feat(my-deepagent): v0.3 PR #5 — plan mode (/plan, /approve, /reject)

Claude Code의 plan mode 등가.  `/plan` 진입 시 write_file / edit_file /
execute / bash / task (sub-agent) 도구가 차단되고 read_file / glob / grep /
ls / write_todos 만 허용.

핵심 동작:
- `PlanModeMiddleware(is_active: Callable[[], bool])` 가 `awrap_tool_call` /
  `wrap_tool_call` 에서 활성 + 차단 도구면 synthetic
  `ToolMessage(status="error")` 반환.  raise 하지 않음 — LLM 이 차단 메시지를
  보고 다른 도구로 전환하거나 plan 다듬기로 자동 복귀.
- `is_active` 는 closure 라서 슬래시 토글 후 agent 재빌드 불필요.
- `InteractiveSessionRow.plan_mode` 영속 + resume 시 복원.

데이터·라이브러리:
- `middleware/plan_mode.py` (신규):
  - `BLOCKED_TOOLS_IN_PLAN_MODE = write_file / edit_file / bash / execute /
    run_command / shell / task`.
  - `PlanModeMiddleware` async + sync 양쪽 구현.

REPL 통합 (`cli/interactive.py`):
- `InteractiveSession._plan_mode: bool` + `set_plan_mode(enabled)` async →
  flag 토글 + `thread_suffix` bump + row 영속.
- resume path 에서 `sess._plan_mode = row.plan_mode` 로 복원.
- `_register_plan_mode_slash`: `/plan`, `/approve`, `/reject` 등록.
- `/reject` 는 thread 까지 리셋해 plan thread 폐기.

테스트 (`tests/integration/test_plan_mode.py`, 9 케이스):
- inactive 시 모든 도구 패스스루
- active 시 write_file / execute / task 차단 (status=error,
  tool_call_id 유지, 메시지에 도구명 + "Plan-mode" 포함)
- active 시 read_file / glob / grep / ls / write_todos 허용
- closure 토글로 동작 변경 (rebuild 없이)
- 동기 wrap_tool_call 도 동일 동작
- BLOCKED_TOOLS_IN_PLAN_MODE 상수 sanity

게이트:
- ruff check / format --check / mypy: PASS
- pytest -q --ignore=tests/integration/test_e2e_workflow.py
  --ignore=tests/integration/test_openrouter_smoke.py: 657 passed (9 신규 포함)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
chungyeong
2026-05-17 20:47:30 +09:00
parent 2685cb26db
commit fb7e67fd20
4 changed files with 380 additions and 2 deletions

