mirror of
https://github.com/ThearchyHelios/AgentLab.git
synced 2026-11-15 03:41:58 +00:00
ThearchyHelios
5b4bd91476
为什么：问数据的答案要和已治理口径接起来（next-round 02 做法 2、3、9、10；总方案 X1、X2、X3、X12，拍板 L2、L4、L8）。
- 比对（ledger/match.py）：探索运行（chat、canvas、trial、personal_app）成功结束、封存完、有查询快照时，在 runner 封存之后和
notify_run 同一个位置放到后台；候选口径是每个工作流最新的已发布未退役版本里查库的调用工具节点（受管、已发布都算），按数据源
和表预筛；逐个用 match_template 比，schema 用查询当时冻结的表结构、列名取同一份快照里冻结的数据目录。每次运行限时 8 秒、在
线程里、同时最多 2 个；超时、快照被改过或读不出、SQL 解析不了照实记 skipped。启动时（run_manager.setup，和调度循环同一个开关）
补比最近 7 天封存了却没比过的运行，只读工件和版本表。
- 记法（ledger/store.py，拍板 L8）：内容寻址工件 kind=caliber_match（agentlab.caliber_match/1）+ run_append 的 sealed 方式追加
ledger.matched（core/events.py 登记）；工件里不放时间和工作流名，同一份封存再比一次哈希一样；meta 记 calibers / related / counts。
- 标签（ledger/labels.py）：单元格按查询的比对状态；指标要每个输入都等价于同一份口径同一版、绑定一致，式子按「哪个取数节点、
第几行、哪一列」换完和口径卡上的一字不差。caliber（how：independent 等价已证明；composed 助手选用留给第 3 波）/ related /
reasoning / unknown；整题判断附还缺的入口字段。
- 读（ledger/view.py）：GET /api/runs/{id}/ledger（ledger_api.runs_router，和证据同一道门），没比对过、能比的按需排上；证据
片段接口加 ledger 键（没比对过的运行没有这个键，返回一字不差）；报告导出的面板最上面多「口径」一步并写明「封存后比对，不在
封存范围内」。遮罩按读的那一刻重滤：绑定的值是这次运行查到过的遮罩值时不给值，差异条目和逐行对照走 _Shown。
- 来源链：POST /api/runs 加 origin_ref（只在 origin=chat 时收，服务端核对对话访问权和轮次归属，写进 Run.origin 和封存的
run.origin）；WorkflowIn 加 origin（run_id、conversation_id、turn_id，核对规则同 from-run）；报告导出和可验证报告下载各记一条
审计 report.exported。可验证报告列封存后追加的事件时给 ledger.matched 一句只有条数的说明。
- 统计（ledger/stats.py）：E1–E4 从工件、运行、查询快照、审计现算，GET /api/ledger/stats（搭建者）；E3 用 app/chat/questions
的聚类加按（数据源、normalize_sql 文本）的确定性重复；评测直接调 compute_stats（可点名 run_ids）。对账反馈、认证转化、按发布
的召回在后面几波，照实写没有数据。
测试：test_ledger_match.py（真跑探索运行：等价、相关、没有相关口径、写死的码值、非探索运行、超时、快照被改、解析失败、预筛、
指标两种、级别、只认最新版、补比）、test_ledger_api.py（账本接口、按需比、遮罩、证据片段、导出、审计、统计）、
test_run_origin_ref.py（origin_ref 与画布来源的核对）。
778 lines
38 KiB
Python
778 lines
38 KiB
Python
from __future__ import annotations
|
||
|
||
from typing import Any
|
||
|
||
from fastapi import APIRouter, Depends, Header, HTTPException
|
||
from pydantic import BaseModel, Field
|
||
from sqlalchemy import func, select
|
||
from sqlalchemy.ext.asyncio import AsyncSession
|
||
from sqlalchemy.orm import aliased
|
||
|
||
from app.core.errors import graph_error
|
||
from app.api.coded import CodedHTTPException
|
||
from app.api.runs import actor_of
|
||
from app.core.artifact_store import graph_hash
|
||
from app.data.catalog_impact import catalog_drift, catalog_versions_for, recorded_versions
|
||
from app.db.base import get_session
|
||
from app.db.models import Run, Workflow, WorkflowVersion
|
||
from app.engine.governance import publish_issues
|
||
from app.engine.lint_process import load_subgraphs, with_subgraphs
|
||
from app.engine.provider_refs import CODES as PROVIDER_CODES, load_issues, load_providers, quiet_unset_model
|
||
from app.engine.schema import GraphSpec, NodeType, validate_graph
|
||
|
||
router = APIRouter(prefix="/api/workflows", tags=["workflows"])
|
||
|
||
_GONE = "工作流不存在，可能已被删除"
|
||
#: 删除被拒：还有应用固定着、或者发布过的版本有生效记录（F4）
|
||
WORKFLOW_IN_USE = "workflow_in_use"
|
||
#: 受管级发布、版本退役只给管理员（Q8）
|
||
ADMIN_ONLY = "admin_only"
|
||
_LEVELS = ("published", "governed")
|
||
_BAD_LEVEL = "发布级别只能是「已发布」或「受管」"
|
||
#: 剩下的挡住发布的只有模型接入的问题、人又要交给助手时的说明：换成哪个接入要人拿主意，助手处理不了。发布弹窗里
|
||
#: 渲染成「助手：……」，说明里不再带冒号；可能是全图默认、也可能只写了模型而提供它的接入停用了，不说「节点」「点名」。
|
||
#: 人刚点了「交给助手」，写「助手无法处理」，不写「无法交给助手处理」（读来像没交出去）
|
||
PROVIDER_LEFT = ("剩余的问题是所用的模型接入不存在或已停用，助手无法处理。"
|
||
"请在「设置 → 模型接入」中核对名称或启用相应的接入")
|
||
#: 剩下的问题里有一条带着修复（provider_refs 按候选算得出来）才补这半句：一个候选都落不稳时那里什么都不给。弹窗里那一组
|
||
#: 选项在每条问题下方，标题是「选择……改用的模型接入」；另起半句，不和前面的「或」连用
|
||
PROVIDER_LEFT_FIX = "；也可在各条问题下方选择改用的模型接入"
|
||
|
||
|
||
class WorkflowOrigin(BaseModel):
|
||
"""新工作流出自哪次运行、哪段对话、哪一轮（next-round 02 做法 9）：问数据的「在画布里打开」带上它，和从运行提取（/copilot/
|
||
from-run）记成同一个形状 {run_id, conversation_id?, turn_id?}。「口径 → 来源问题」、认证队列的「在画布里打开过」都读它。"""
|
||
|
||
run_id: str = Field(min_length=1, max_length=32)
