mirror of
https://github.com/hansjone/oclaw.git
synced 2026-10-09 03:30:48 +08:00
feat(scheduler): workflow recipes with confirm-before-create
Store self-contained playbook recipes on scheduled jobs, add schedule_propose draft gate, and run playbook instructions at fire time instead of only short reminders. Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
parent
ad5d1db3c6
commit
d036190195
10 changed files with 989 additions and 49 deletions
339
runtime/scheduler/recipe.py
Normal file
339
runtime/scheduler/recipe.py
Normal file
|
|
@ -0,0 +1,339 @@
|
|||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
import re
|
||||
from typing import Any
|
||||
|
||||
|
||||
COMPLEX_PROMPT_HINTS = (
|
||||
"刚才",
|
||||
"之前",
|
||||
"继续",
|
||||
"按刚才",
|
||||
"同样的",
|
||||
"那件事",
|
||||
"那套",
|
||||
"流程",
|
||||
"步骤",
|
||||
"生成",
|
||||
"报告",
|
||||
"文档",
|
||||
"pdf",
|
||||
"xlsx",
|
||||
"附件",
|
||||
"发群",
|
||||
"发给",
|
||||
"执行",
|
||||
"整理",
|
||||
"汇总",
|
||||
"拉取",
|
||||
"爬取",
|
||||
"监控",
|
||||
"same as",
|
||||
"continue",
|
||||
"as before",
|
||||
"workflow",
|
||||
"report",
|
||||
"generate",
|
||||
"attach",
|
||||
)
|
||||
|
||||
|
||||
def _as_str_list(raw: Any) -> list[str]:
|
||||
if raw is None:
|
||||
return []
|
||||
if isinstance(raw, str):
|
||||
item = raw.strip()
|
||||
return [item] if item else []
|
||||
if not isinstance(raw, list):
|
||||
return []
|
||||
out: list[str] = []
|
||||
for item in raw:
|
||||
text = str(item or "").strip()
|
||||
if text:
|
||||
out.append(text)
|
||||
return out
|
||||
|
||||
|
||||
def _as_str_dict(raw: Any) -> dict[str, str]:
|
||||
if not isinstance(raw, dict):
|
||||
return {}
|
||||
out: dict[str, str] = {}
|
||||
for key, value in raw.items():
|
||||
k = str(key or "").strip()
|
||||
if not k:
|
||||
continue
|
||||
out[k] = str(value if value is not None else "").strip()
|
||||
return out
|
||||
|
||||
|
||||
def parse_recipe_arg(raw: Any) -> dict[str, Any] | None:
|
||||
if raw is None:
|
||||
return None
|
||||
if isinstance(raw, dict):
|
||||
return raw
|
||||
if isinstance(raw, str) and raw.strip():
|
||||
try:
|
||||
data = json.loads(raw)
|
||||
return data if isinstance(data, dict) else None
|
||||
except Exception:
|
||||
return None
|
||||
return None
|
||||
|
||||
|
||||
def normalize_recipe(raw: Any) -> dict[str, Any]:
|
||||
data = parse_recipe_arg(raw) or {}
|
||||
goal = str(data.get("goal") or "").strip()
|
||||
steps = _as_str_list(data.get("steps"))
|
||||
constraints = _as_str_list(data.get("constraints"))
|
||||
success_criteria = _as_str_list(data.get("success_criteria") or data.get("successCriteria"))
|
||||
inputs_raw = data.get("inputs") if isinstance(data.get("inputs"), dict) else {}
|
||||
constants = _as_str_dict(inputs_raw.get("constants"))
|
||||
from_context = _as_str_list(inputs_raw.get("from_context") or inputs_raw.get("fromContext"))
|
||||
output_raw = data.get("output") if isinstance(data.get("output"), dict) else {}
|
||||
source_raw = data.get("source") if isinstance(data.get("source"), dict) else {}
|
||||
version = int(data.get("version") or 1)
|
||||
recipe: dict[str, Any] = {
|
||||
"version": max(1, version),
|
||||
"goal": goal,
|
||||
"steps": steps,
|
||||
"constraints": constraints,
|
||||
"success_criteria": success_criteria,
|
||||
"inputs": {
|
||||
"constants": constants,
|
||||
"from_context": from_context,
|
||||
},
|
||||
"output": {
|
||||
"style": str(output_raw.get("style") or "channel_update").strip() or "channel_update",
|
||||
"need_attachments": bool(output_raw.get("need_attachments") or output_raw.get("needAttachments")),
|
||||
},
|
||||
"source": {
|
||||
"session_id": str(source_raw.get("session_id") or source_raw.get("sessionId") or "").strip(),
|
||||
"compiled_at": str(source_raw.get("compiled_at") or source_raw.get("compiledAt") or "").strip(),
