Two fixes on top of the paged transport:
- Near-limit v2 framing no longer materializes the full base64 transport:
chunk lines are generated lazily from a single serialization, the 64 MiB
reassembly ceiling is enforced via Buffer.byteLength before any
full-payload allocation, and RPC stdout writes drain with backpressure
one physical line at a time. Peak RSS for a 63 MiB response drops
~686 MB -> ~521 MB; a rejected 80 MiB response drops ~507 MB -> ~259 MB
(parity with the v1 path).
- get_messages_page errors now carry a machine-readable code
(session_busy | stale_cursor). Both bundled clients' high-level
getMessages() drains discard partial pages and fall back to the legacy
snapshot on either code — previously a cursor invalidated by a
background mutation (e.g. an appended bash message) threw instead of
falling back. Direct page calls remain strict.
- Introduced `Max` as a first-class reasoning effort tier across all packages, including AI providers, coding agent configurations, and RPC protocols.
- Refactored model effort ladders to use wire-exact mappings and removed legacy effort aliasing (e.g., `max-to-xhigh` mapping).
- Updated model registry and provider configurations to support `Max` tier routing, color themes, and UI icon associations.
- Expanded test suites to provide end-to-end coverage for the new reasoning tier, including updated compatibility and fallback scenarios.
- Updated `ThinkingConfig` and related types to model effort-based thinking settings with optional defaults and routing metadata.
- Added a dedicated parser for `model.thinking` payloads to validate efforts and normalize new optional thinking fields.
- Extended protocol tests to verify effort-based thinking data is parsed correctly and unknown efforts are rejected.
- Updated `RpcClient` to spawn `omp` in a new session, cache its process group, and terminate descendants on stop.
- Fixed `stop()` to signal the cached process group with SIGTERM then SIGKILL so leaked grandchildren are terminated.
- Moved status-query reads in `create_app` to `asyncio.to_thread`, preventing FastAPI event-loop stalls.
- Added regression coverage for stopping a spawned grandchild process in the client tests.
- Added homepage metadata entries to the Rust and Python package manifests.
- Replaced OpenRouter HTTP-Referer values with https://omp.sh/ in completion and image requests.
- Updated Codex WebSocket typing and construction to use Bun.WebSocket for handshake header capture.
Fixes#1102
- Stop now marks the client as closed via `_mark_closed` with `RpcProcessExitError`, so `_wait_for_agent_end` wakes immediately instead of waiting for a timeout.
- Added a regression test using a hanging server subprocess to verify `stop()` unblocks `prompt_and_wait` promptly and raises `RpcProcessExitError`.
- Updated Kimi compatibility tests to distinguish Moonshot-hosted models from OpenCode-hosted ones for `reasoning_content` and assistant-content tool-call behavior.
- Added user, group, and extra_groups parameters to RpcClient initialization.
- Propagated those parameters through to subprocess startup calls.
- Added tests to verify the new kwargs are passed correctly, including None defaults and empty extra groups.
- Adjusted the test server in test_host_uris.py to re-emit host_uri_result frames as uri_echo notifications.
- Updated test client setup to attach an unknown-notification capture hook for uri_echo frames.
- Changed _await_echo to read captured uri_echo frames from that hook and fail fast if capture is missing.
- Added `_optional_str_list` to normalize `systemPrompt` payloads into a tuple when absent, a bare string, or a string array, and to reject invalid shapes.
- Updated `SessionState` to hold `system_prompt` as a tuple and to parse daemon payloads with the new helper in `parse_session_state`.
- Extended protocol tests to verify array and legacy string parsing, default empty tuple behavior, and invalid `systemPrompt` values.
- Added `toolStrictMode` support with `all_strict`/`none`/`mixed` options to OpenAI compatibility.
- Fixed OpenAI-completion strict-mode flows by capturing failed HTTP responses and retrying once as non-strict.
- Fixed completion error reporting by surfacing captured status, headers, and JSON `type`/`param`/`code` details.
- Improved strict-schema enforcement with WeakMap memoization and circular-schema detection in sanitization.
- Fixed OpenRouter provider lookup by resolving fallback model IDs for suffix and date variants in registry resolution.
- Refactored benchmark tooling and added async RPC error-window tracking for scheduled run execution.
- Added host tool execution framework with HostTool, HostToolContext, and host_tool() factory for custom tool integration.
- Added RpcConcurrencyError exception and _PromptLifecycleCoordinator to enforce single-flight constraint on prompt lifecycle methods.
- Enhanced JSON parsing with 10 validation helpers and enum frozensets for safe field extraction with detailed error messages.
- Replaced manual event/error list management with _BoundedHistory for bounded-size history with offset tracking.
- Added deep JSON cloning to prevent external mutations of stored payloads and improved UTF-8 error handling in subprocess stderr.
- Added custom_tools parameter to RpcClient and set_custom_tools() method for runtime tool registration.
- Added error handling callbacks (on_protocol_error, on_listener_error) and error history tracking (protocol_errors, listener_errors properties) for RPC failures.
- Added max_event_history and max_stderr_chunks parameters to bound retained history in long-lived hosts, preventing unbounded memory growth.
- Added assistant_text_with_thinking() and message_text_with_thinking() helper functions for backward compatibility with thinking block inclusion.
- Improved error correlation for id-less parse/unknown-command failures to waiting requests and changed prompt scheduling failures to raise instead of timing out.
- Refactored event dispatch to _dispatch_listeners() with exception handling and error recording for robust listener failure management.
- Added typed event listeners and granular event handling for all RPC notification types.
- Added set_todos RPC command and todoPhases session state field for todo phase management.
- Added RpcClient initialization parameters (thinking, tools, no_session, rpc_defaults) for startup configuration.
- Added install_headless_ui() method and todo management methods (get_todos, set_todos, clear_todos).
- Added TodoItem and TodoPhase dataclasses with parser functions for structured todo representation.
- Added RPC mode behavior: disables session title generation by default and resets workflow settings to built-in defaults.
- Added RpcClient class with subprocess-based RPC communication, thread-safe request/response handling, and 30+ public methods for agent operations.
- Defined 65+ public API symbols including protocol types, dataclasses, parser functions, and custom exception types for RPC communication.
- Implemented event listener system with on_event, on_ui_request, and on_extension_error callbacks for asynchronous RPC event handling.
- Added comprehensive test suite with integration tests for RpcClient and unit tests for protocol parsing functions.
- Configured Python package with setuptools, PEP 561 type hints support, and documentation for usage and API reference.
can1357