|
||
conversation_id: str | None = Field(default=None, max_length=32)
|
||
turn_id: str | None = Field(default=None, max_length=32)
|
||
|
||
|
||
class WorkflowIn(BaseModel):
|
||
name: str = Field(min_length=1, max_length=200)
|
||
description: str = ""
|
||
graph: dict[str, Any] = Field(default_factory=dict)
|
||
tags: list[str] = Field(default_factory=list)
|
||
#: 来源（WorkflowOrigin）：服务端核对运行读得到、对话用得了、这一轮属于这段对话，才记进 workflows.origin
|
||
origin: WorkflowOrigin | None = None
|
||
|
||
|
||
class WorkflowPatch(BaseModel):
|
||
name: str | None = None
|
||
description: str | None = None
|
||
graph: dict[str, Any] | None = None
|
||
tags: list[str] | None = None
|
||
note: str = ""
|
||
|
||
|
||
class WorkflowOut(BaseModel):
|
||
id: str
|
||
name: str
|
||
description: str
|
||
graph: dict[str, Any]
|
||
tags: list[str]
|
||
version: int
|
||
is_template: bool
|
||
status: str = "draft"
|
||
published_version: int | None = None
|
||
published_by: str | None = None
|
||
#: 当前发布版本的级别（published / governed）。status 在画布上改过之后会回到 draft，「这是不是受管工作流」看这一项
|
||
published_level: str | None = None
|
||
#: 画布这一版的图和当前发布版的图内容相同（按 graph_hash 比）：撤回改动后保存出来的新版本号更大，内容却和发布版
|
||
#: 一样，页头照实写「与发布版内容相同」，不写「领先 N 版」。没发布过、老数据没记哈希的为 false
|
||
same_as_published: bool = False
|
||
created_at: Any = None
|
||
updated_at: Any = None
|
||
run_count: int = 0
|
||
|
||
model_config = {"from_attributes": True}
|
||
|
||
|
||
async def _published_levels(session: AsyncSession, ids: list[str] | None = None) -> dict[str, str]:
|
||
"""各工作流当前发布版本的级别 {工作流 id: level}。没发布过、或老数据没记级别的不在里面。"""
|
||
stmt = select(WorkflowVersion.workflow_id, WorkflowVersion.level).join(
|
||
Workflow, (Workflow.id == WorkflowVersion.workflow_id) & (Workflow.published_version == WorkflowVersion.version))
|
||
if ids is not None:
|
||
stmt = stmt.where(WorkflowVersion.workflow_id.in_(ids))
|
||
return {wid: level for wid, level in (await session.execute(stmt)).all() if level}
|
||
|
||
|
||
async def _same_as_published(session: AsyncSession, ids: list[str] | None = None) -> set[str]:
|
||
"""画布这一版（Workflow.version）的图哈希和当前发布版相同的工作流 id。两边哈希都记着才比，不在前端比图。"""
|
||
current, published = aliased(WorkflowVersion), aliased(WorkflowVersion)
|
||
stmt = select(Workflow.id, current.graph_hash, published.graph_hash).join(
|
||
current, (current.workflow_id == Workflow.id) & (current.version == Workflow.version)).join(
|
||
published, (published.workflow_id == Workflow.id) & (published.version == Workflow.published_version))
|
||
if ids is not None:
|
||
stmt = stmt.where(Workflow.id.in_(ids))
|
||
return {wid for wid, mine, theirs in (await session.execute(stmt)).all() if mine and mine == theirs}
|
||
|
||
|
||
async def _run_counts(session: AsyncSession, ids: list[str] | None = None) -> dict[str, int]:
|
||
"""各工作流的运行次数 {工作流 id: 次数}，ids 为 None 时算全部。列表和单个工作流的接口都按它算：以前只有列表
|
||
计数，详情、保存、恢复的返回落在 WorkflowOut 的默认值 0 上，同一个工作流两处说的次数不一样。"""
|
||
stmt = select(Run.workflow_id, func.count(Run.id)).group_by(Run.workflow_id)
|
||
if ids is not None:
|
||
stmt = stmt.where(Run.workflow_id.in_(ids))
|
||
return dict((await session.execute(stmt)).all())
|
||
|
||
|
||
async def _workflow_out(session: AsyncSession, workflow: Workflow) -> WorkflowOut:
|
||
"""返回单个工作流的接口都走这里，run_count 和列表一致。"""
|
||
item = WorkflowOut.model_validate(workflow)
|
||
item.run_count = (await _run_counts(session, [workflow.id])).get(workflow.id, 0)
|
||
item.published_level = (await _published_levels(session, [workflow.id])).get(workflow.id)
|
||
item.same_as_published = workflow.id in await _same_as_published(session, [workflow.id])
|
||
return item
|
||
|
||
|
||
@router.get("", response_model=list[WorkflowOut])
|
||
async def list_workflows(session: AsyncSession = Depends(get_session)) -> list[WorkflowOut]:
|
||
rows = list(
|
||
(await session.execute(select(Workflow).order_by(Workflow.updated_at.desc()))).scalars()