|
||||
},
|
||||
}
|
||||
return recipe
|
||||
|
||||
|
||||
def recipe_is_empty(recipe: dict[str, Any] | None) -> bool:
|
||||
if not recipe:
|
||||
return True
|
||||
norm = normalize_recipe(recipe)
|
||||
return not (norm.get("goal") or norm.get("steps") or norm.get("success_criteria") or norm.get("constraints"))
|
||||
|
||||
|
||||
def recipe_has_playbook(recipe: dict[str, Any] | None) -> bool:
|
||||
if not recipe:
|
||||
return False
|
||||
norm = normalize_recipe(recipe)
|
||||
steps = list(norm.get("steps") or [])
|
||||
goal = str(norm.get("goal") or "").strip()
|
||||
return bool(goal) and len(steps) >= 2
|
||||
|
||||
|
||||
def recipe_missing_fields(recipe: dict[str, Any] | None) -> list[str]:
|
||||
norm = normalize_recipe(recipe or {})
|
||||
missing: list[str] = []
|
||||
if not str(norm.get("goal") or "").strip():
|
||||
missing.append("goal")
|
||||
steps = list(norm.get("steps") or [])
|
||||
if len(steps) < 2:
|
||||
missing.append("steps")
|
||||
if not list(norm.get("success_criteria") or []):
|
||||
missing.append("success_criteria")
|
||||
return missing
|
||||
|
||||
|
||||
def looks_like_complex_schedule_prompt(prompt_text: str, *, recipe: dict[str, Any] | None = None) -> bool:
|
||||
"""Heuristic: multi-step / referential prompts need a recipe."""
|
||||
if recipe_has_playbook(recipe):
|
||||
return True
|
||||
text = str(prompt_text or "").strip()
|
||||
if not text:
|
||||
return False
|
||||
low = text.lower()
|
||||
if any(hint in low or hint in text for hint in COMPLEX_PROMPT_HINTS):
|
||||
# Short pure reminders like "提醒喝水" should stay simple.
|
||||
if len(text) <= 12 and ("提醒" in text or "remind" in low) and "步骤" not in text:
|
||||
return False
|
||||
return True
|
||||
if len(text) >= 80:
|
||||
return True
|
||||
if text.count("\n") >= 2:
|
||||
return True
|
||||
if re.search(r"(1[\.\)]|第一步|step\s*1)", text, flags=re.I):
|
||||
return True
|
||||
return False
|
||||
|
||||
|
||||
def prompt_summary_from_recipe(recipe: dict[str, Any] | None, *, fallback: str = "") -> str:
|
||||
norm = normalize_recipe(recipe or {})
|
||||
goal = str(norm.get("goal") or "").strip()
|
||||
if goal:
|
||||
return goal
|
||||
steps = list(norm.get("steps") or [])
|
||||
if steps:
|
||||
return steps[0]
|
||||
return str(fallback or "").strip()
|
||||
|
||||
|
||||
def preview_markdown(
|
||||
*,
|
||||
name: str,
|
||||
schedule_kind: str,
|
||||
schedule_expr: str,
|
||||
timezone_name: str,
|
||||
recipe: dict[str, Any],
|
||||
lang: str = "zh",
|
||||
) -> str:
|
||||
norm = normalize_recipe(recipe)
|
||||
is_en = str(lang or "").lower().startswith("en")
|
||||
lines: list[str] = []
|
||||
if is_en:
|
||||
lines.append("## Scheduled workflow draft (confirm before create)")
|
||||
lines.append(f"- **Name**: {name or '(untitled)'}")
|
||||
lines.append(f"- **Schedule**: `{schedule_kind}` `{schedule_expr}` ({timezone_name or 'system'})")
|
||||
lines.append(f"- **Goal**: {norm.get('goal') or '(missing)'}")
|
||||
lines.append("### Steps")
|
||||
for i, step in enumerate(list(norm.get("steps") or []), start=1):
|
||||
lines.append(f"{i}. {step}")
|
||||
constraints = list(norm.get("constraints") or [])
|
||||
if constraints:
|
||||
lines.append("### Constraints")
|
||||
for item in constraints:
|
||||
lines.append(f"- {item}")
|
||||
criteria = list(norm.get("success_criteria") or [])
|
||||
if criteria:
|
||||
lines.append("### Success criteria")
|
||||
for item in criteria:
|
||||
lines.append(f"- {item}")
|
||||
constants = dict((norm.get("inputs") or {}).get("constants") or {})
|
||||
if constants:
|
||||
lines.append("### Fixed inputs")
|
||||
for key, value in constants.items():
|
||||
lines.append(f"- `{key}`: {value}")
|
||||
lines.append("")
|
||||
lines.append("Reply **confirm** to create, or list the edits you want.")