View File

@@ -46,6 +46,7 @@ from ..memory import (
)
from ..middleware.audit import AuditToolMiddleware
from ..middleware.cost import CostMiddleware
from ..middleware.plan_mode import PlanModeMiddleware
from ..monitoring.pricing import ModelPrice, PricingCache
from ..monitoring.token_budget import count_tokens
from ..persistence.checkpointer import get_checkpointer_ctx
@@ -169,6 +170,9 @@ class InteractiveSession:
# users drop `<name>/SKILL.md` directories under here to register skills.
self.skills_dir: Path = user_skills_dir(config)
ensure_skills_initialized(self.skills_dir)
# v0.3 PR #5: plan-mode flag. PlanModeMiddleware reads this via closure
# every tool call — no agent rebuild needed when toggling on/off.
self._plan_mode: bool = False
@property
def thread_id(self) -> str:
@@ -216,6 +220,28 @@ class InteractiveSession:
self._agent = None
self._thread_suffix += 1
@property
def plan_mode(self) -> bool:
"""Whether plan mode is currently active for this session."""
return self._plan_mode
async def set_plan_mode(self, enabled: bool) -> None:
"""Toggle plan mode + persist to the session row.
PlanModeMiddleware re-reads via closure each tool call → no agent
rebuild required. We DO bump the thread suffix on each toggle so the
model doesn't carry over "I was about to write a file" state into the
new mode. Persists `plan_mode` on the InteractiveSessionRow so resumes
re-establish the mode.
"""
self._plan_mode = enabled
self._thread_suffix += 1
async with self.db.session() as s:
row = await s.get(InteractiveSessionRow, str(self.session_id))
if row is not None:
row.plan_mode = enabled
await s.commit()
def build_agent_if_needed(self) -> Any:
if self._agent is not None:
return self._agent
@@ -231,6 +257,9 @@ class InteractiveSession:
interactive_session_id=self.session_id,
file_recorder=make_audit_recorder(self.config.state_dir),
)
# v0.3 PR #5: plan-mode middleware reads `self._plan_mode` via closure
# every tool call → toggling /plan vs /approve doesn't require rebuild.
plan_mw = PlanModeMiddleware(is_active=lambda: self._plan_mode)
# Re-glob memory paths every time the agent is rebuilt — `/remember` and
# `/forget` call `clear_agent_cache()` so this picks up new/removed files.
memory_paths = list_memory_paths(self.memory_dir)
@@ -239,7 +268,7 @@ class InteractiveSession:
self._persona,
self.config,
root_dir=self.repo_root,
middleware=[cost_mw, audit_mw],
middleware=[plan_mw, cost_mw, audit_mw],
model_override=self._model_override,
checkpointer=self.saver,
memory_paths_override=memory_paths,
@@ -596,6 +625,43 @@ def _register_skills_slash(reg: SlashRegistry, sess: InteractiveSession) -> None
reg.register("skill", _skill, help="show a skill's body: /skill <name>")
def _register_plan_mode_slash(reg: SlashRegistry, sess: InteractiveSession) -> None:
"""Register /plan, /approve, /reject slash handlers (v0.3 PR #5)."""
async def _plan(_: SlashParsed) -> bool:
if sess.plan_mode:
_CONSOLE.print("[yellow]plan-mode is already active.[/]")
return False
await sess.set_plan_mode(True)
_CONSOLE.print(
"[bold yellow]plan-mode ON[/] — write_file / edit_file / "
"execute / task tools are blocked. Use /approve to leave, "
"or /reject to discard the plan."
)
return False
async def _approve(_: SlashParsed) -> bool:
if not sess.plan_mode:
_CONSOLE.print("[yellow]plan-mode is not active.[/]")
return False
await sess.set_plan_mode(False)
_CONSOLE.print("[green]plan approved → leaving plan-mode (writes re-enabled).[/]")
return False
async def _reject(_: SlashParsed) -> bool:
if not sess.plan_mode:
_CONSOLE.print("[yellow]plan-mode is not active.[/]")
return False
await sess.set_plan_mode(False)
sess.clear_agent_cache() # drop the plan thread entirely
_CONSOLE.print("[red]plan rejected → fresh thread, writes re-enabled.[/]")
return False
reg.register("plan", _plan, help="enter plan-mode (block writes until /approve)")
reg.register("approve", _approve, help="leave plan-mode, allow writes")
reg.register("reject", _reject, help="leave plan-mode, discard plan thread")
def _register_slash(reg: SlashRegistry, sess: InteractiveSession) -> None:
_register_navigation_slash(reg, sess)
_register_persona_slash(reg, sess)
@@ -603,6 +669,7 @@ def _register_slash(reg: SlashRegistry, sess: InteractiveSession) -> None:
_register_compaction_slash(reg, sess)
_register_memory_slash(reg, sess)
_register_skills_slash(reg, sess)
_register_plan_mode_slash(reg, sess)
def _completer(personas: list[Persona], slash_names: list[str]) -> WordCompleter:
@@ -818,9 +885,12 @@ async def _interactive_loop_async(
sess._thread_suffix = 0
# Now persist the session row (or load existing).
await _load_or_create_session_row(
row = await _load_or_create_session_row(
db, session_id, sess.persona, Path.cwd(), create=creating
)
# v0.3 PR #5: restore plan_mode flag from row on resume so the
# session remembers it across REPL restarts.
sess._plan_mode = bool(row.plan_mode)
reg = SlashRegistry()
_register_slash(reg, sess)