|
||
)
|
||
counts = await _run_counts(session)
|
||
levels = await _published_levels(session)
|
||
same = await _same_as_published(session)
|
||
out: list[WorkflowOut] = []
|
||
for row in rows:
|
||
item = WorkflowOut.model_validate(row)
|
||
item.run_count = counts.get(row.id, 0)
|
||
item.published_level = levels.get(row.id)
|
||
item.same_as_published = row.id in same
|
||
out.append(item)
|
||
return out
|
||
|
||
|
||
@router.post("", response_model=WorkflowOut, status_code=201)
|
||
async def create_workflow(
|
||
payload: WorkflowIn, session: AsyncSession = Depends(get_session)
|
||
) -> WorkflowOut:
|
||
origin = await _checked_origin(session, payload.origin) if payload.origin is not None else None
|
||
workflow = Workflow(
|
||
name=payload.name,
|
||
description=payload.description,
|
||
graph=payload.graph,
|
||
tags=payload.tags,
|
||
origin=origin,
|
||
)
|
||
session.add(workflow)
|
||
await session.flush()
|
||
session.add(
|
||
WorkflowVersion(workflow_id=workflow.id, version=1, graph=payload.graph,
|
||
graph_hash=graph_hash(payload.graph), note="初始版本")
|
||
)
|
||
await session.commit()
|
||
await session.refresh(workflow)
|
||
return await _workflow_out(session, workflow)
|
||
|
||
|
||
#: 新工作流的来源核对不过（422）：运行读不到、这一轮不属于这段对话
|
||
WORKFLOW_ORIGIN_INVALID = "workflow_origin_invalid"
|
||
|
||
|
||
async def _checked_origin(session: AsyncSession, origin: WorkflowOrigin) -> dict[str, str]:
|
||
"""核对新工作流的来源，规则同从运行提取（/copilot/from-run）：运行这个人读得到（读不到的当它不存在）；对话这个人用得了
|
||
（不能用的 403）；给了轮次的，这一轮要是这段对话的。对话不存在的照 from-run 记下 id（对话删了，来源仍是那次运行）。"""
|
||
from app.api.deps import can_read_run, current_principal, ensure_conversation_access
|
||
from app.db.models import ConversationTurn
|
||
|
||
run = await session.get(Run, origin.run_id)
|
||
if run is None or not await can_read_run(current_principal(), run):
|
||
raise CodedHTTPException(422, "来源的运行不存在，可能已被删除", WORKFLOW_ORIGIN_INVALID, field="origin")
|
||
conversation = await ensure_conversation_access(session, origin.conversation_id)
|
||
if origin.turn_id:
|
||
turn = await session.get(ConversationTurn, origin.turn_id)
|
||
if turn is None or not origin.conversation_id or turn.conversation_id != origin.conversation_id:
|
||
raise CodedHTTPException(422, "来源的这一轮不属于这段对话", WORKFLOW_ORIGIN_INVALID, field="origin")
|
||
return {"run_id": run.id,
|
||
**({"conversation_id": conversation.id if conversation else origin.conversation_id}
|
||
if origin.conversation_id else {}),
|
||
**({"turn_id": origin.turn_id} if origin.turn_id else {})}
|
||
|
||
|
||
@router.get("/{workflow_id}", response_model=WorkflowOut)
|
||
async def get_workflow(
|
||
workflow_id: str, session: AsyncSession = Depends(get_session)
|
||
) -> WorkflowOut:
|
||
workflow = await session.get(Workflow, workflow_id)
|
||
if not workflow:
|
||
raise HTTPException(404, _GONE)
|
||
return await _workflow_out(session, workflow)
|
||
|
||
|
||
@router.patch("/{workflow_id}", response_model=WorkflowOut)
|
||
async def update_workflow(
|
||
workflow_id: str, payload: WorkflowPatch, session: AsyncSession = Depends(get_session)
|
||
) -> WorkflowOut:
|
||
workflow = await session.get(Workflow, workflow_id)
|
||
if not workflow:
|
||
raise HTTPException(404, _GONE)
|
||
|
||
if payload.name is not None:
|
||
workflow.name = payload.name
|
||
if payload.description is not None:
|
||
workflow.description = payload.description
|
||
if payload.tags is not None:
|
||
workflow.tags = payload.tags
|
||
if payload.graph is not None and payload.graph != workflow.graph:
|
||
# 图有实际变化才留快照，避免改个名字就堆一堆版本
|
||
workflow.graph = payload.graph
|
||
workflow.version += 1
|
||
session.add(
|
||
WorkflowVersion(
|
||
workflow_id=workflow.id,
|
||
version=workflow.version,
|
||
graph=payload.graph,
|
||
graph_hash=graph_hash(payload.graph),
|
||
note=payload.note,
|
||
)
|
||