|
||||
else:
|
||||
lines.append("## 定时工作流草稿（请确认后再创建）")
|
||||
lines.append(f"- **名称**：{name or '（未命名）'}")
|
||||
lines.append(f"- **时间**：`{schedule_kind}` `{schedule_expr}`（{timezone_name or '系统时区'}）")
|
||||
lines.append(f"- **目标**：{norm.get('goal') or '（缺失）'}")
|
||||
lines.append("### 步骤")
|
||||
for i, step in enumerate(list(norm.get("steps") or []), start=1):
|
||||
lines.append(f"{i}. {step}")
|
||||
constraints = list(norm.get("constraints") or [])
|
||||
if constraints:
|
||||
lines.append("### 约束")
|
||||
for item in constraints:
|
||||
lines.append(f"- {item}")
|
||||
criteria = list(norm.get("success_criteria") or [])
|
||||
if criteria:
|
||||
lines.append("### 成功标准")
|
||||
for item in criteria:
|
||||
lines.append(f"- {item}")
|
||||
constants = dict((norm.get("inputs") or {}).get("constants") or {})
|
||||
if constants:
|
||||
lines.append("### 固定输入")
|
||||
for key, value in constants.items():
|
||||
lines.append(f"- `{key}`：{value}")
|
||||
lines.append("")
|
||||
lines.append("请回复**确认**以创建，或说明要修改的地方。")
|
||||
return "\n".join(lines).strip()
|
||||
|
||||
|
||||
def compile_playbook_instruction(*, recipe: dict[str, Any], lang: str = "zh") -> str:
|
||||
norm = normalize_recipe(recipe)
|
||||
is_en = str(lang or "").lower().startswith("en")
|
||||
steps = list(norm.get("steps") or [])
|
||||
constraints = list(norm.get("constraints") or [])
|
||||
criteria = list(norm.get("success_criteria") or [])
|
||||
constants = dict((norm.get("inputs") or {}).get("constants") or {})
|
||||
from_context = list((norm.get("inputs") or {}).get("from_context") or [])
|
||||
need_attachments = bool((norm.get("output") or {}).get("need_attachments"))
|
||||
|
||||
if is_en:
|
||||
lines = [
|
||||
"[Scheduled playbook — internal instruction, not a user message]",
|
||||
f"Goal: {norm.get('goal') or '(unspecified)'}",
|
||||
"Execute this recurring playbook end-to-end. Use tools as needed.",
|
||||
"Do not reply with only a short reminder unless the playbook is truly reminder-only.",
|
||||
"Steps:",
|
||||
]
|
||||
for i, step in enumerate(steps, start=1):
|
||||
lines.append(f"{i}. {step}")
|
||||
if constraints:
|
||||
lines.append("Constraints:")
|
||||
lines.extend(f"- {c}" for c in constraints)
|
||||
if criteria:
|
||||
lines.append("Success criteria:")
|
||||
lines.extend(f"- {c}" for c in criteria)
|
||||
if constants:
|
||||
lines.append("Fixed inputs:")
|
||||
lines.extend(f"- {k}: {v}" for k, v in constants.items())
|
||||
if from_context:
|
||||
lines.append("Pull from context when needed:")
|
||||
lines.extend(f"- {item}" for item in from_context)
|
||||
if need_attachments:
|
||||
lines.append(
|
||||
"If files are produced, call save_deliverable_attachment so channel delivery includes them."
