core/agent+cli: wire max_context_chars and re-raise CancelledError

Propagate context budget from [agent] config; preserve cancel through
tool invoke for mid-turn interruption.

CLAUDE.md

@@ -54,10 +54,10 @@ Async SQLAlchemy + aiosqlite. `MemoryStore`: schema init, default local user, se
 
 - **`ChatClient`** Protocol for `chat_completions`.
 - **`@tool` / `ToolRegistry`**: decorator infers JSON Schema from type hints; execute tools by name.
-- **`run_chat_loop`**: multi-round tool loop; default **streaming** text deltas + stream tool-call merge; parallel tools; tool-result char budget; optional `on_limit`.
+- **`run_chat_loop`**: multi-round tool loop; default **streaming** text deltas + stream tool-call merge; parallel tools; tool-result char budget; soft context compact on request; cooperative cancel points; optional `on_limit`.
 - **`ChatAgent`**: optional `MemoryStore` (persist on success only); `stream`; system prompt; `pending_retry_text` + `retry()`.
-- Events: text_delta, assistant_message, tool_call/result, max_rounds, **error**, **cancelled**.
-- Config ``[agent]``: `system_prompt`, `max_tool_result_chars`, `parallel_tools`, `confirm_destructive`, `path_denylist`.
+- Events: text_delta, assistant_message, tool_call/result, max_rounds, **error** (`retryable`/`source`), **cancelled** (`reason`).
+- Config ``[agent]``: `system_prompt`, `max_tool_result_chars`, `parallel_tools`, `confirm_destructive`, `path_denylist`, `max_context_chars`.
 
 ### Tools (`tools/`)
 

src/plyngent/agent/chat.py

@@ -4,7 +4,7 @@ from typing import TYPE_CHECKING
 
 from plyngent.lmproto.openai_compatible.model import SystemChatMessage, UserChatMessage
 
-from .budget import DEFAULT_TOOL_RESULT_MAX_CHARS
+from .budget import DEFAULT_CONTEXT_MAX_CHARS, DEFAULT_TOOL_RESULT_MAX_CHARS
 from .loop import DEFAULT_MAX_ROUNDS, run_chat_loop
 
 if TYPE_CHECKING:
@@ -35,6 +35,7 @@ class ChatAgent:
     system_prompt: str | None
     max_tool_result_chars: int
     parallel_tools: bool
+    max_context_chars: int
     messages: list[AnyChatMessage]
     pending_retry_text: str | None
 
@@ -54,6 +55,7 @@ class ChatAgent:
         system_prompt: str | None = None,
         max_tool_result_chars: int = DEFAULT_TOOL_RESULT_MAX_CHARS,
         parallel_tools: bool = True,
+        max_context_chars: int = DEFAULT_CONTEXT_MAX_CHARS,
     ) -> None:
         self.client = client
         self.model = model
@@ -67,6 +69,7 @@ class ChatAgent:
         self.system_prompt = system_prompt
         self.max_tool_result_chars = max_tool_result_chars
         self.parallel_tools = parallel_tools
+        self.max_context_chars = max_context_chars
         self.messages = list(messages) if messages is not None else []
         self.pending_retry_text = None
         self._ensure_system_prompt()
@@ -125,6 +128,7 @@ class ChatAgent:
                 stream=self.stream,
                 max_tool_result_chars=self.max_tool_result_chars,
                 parallel_tools=self.parallel_tools,
+                max_context_chars=self.max_context_chars,
             ):
                 yield event
             completed = True

src/plyngent/agent/tools.py

@@ -193,6 +193,8 @@ class ToolRegistry:
             result = definition.handler(**args)
             if inspect.isawaitable(result):
                 result = await result
+        except asyncio.CancelledError:
+            raise
         except TypeError as exc:
             return f"error: invalid tool arguments: {exc}"
         except Exception as exc:  # noqa: BLE001 — surface tool failures to the model

src/plyngent/cli/editor.py

@@ -28,6 +28,7 @@ _MINIMAL_CONFIG = """\
 # parallel_tools = true
 # confirm_destructive = true
 # path_denylist = ["/secrets/", ".ssh/"]
+# max_context_chars = 200000
 
 # [providers.example]
 # preset = "openai-compatible"

src/plyngent/cli/state.py

@@ -69,6 +69,7 @@ class ReplState:
             system_prompt=system_prompt,
             max_tool_result_chars=agent_cfg.max_tool_result_chars,
             parallel_tools=agent_cfg.parallel_tools,
+            max_context_chars=agent_cfg.max_context_chars,
         )
 
     def rebuild_client(self) -> None:

src/plyngent/config/models.py

@@ -18,6 +18,7 @@ class AgentConfig(Struct, omit_defaults=True):
     parallel_tools: bool = True
     confirm_destructive: bool = True
     path_denylist: list[str] = field(default_factory=list)
+    max_context_chars: int = 200_000
 
 
 class ModelConfig(Struct, omit_defaults=True):

tests/test_config/test_agent_section.py

@@ -17,6 +17,7 @@ def test_agent_section_defaults(tmp_path: Path) -> None:
     assert store.agent_config.parallel_tools is True
     assert store.agent_config.confirm_destructive is True
     assert store.agent_config.path_denylist == []
+    assert store.agent_config.max_context_chars == 200_000  # noqa: PLR2004
 
 
 def test_agent_section_parse(tmp_path: Path) -> None:
@@ -29,6 +30,7 @@ max_tool_result_chars = 100
 parallel_tools = false
 confirm_destructive = false
 path_denylist = ["/secrets/", ".ssh/"]
+max_context_chars = 5000
 """,
         encoding="utf-8",
     )
@@ -38,3 +40,4 @@ path_denylist = ["/secrets/", ".ssh/"]
     assert store.agent_config.parallel_tools is False
     assert store.agent_config.confirm_destructive is False
     assert store.agent_config.path_denylist == ["/secrets/", ".ssh/"]
+    assert store.agent_config.max_context_chars == 5000  # noqa: PLR2004