)
|
||
# status 说的是**当前画布**的状态，不是"这个工作流曾经发布过"。
|
||
# 改完图还挂着 governed，界面上就会显示一张从未过闸的图是受管模板——
|
||
# 已发布的那一版仍由 published_version 指着，formal 运行不受影响。
|
||
if workflow.status in ("published", "governed"):
|
||
workflow.status = "draft"
|
||
await session.commit()
|
||
await session.refresh(workflow)
|
||
return await _workflow_out(session, workflow)
|
||
|
||
|
||
@router.delete("/{workflow_id}", status_code=204)
|
||
async def delete_workflow(
|
||
workflow_id: str, session: AsyncSession = Depends(get_session)
|
||
) -> None:
|
||
workflow = await session.get(Workflow, workflow_id)
|
||
if not workflow:
|
||
raise HTTPException(404, _GONE)
|
||
# 运行记录是 ON DELETE SET NULL：删掉工作流，已结束的运行留着、照样能核对封存；
|
||
# 还在跑的那次会失去归属，再也回不到它的画布，所以和删除运行记录一样先停再删
|
||
live = await session.scalar(
|
||
select(func.count()).select_from(Run).where(
|
||
Run.workflow_id == workflow_id, Run.status.in_(("queued", "running"))
|
||
)
|
||
)
|
||
if live:
|
||
raise HTTPException(409, f"该工作流还有 {live} 次运行正在进行，现在删除会使这些运行失去所属工作流。请先停止运行，再删除")
|
||
# 版本随工作流级联删除（F4）：应用固定着的版本、带生效记录的版本删掉，应用发起不了、生效记录对不上承诺。这两种拒绝
|
||
if reasons := await _delete_blockers(session, workflow_id):
|
||
raise CodedHTTPException(409, "该工作流还在使用中，不能删除：" + "；".join(reasons), WORKFLOW_IN_USE)
|
||
await session.delete(workflow)
|
||
await session.commit()
|
||
|
||
|
||
async def _delete_blockers(session: AsyncSession, workflow_id: str) -> list[str]:
|
||
"""拒绝删除的原因：应用固定着；某个版本有生效记录（演练不算）。
|
||
|
||
生效记录按全部版本数，不只看在用的发布版：退役只是不再接受新的运行，退役之前写进外部系统的那些事照样发生过，
|
||
版本删掉了，生效记录就对不上它当时的承诺（D-n）。"""
|
||
from app.apps.versions import pinning_apps
|
||
from app.db.models import ActionRecord
|
||
|
||
reasons = [f"应用「{a['name']}」v{a['app_version']} 固定了 v{a['version']}"
|
||
for a in await pinning_apps(session, workflow_id)]
|
||
rows = (await session.execute(
|
||
select(ActionRecord.workflow_version, func.count(ActionRecord.id))
|
||
.where(ActionRecord.workflow_id == workflow_id, ActionRecord.state != "simulated")
|
||
.group_by(ActionRecord.workflow_version).order_by(ActionRecord.workflow_version))).all()
|
||
retired = set((await session.execute(select(WorkflowVersion.version).where(
|
||
WorkflowVersion.workflow_id == workflow_id, WorkflowVersion.retired_at.is_not(None)))).scalars())
|
||
for version, count in rows:
|
||
which = f"v{version}" if version is not None else "未记版本的运行"
|
||
reasons.append(f"{which}{'（已退役）' if version in retired else ''} 有 {count} 条生效记录")
|
||
return reasons
|
||
|
||
|
||
@router.post("/{workflow_id}/duplicate", response_model=WorkflowOut, status_code=201)
|
||
async def duplicate_workflow(
|
||
workflow_id: str, session: AsyncSession = Depends(get_session)
|
||
) -> WorkflowOut:
|
||
source = await session.get(Workflow, workflow_id)
|
||
if not source:
|
||
raise HTTPException(404, _GONE)
|
||
copy = Workflow(
|
||
name=f"{source.name} 副本",
|
||
description=source.description,
|
||
graph=source.graph,
|
||
tags=list(source.tags or []),
|
||
)
|
||
session.add(copy)
|
||
await session.commit()
|
||
await session.refresh(copy)
|
||
return await _workflow_out(session, copy)
|
||
|
||
|
||
class VersionOut(BaseModel):
|
||
id: str
|
||
version: int
|
||
note: str
|
||
created_at: Any = None
|
||
#: 这一版按哪一档发布的（published / governed）。没发布过、或者是记级别之前发布的为 null
|
||
level: str | None = None
|
||
#: 这一版是谁、什么时候发布的；退役了的是什么时候、谁退的、退役时写的说明（没写为 null）。存量为 null
|
||
published_at: Any = None
|
||
published_by: str | None = None
|
||
retired_at: Any = None
|
||
retired_by: str | None = None
|
||
retire_note: str | None = None
|
||
|
||
model_config = {"from_attributes": True}
|
||
|
||
|
||
@router.get("/{workflow_id}/versions", response_model=list[VersionOut])
|
||
async def list_versions(
|
||
workflow_id: str, session: AsyncSession = Depends(get_session)
|
||
) -> list[WorkflowVersion]:
|
||
rows = await session.execute(
|
||
select(WorkflowVersion)
|
||