|
||||
)
|
||||
lines.append("Deliver a useful channel update that reflects completed work.")
|
||||
return "\n".join(lines)
|
||||
|
||||
lines = [
|
||||
"【定时工作流·内部指令，不是用户发言】",
|
||||
f"目标：{norm.get('goal') or '（未指定）'}",
|
||||
"请按下方 playbook 完整执行本轮定时任务；按需调用工具。",
|
||||
"除非任务本身只是提醒，否则不要只回一句短提醒。",
|
||||
"步骤：",
|
||||
]
|
||||
for i, step in enumerate(steps, start=1):
|
||||
lines.append(f"{i}. {step}")
|
||||
if constraints:
|
||||
lines.append("约束：")
|
||||
lines.extend(f"- {c}" for c in constraints)
|
||||
if criteria:
|
||||
lines.append("成功标准：")
|
||||
lines.extend(f"- {c}" for c in criteria)
|
||||
if constants:
|
||||
lines.append("固定输入：")
|
||||
lines.extend(f"- {k}：{v}" for k, v in constants.items())
|
||||
if from_context:
|
||||
lines.append("需要时从上下文获取：")
|
||||
lines.extend(f"- {item}" for item in from_context)
|
||||
if need_attachments:
|
||||
lines.append("若产生文件，必须调用 save_deliverable_attachment，渠道才会随消息发送附件。")
|
||||
lines.append("完成后向渠道发送能体现已完成工作的更新消息。")
|
||||
return "\n".join(lines)
|
||||
|
||||
|
||||
def load_recipe_from_job(job: Any) -> dict[str, Any]:
|
||||
raw = getattr(job, "recipe_json", None)
|
||||
if raw is None and isinstance(job, dict):
|
||||
raw = job.get("recipe_json") or job.get("recipe")
|
||||
if isinstance(raw, dict):
|
||||
return normalize_recipe(raw)
|
||||
text = str(raw or "").strip()
|
||||
if not text:
|
||||
return normalize_recipe({})
|
||||
try:
|
||||
data = json.loads(text)
|
||||
except Exception:
|
||||
return normalize_recipe({})
|
||||
return normalize_recipe(data if isinstance(data, dict) else {})
|
||||
|
||||
|
||||
__all__ = [
|
||||
"COMPLEX_PROMPT_HINTS",
|
||||
"compile_playbook_instruction",
|
||||
"load_recipe_from_job",
|
||||
"looks_like_complex_schedule_prompt",
|
||||
"normalize_recipe",
|
||||
"parse_recipe_arg",
|
||||
"preview_markdown",
|
||||
"prompt_summary_from_recipe",
|
||||
"recipe_has_playbook",
|
||||
"recipe_is_empty",
|
||||
"recipe_missing_fields",
|
||||
]
|
||||
|
|
@ -7,6 +7,7 @@ import uuid
|
|||
from datetime import datetime, timezone
|
||||
from typing import Any
|
||||
|
||||
from runtime.scheduler.recipe import load_recipe_from_job, recipe_has_playbook
|
||||
from runtime.scheduler.session_resolver import resolve_scheduled_session, resolve_scheduled_viewer_username
|
||||
from runtime.scheduler.turn_text import build_scheduled_turn_instruction
|
||||
from runtime.worker import ensure_worker_started
|
||||
|
|
@ -76,7 +77,14 @@ def enqueue_scheduled_job_run(
|
|||
agent_run_id = uuid.uuid4().hex
|
||||
prompt_text = str(getattr(job, "prompt_text", "") or "").strip()
|
||||
lang = str(getattr(job, "lang", "") or "zh")
|
||||
user_text = build_scheduled_turn_instruction(prompt_text=prompt_text, mode=mode, lang=lang)
|
||||
recipe = load_recipe_from_job(job)
|
||||
playbook = recipe_has_playbook(recipe)
|
||||
user_text = build_scheduled_turn_instruction(
|
||||
prompt_text=prompt_text,
|
||||
mode=mode,
|
||||
lang=lang,
|
||||
recipe=recipe if playbook else None,
|
||||
)
|
||||
viewer_username = resolve_scheduled_viewer_username(