View File

@@ -0,0 +1,106 @@
"""PlanModeMiddleware (v0.3 PR #5) — block write tools when plan-mode is active.
Claude Code's plan mode lets the user say "design this, don't write code" — the
agent can read, search, plan via `write_todos`, but cannot mutate the
filesystem or run shell commands until the user `/approve`s.
Implementation strategy:
- A callable ``is_active()`` is passed in at construction time. The REPL flips
a flag on/off via slash commands; the middleware re-reads on every tool call.
This avoids rebuilding the agent on every `/plan` / `/approve` toggle.
- When plan-mode is on and the LLM calls a blocked tool, we return a synthetic
``ToolMessage(status="error", ...)`` so the LLM sees feedback and can adjust
("ok, I'll keep planning instead"). We do NOT raise — that would crash the
turn and the user would lose the partial response.
Blocked tools (matches Claude Code):
- ``write_file``, ``edit_file`` — fs mutation
- ``bash`` / ``execute`` / ``run_command`` / ``shell`` — shell exec
- ``task`` — sub-agent spawn (a sub-agent could bypass plan mode)
- ``write_todos`` is allowed — plan mode IS planning, todos are the artifact
"""
from __future__ import annotations
from collections.abc import Callable
from typing import Any
from langchain.agents.middleware import AgentMiddleware
from langchain_core.messages import ToolMessage
#: Tool names that mutate the filesystem.
_FS_WRITE_TOOLS: frozenset[str] = frozenset({"write_file", "edit_file"})
#: Tool names that execute shell commands.
_SHELL_TOOLS: frozenset[str] = frozenset({"bash", "execute", "run_command", "shell"})
#: Tool names that spawn sub-agents (which would bypass plan mode in the parent).
_SUBAGENT_TOOLS: frozenset[str] = frozenset({"task"})
#: Full blocklist applied while plan mode is on.
BLOCKED_TOOLS_IN_PLAN_MODE: frozenset[str] = _FS_WRITE_TOOLS | _SHELL_TOOLS | _SUBAGENT_TOOLS
def _block_message(tool_name: str) -> str:
return (
f"Plan-mode is active — `{tool_name}` is blocked. "
"Keep planning with read_file / glob / grep / write_todos, "
"or ask the user to `/approve` to leave plan mode."
)
class PlanModeMiddleware(AgentMiddleware):
"""Block mutating tool calls while plan-mode is active.
Construction takes an ``is_active`` callable that returns the current plan
mode state. The REPL toggles this state via slash commands without
rebuilding the agent — the middleware reads it fresh per tool call.
Tools that are read-only (``read_file``, ``glob``, ``grep``, ``ls``,
``write_todos``) are allowed in plan mode unconditionally.
"""
def __init__(self, *, is_active: Callable[[], bool]) -> None:
self._is_active = is_active
async def awrap_tool_call(self, request: Any, handler: Any) -> Any:
if not self._is_active():
return await handler(request)
name = _tool_name(request)
if name in BLOCKED_TOOLS_IN_PLAN_MODE:
return ToolMessage(
content=_block_message(name),
tool_call_id=_tool_call_id(request),
name=name,
status="error",
)
return await handler(request)
def wrap_tool_call(self, request: Any, handler: Any) -> Any:
# Sync path mirrors the async one for parity (e.g. when the agent is
# invoked synchronously in unit tests). Real REPL/Web paths are async.
if not self._is_active():
return handler(request)
name = _tool_name(request)
if name in BLOCKED_TOOLS_IN_PLAN_MODE:
return ToolMessage(
content=_block_message(name),
tool_call_id=_tool_call_id(request),
name=name,
status="error",
)
return handler(request)
def _tool_name(request: Any) -> str:
tool_call = getattr(request, "tool_call", None)
if isinstance(tool_call, dict):
return str(tool_call.get("name") or "")
return str(getattr(request, "name", "") or "")
def _tool_call_id(request: Any) -> str:
tool_call = getattr(request, "tool_call", None)
if isinstance(tool_call, dict):
return str(tool_call.get("id") or "")
return str(getattr(request, "id", "") or "")