.where(WorkflowVersion.workflow_id == workflow_id)
|
||
.order_by(WorkflowVersion.version.desc())
|
||
)
|
||
return list(rows.scalars())
|
||
|
||
|
||
class VersionDetailOut(VersionOut):
|
||
workflow_id: str
|
||
graph: dict[str, Any]
|
||
graph_hash: str | None = None
|
||
#: 这一版入口节点声明的字段。正式运行跑的是已发布的那一版，表单得照它来填——
|
||
#: 画布上改过入口字段时，拿当前画布的字段去跑已发布版本会传错参数
|
||
input_fields: list[dict[str, Any]] = Field(default_factory=list)
|
||
published: bool = False
|
||
#: 发布时记下的数据目录版本 {"direct": {源名: {表名: 版本}}, "possible": {源名: {表名: 版本}}}：direct 写死的
|
||
#: SQL 用到的表，possible Agent 可能查询的表（catalog_impact.recorded_versions，只记了直接引用的旧记录 possible
|
||
#: 为空）；没发布过、或者记录之前发布的为 null
|
||
catalog_versions: dict[str, Any] | None = None
|
||
#: 发布之后目录有变化的表：[{source, source_id, table, label, published, current, impact}]
|
||
catalog_changes: list[dict[str, Any]] = Field(default_factory=list)
|
||
#: 入口字段换成表单（agentlab.form/1）：发起表单、应用入口用同一个渲染器（X2）
|
||
input_schema: dict[str, Any] = Field(default_factory=dict)
|
||
#: 账本（L4）：固定这一版的应用、生效记录数、正式运行数
|
||
usage: dict[str, Any] = Field(default_factory=dict)
|
||
|
||
|
||
@router.get("/{workflow_id}/versions/{version}", response_model=VersionDetailOut)
|
||
async def get_version(
|
||
workflow_id: str, version: int, session: AsyncSession = Depends(get_session)
|
||
) -> VersionDetailOut:
|
||
"""取某一版的完整图。"""
|
||
workflow = await session.get(Workflow, workflow_id)
|
||
if not workflow:
|
||
raise HTTPException(404, _GONE)
|
||
snapshot = (
|
||
await session.execute(
|
||
select(WorkflowVersion).where(
|
||
WorkflowVersion.workflow_id == workflow_id,
|
||
WorkflowVersion.version == version,
|
||
)
|
||
)
|
||
).scalar_one_or_none()
|
||
if not snapshot:
|
||
raise HTTPException(404, f"「{workflow.name}」不存在 v{version} 版本")
|
||
|
||
graph = snapshot.graph or {}
|
||
fields: list[dict[str, Any]] = []
|
||
for node in graph.get("nodes") or []:
|
||
if node.get("type") == "input":
|
||
fields.extend(
|
||
f for f in ((node.get("data") or {}).get("config") or {}).get("fields") or []
|
||
if isinstance(f, dict)
|
||
)
|
||
from app.apps.forms import input_schema
|
||
from app.apps.versions import usage
|
||
|
||
return VersionDetailOut(
|
||
id=snapshot.id, version=snapshot.version, note=snapshot.note,
|
||
created_at=snapshot.created_at, workflow_id=workflow_id, graph=graph,
|
||
graph_hash=snapshot.graph_hash or graph_hash(graph), input_fields=fields,
|
||
published=workflow.published_version == snapshot.version, level=snapshot.level,
|
||
catalog_versions=recorded_versions(snapshot.catalog_versions),
|
||
catalog_changes=await catalog_drift(session, snapshot.catalog_versions),
|
||
input_schema=input_schema(fields), published_at=snapshot.published_at, published_by=snapshot.published_by,
|
||
retired_at=snapshot.retired_at, retired_by=snapshot.retired_by, retire_note=snapshot.retire_note,
|
||
usage=await usage(session, workflow_id, snapshot.version),
|
||
)
|
||
|
||
|
||
class RetireIn(BaseModel):
|
||
note: str = Field(default="", max_length=2000)
|
||
|
||
|
||
@router.post("/{workflow_id}/versions/{version}/retire")
|
||
async def retire_version(
|
||
workflow_id: str, version: int, payload: RetireIn, session: AsyncSession = Depends(get_session),
|
||
x_actor: str | None = Header(default=None),
|
||
) -> dict[str, Any]:
|
||
"""退役一个发布过的版本：不再接受新的正式运行，进行中的实例照常走完。当前发布版不能退役；返回受影响的应用
|
||
（它们标「需升级」）。只有管理员能做（Q8），关登录时放行。"""
|
||
from app.apps.versions import RetireRefused, retire_version as retire
|
||
|
||
_require_admin("退役版本")
|
||
try:
|
||
return await retire(session, workflow_id, version, actor=actor_of(x_actor), note=payload.note)
|
||
except RetireRefused as e:
|
||
if e.code:
|
||
raise CodedHTTPException(e.status_code, e.message, e.code) from e
|
||
raise HTTPException(e.status_code, e.message) from e
|
||
|
||
|
||
def _require_admin(what: str) -> None:
|
||
"""受管级发布、版本退役只给管理员（Q8）。关登录时所有人等同管理员。"""