|
||||
store,
|
||||
tenant_id=tenant_id,
|
||||
|
|
@ -95,6 +103,7 @@ def enqueue_scheduled_job_run(
|
|||
"lang": lang,
|
||||
"text": user_text,
|
||||
"prompt_text": prompt_text,
|
||||
"recipe": recipe if playbook else {},
|
||||
"attachments": [],
|
||||
"metadata": {
|
||||
"scheduled_job_id": job_id,
|
||||
|
|
@ -103,6 +112,7 @@ def enqueue_scheduled_job_run(
|
|||
"selected_specialist": str(getattr(job, "specialist", "") or "generalist"),
|
||||
"scheduled_mode": mode,
|
||||
"scheduled_proactive": True,
|
||||
"scheduled_playbook": playbook,
|
||||
},
|
||||
"interaction_mode": str(getattr(job, "interaction_mode", "") or "expert"),
|
||||
"requested_specialist": str(getattr(job, "specialist", "") or "generalist"),
|
||||
|
|
|
|||
|
|
@ -1,5 +1,9 @@
|
|||
from __future__ import annotations
|
||||
|
||||
from typing import Any
|
||||
|
||||
from runtime.scheduler.recipe import compile_playbook_instruction, recipe_has_playbook
|
||||
|
||||
|
||||
def format_scheduled_user_reminder(prompt_text: str) -> str:
|
||||
body = str(prompt_text or "").strip()
|
||||
|
|
@ -10,10 +14,19 @@ def format_scheduled_user_reminder(prompt_text: str) -> str:
|
|||
return f"⏰ 提醒：{body}"
|
||||
|
||||
|
||||
def build_scheduled_turn_instruction(*, prompt_text: str, mode: str, lang: str) -> str:
|
||||
"""Internal LLM instruction for proactive scheduled reminders (not user-facing)."""
|
||||
intent = str(prompt_text or "").strip()
|
||||
def build_scheduled_turn_instruction(
|
||||
*,
|
||||
prompt_text: str,
|
||||
mode: str,
|
||||
lang: str,
|
||||
recipe: dict[str, Any] | None = None,
|
||||
) -> str:
|
||||
"""Internal LLM instruction for proactive scheduled reminders/playbooks (not user-facing)."""
|
||||
_ = str(mode or "scheduled").strip()
|
||||
if recipe_has_playbook(recipe):
|
||||
return compile_playbook_instruction(recipe=recipe or {}, lang=lang)
|
||||
|
||||
intent = str(prompt_text or "").strip()
|
||||
is_en = str(lang or "").lower().startswith("en")
|
||||
if is_en:
|
||||
return (
|
||||
|
|
@ -30,8 +43,21 @@ def build_scheduled_turn_instruction(*, prompt_text: str, mode: str, lang: str)
|
|||
)
|
||||
|
||||
|
||||
def scheduled_turn_system_suffix(*, lang: str) -> str:
|
||||
def scheduled_turn_system_suffix(*, lang: str, playbook: bool = False) -> str:
|
||||
is_en = str(lang or "").lower().startswith("en")
|
||||
if playbook:
|
||||
if is_en:
|
||||
return (
|
||||
"\n\n[Scheduled playbook mode] You are executing a recurring workflow for the user. "
|
||||
"Follow the playbook steps, use tools as needed, and deliver a useful update "
|
||||
"(including save_deliverable_attachment for generated files). "
|
||||
"Do not pretend the user just messaged you."
|
||||
)
|
||||
return (
|
||||
"\n\n【定时工作流模式】你正在执行周期性工作流。"
|
||||
"按 playbook 步骤完成任务，按需调用工具；若生成文件须 save_deliverable_attachment。"
|
||||
"不要假装用户刚刚发了消息，不要只回一句空提醒。"
|
||||
)
|
||||
if is_en:
|
||||
return (
|
||||
"\n\n[Scheduled job mode] You are sending a proactive reminder to the user. "
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue