absolute to relative imports. No behaviour changes are introduced, the test suite and pyright both pass per the PR description, and the patterns used match the rest of the codebase.

No files require special attention.

Important Files Changed

Sequence Diagram

sequenceDiagram
    participant Caller
    participant AgentsResource
    participant rpc / rpc_by_name
    participant agent_rpc_response

    Caller->>AgentsResource: create_task(agent_id, params)
    AgentsResource->>rpc / rpc_by_name: method="task/create", id=NOT_GIVEN, jsonrpc=NOT_GIVEN
    rpc / rpc_by_name-->>AgentsResource: AgentRpcResponse
    AgentsResource->>agent_rpc_response: CreateTaskResponse.model_validate(...)
    agent_rpc_response-->>Caller: CreateTaskResponse

    Caller->>AgentsResource: cancel_task(agent_name, params)
    AgentsResource->>rpc / rpc_by_name: method="task/cancel"
    rpc / rpc_by_name-->>AgentsResource: AgentRpcResponse
    AgentsResource->>agent_rpc_response: CancelTaskResponse.model_validate(...)
    agent_rpc_response-->>Caller: CancelTaskResponse

    Caller->>AgentsResource: send_message(agent_id, params)
    AgentsResource->>rpc / rpc_by_name: method="message/send" (streaming)
    rpc / rpc_by_name-->>AgentsResource: SSE stream
    AgentsResource->>agent_rpc_response: SendMessageResponse / SendMessageStreamResponse
    agent_rpc_response-->>Caller: SendMessageResponse

Reviews (3): Last reviewed commit: "fix: sort imports in agents resource" | Re-trigger Greptile