|
||
from app.api.deps import current_principal
|
||
|
||
if not current_principal().at_least("admin"):
|
||
raise CodedHTTPException(403, f"{what}需要管理员权限", ADMIN_ONLY)
|
||
|
||
|
||
@router.post("/{workflow_id}/versions/{version}/restore", response_model=WorkflowOut)
|
||
async def restore_version(
|
||
workflow_id: str, version: int, session: AsyncSession = Depends(get_session)
|
||
) -> WorkflowOut:
|
||
workflow = await session.get(Workflow, workflow_id)
|
||
if not workflow:
|
||
raise HTTPException(404, _GONE)
|
||
snapshot = (
|
||
await session.execute(
|
||
select(WorkflowVersion).where(
|
||
WorkflowVersion.workflow_id == workflow_id,
|
||
WorkflowVersion.version == version,
|
||
)
|
||
)
|
||
).scalar_one_or_none()
|
||
if not snapshot:
|
||
raise HTTPException(404, f"不存在 v{version} 版本")
|
||
|
||
workflow.graph = snapshot.graph
|
||
workflow.version += 1
|
||
session.add(
|
||
WorkflowVersion(
|
||
workflow_id=workflow.id,
|
||
version=workflow.version,
|
||
graph=snapshot.graph,
|
||
graph_hash=graph_hash(snapshot.graph),
|
||
note=f"回滚到 v{version}",
|
||
)
|
||
)
|
||
# 恢复旧版也是改图，和 PATCH 同一条规矩：旧图没过这次的闸，不能挂着受管 / 已发布
|
||
if workflow.status in ("published", "governed"):
|
||
workflow.status = "draft"
|
||
await session.commit()
|
||
await session.refresh(workflow)
|
||
return await _workflow_out(session, workflow)
|
||
|
||
|
||
class PublishIn(BaseModel):
|
||
level: str = "published" # published | governed
|
||
version: int | None = None # 默认发布当前最新版本
|
||
|
||
|
||
@router.post("/{workflow_id}/publish")
|
||
async def publish_workflow(
|
||
workflow_id: str,
|
||
payload: PublishIn,
|
||
session: AsyncSession = Depends(get_session),
|
||
x_actor: str | None = Header(default=None),
|
||
) -> dict[str, Any]:
|
||
"""把某个版本立为正式版本。
|
||
|
||
formal 运行只能从这里发布过的版本发起。governed 级别额外过治理 lint：
|
||
错误会挡住发布——这就是"必经检查点"落地的地方。
|
||
返回 200 + ok/issues 而不是抛错，让前端能把问题列表渲染出来。
|
||
"""
|
||
if payload.level not in _LEVELS:
|
||
raise HTTPException(400, _BAD_LEVEL)
|
||
if payload.level == "governed":
|
||
_require_admin("以受管级别发布")
|
||
workflow = await session.get(Workflow, workflow_id)
|
||
if not workflow:
|
||
raise HTTPException(404, _GONE)
|
||
if await downgrades_governed(session, workflow, payload.level):
|
||
# 把受管工作流降成已发布，之后的正式运行就不再过治理门禁：和以受管级别发布是同一级的决定（Q8）
|
||
_require_admin("把受管工作流降为已发布")
|
||
|
||
version = payload.version or workflow.version
|
||
snapshot = (
|
||
await session.execute(
|
||
select(WorkflowVersion).where(
|
||
WorkflowVersion.workflow_id == workflow_id,
|
||
WorkflowVersion.version == version,
|
||
)
|
||
)
|
||
).scalar_one_or_none()
|
||
if not snapshot:
|
||
raise HTTPException(404, f"不存在 v{version} 版本：画布上的修改还没有保存，请先保存再发布")
|
||
|
||
issues = await gate_issues(session, snapshot.graph, payload.level)
|
||
if issues is None:
|
||
return {"ok": False, "issues": [{"level": "error", "message": _unparsed(snapshot.graph)}]}
|
||
if any(i["level"] == "error" for i in issues):
|
||
return {"ok": False, "level": payload.level, "version": version, "issues": issues}
|
||
return {**await commit_publish(session, workflow, snapshot, payload.level, actor_of(x_actor)),
|
||
"issues": issues} # 剩下的都是警告
|
||
|
||
|
||
async def downgrades_governed(session: AsyncSession, workflow: Workflow, level: str) -> bool:
|
||
"""这次发布会不会把受管工作流降成已发布（Q8：只给管理员）。发布接口和发布单的预检共用。"""
|
||
return level != "governed" and bool(workflow.published_version) and \
|
||
(await _published_levels(session, [workflow.id])).get(workflow.id) == "governed"
|
||
|
||
|
||
def _unparsed(graph: Any) -> str:
|
||
try:
|
||
GraphSpec.model_validate(graph)
|
||
except Exception as e: # noqa: BLE001
|
||
return f"无法解析工作流结构：{graph_error(e)}"
|
||
return "无法解析工作流结构"
|
||
|
||
|
||
async def gate_issues(session: AsyncSession, graph: Any, level: str) -> list[dict[str, Any]] | None:
|
||
"""发布这张图会遇到的问题（validate + 门禁），只读：不改库、不发布。图不成形时为 None。
|
||
发布接口和发布单的预检（blueprints/publish_plan.py）共用，口径只有这一份。"""
|
||
try:
|
||
spec = GraphSpec.model_validate(graph)
|
||
except Exception: # noqa: BLE001
|
||
return None
|
||
checkers = await _sql_checkers(session, spec)
|
||
providers = await _providers(session, spec)
|
||
with with_subgraphs(await load_subgraphs(session, spec)):
|
||
return [i.model_dump() for i in publish_issues(spec, level=level, checkers=checkers, providers=providers)]
|
||
|
||
|
||
async def commit_publish(session: AsyncSession, workflow: Workflow, snapshot: WorkflowVersion, level: str,
|
||
actor: str | None) -> dict[str, Any]:
|
||
"""把这一版立为正式版本并提交（调用方先过了 gate_issues、权限）。发布接口和发布单共用。"""
|
||
# 发布时用到的能力的治理版本（F5）。在改这一版之前取：能力登记缺了的要另开会话同步、提交，这个会话手里不能先有写
|
||
from app.capabilities.registry import capability_versions_for
|
||
from app.db.base import utcnow
|
||
from app.engine.launch import graph_tool_names
|
||
|
||
spec = GraphSpec.model_validate(snapshot.graph)
|
||
capabilities = await capability_versions_for(session, graph_tool_names(snapshot.graph))
|
||
if not snapshot.graph_hash:
|
||
snapshot.graph_hash = graph_hash(snapshot.graph)
|
||
# 级别记在这一版上：之后改画布（status 退回 draft），从这一版发起的正式运行照样按发布时的级别出具
|
||
snapshot.level = level
|
||
# 发布时的目录版本也记在这一版上：重新发布同一版等于按现在的目录重新认可它，跟着更新
|
||
snapshot.catalog_versions = await catalog_versions_for(session, spec)
|
||
# 发布人、发布时间、用到的能力版本也记在这一版上（F5）：工作流上的 published_by 只记最近一次。退役过的版本重新
|
||
# 发布等于重新认可它，退役信息清掉
|
||
snapshot.published_at = utcnow()
|
||
snapshot.published_by = actor
|
||
snapshot.capability_versions = capabilities
|
||
snapshot.retired_at = snapshot.retired_by = snapshot.retire_note = None
|
||
workflow.status = level
|
||
workflow.published_version = snapshot.version
|
||
workflow.published_by = actor
|
||
await session.commit()
|
||
return {
|
||
"ok": True,
|
||
"level": level,
|
||
"version": snapshot.version,
|
||
"graph_hash": snapshot.graph_hash,
|
||
"published_by": workflow.published_by,
|
||
"catalog_versions": recorded_versions(snapshot.catalog_versions),
|
||
}
|
||
|
||
|
||
# --------------------------------------------------------------------------
|
||
# 发布前检查与自动修复：两个接口都只读，不改库、不发布
|
||
# --------------------------------------------------------------------------
|
||
|
||
class PublishCheckIn(BaseModel):
|
||
level: str = "published"
|
||
#: 画布上还没保存的图。不带就用库里当前的草稿
|
||
graph: dict[str, Any] | None = None
|
||
|
||
|
||
class AutofixIn(PublishCheckIn):
|
||
#: 要应用的修复 id（publish-check 给的 fixes[].id）
|
||
apply: list[str] = Field(default_factory=list)
|
||
#: choice 类修复的选择：{fix_id: 值}，可以多选的给列表
|
||
choices: dict[str, Any] = Field(default_factory=dict)
|
||
#: 剩下的 error 交给 Copilot 试着修（结果同样只是预览）
|
||
assist: bool = False
|
||
provider: str | None = None
|
||
model: str | None = None
|
||
|
||
|
||
async def _draft(session: AsyncSession, workflow_id: str, payload: PublishCheckIn) -> dict[str, Any]:
|
||
if payload.level not in _LEVELS:
|
||
raise HTTPException(400, _BAD_LEVEL)
|
||
workflow = await session.get(Workflow, workflow_id)
|
||
if not workflow:
|
||
raise HTTPException(404, _GONE)
|
||
return payload.graph if payload.graph is not None else (workflow.graph or {})
|
||
|
||
|
||
async def _gate_context(session: AsyncSession, spec: GraphSpec) -> tuple[dict[str, Any], dict[str, Any]]:
|
||
"""修复要用、但门禁本身不查库的东西：子工作流当前的发布版本，钉在别处的口径卡有哪些指标。"""
|
||
versions: dict[str, Any] = {}
|
||
cards: dict[str, Any] = {}
|
||
for node in spec.nodes:
|
||
cfg = node.config
|
||
if node.type == NodeType.SUBGRAPH and cfg.get("workflow_id") and str(cfg["workflow_id"]) not in versions:
|
||
upstream = await session.get(Workflow, str(cfg["workflow_id"]))
|
||
versions[str(cfg["workflow_id"])] = upstream.published_version if upstream else None
|
||
elif node.type == NodeType.METRICS and isinstance(cfg.get("caliber_from"), dict):
|
||
ref = cfg["caliber_from"]
|
||
version = ref.get("workflow_version")
|
||
if not ref.get("workflow_id") or not str(version or "").isdigit():
|
||
continue
|
||
snapshot = (await session.execute(select(WorkflowVersion).where(
|
||
WorkflowVersion.workflow_id == str(ref["workflow_id"]),
|
||
WorkflowVersion.version == int(version)))).scalar_one_or_none()
|
||
card = next((n for n in ((snapshot.graph if snapshot else None) or {}).get("nodes") or []
|
||
if isinstance(n, dict) and n.get("id") == ref.get("node_id")), None)
|
||
if card is not None:
|
||
cards[node.id] = ((card.get("data") or {}).get("config") or {}).get("metrics") or []
|
||
return versions, cards
|
||
|
||
|
||
async def _sql_checkers(session: AsyncSession, spec: GraphSpec) -> dict[str, Any]:
|
||
"""图里调用工具节点写死的 SQL 用到的数据源，各建一个基于数据目录的 SQL 检查器。发布、发布前检查、
|
||
自动修复都传这一份给门禁（governance.lint_for_publish）：门禁本身不查库，三处口径才一致。"""
|
||
from app.data.sqlcheck import graph_sources, load_checkers_by_name
|
||
|
||
return await load_checkers_by_name(session, graph_sources(spec.nodes))
|
||
|
||
|
||
async def _providers(session: AsyncSession, spec: GraphSpec) -> list[Any]:
|
||
"""模型接入表：节点点名的接入不存在或已停用、只写了模型而提供它的接入已停用时，发布、发布前检查、自动修复的复核
|
||
给 error、挡住发布，修复从启用的接入里列候选（engine/provider_refs.py、autofix 的 provider.*）。和 _sql_checkers
|
||
一样由接口这一层读好交给门禁，门禁本身不查库；三处传同一份。图里取模型的几处都没写接入和模型时不读库。"""
|
||
return await load_providers(session, spec)
|
||
|
||
|
||
def _unreadable(level: str, e: Exception) -> dict[str, Any]:
|
||
return {"level": level, "ok": False, "fixes": [], "issues": [
|
||
{"level": "error", "node_id": None, "edge_id": None, "field": None, "code": None, "fix": None,
|
||
"message": f"无法解析工作流结构：{graph_error(e)}"}]}
|
||
|
||
|
||
@router.post("/{workflow_id}/publish-check")
|
||
async def publish_check(
|
||
workflow_id: str, payload: PublishCheckIn, session: AsyncSession = Depends(get_session)
|
||
) -> dict[str, Any]:
|
||
"""发布前检查：和真正发布同一套口径（validate + 门禁）列出问题，以及能怎么修。
|
||
|
||
只读：不改库、不发布。发布弹窗一打开就调它，不用等点了发布才知道被拦。
|
||
"""
|
||
return await check_graph(session, await _draft(session, workflow_id, payload), payload.level)
|
||
|
||
|
||
async def check_graph(session: AsyncSession, graph: Any, level: str) -> dict[str, Any]:
|
||
"""发布前检查的本体：问题加上能怎么修（每条问题带 fix 的种类 auto / choice / assist）。只读。
|
||
发布前检查接口和发布单的预检共用（发布单按修复的种类把问题分给「你来定」「助手可以修」）。"""
|
||
from app.engine.autofix import check
|
||
|
||
try:
|
||
spec = GraphSpec.model_validate(graph)
|
||
except Exception as e: # noqa: BLE001
|
||
return _unreadable(level, e)
|
||
versions, cards = await _gate_context(session, spec)
|
||
checkers = await _sql_checkers(session, spec)
|
||
providers = await _providers(session, spec)
|
||
with with_subgraphs(await load_subgraphs(session, spec)):
|
||
return check(spec, level=level, versions=versions, cards=cards, checkers=checkers, providers=providers)
|
||
|
||
|
||
@router.post("/{workflow_id}/autofix")
|
||
async def autofix(
|
||
workflow_id: str, payload: AutofixIn, session: AsyncSession = Depends(get_session)
|
||
) -> dict[str, Any]:
|
||
"""应用选中的修复，返回修复后的整张图——只是预览。
|
||
|
||
不自动保存、不自动发布：人看过逐项变更再确认，前端走现有的保存接口存草稿，再检查、再发布。
|
||
每条修复都复核过（降低要求的、错误没变少或者冒出新错误的丢弃并写明原因）。assist 为真时，
|
||
剩下的 error 交给 Copilot，同一道复核通过才采纳。
|
||
"""
|
||
from app.engine.autofix import apply_fixes, diff_changes
|
||
|
||
graph = await _draft(session, workflow_id, payload)
|
||
try:
|
||
spec = GraphSpec.model_validate(graph)
|
||
except Exception as e: # noqa: BLE001
|
||
bad = _unreadable(payload.level, e)
|
||
return {"graph": graph, "changes": [], "ops": [], "applied": [], "rejected": [
|
||
{"fix_id": fid, "reason": "无法解析工作流结构，不能应用修复"} for fid in payload.apply],
|
||
"remaining": bad["issues"], "fixes": [], "handoff": [], "assist": None, "ok": False}
|
||
versions, cards = await _gate_context(session, spec)
|
||
checkers = await _sql_checkers(session, spec)
|
||
providers = await _providers(session, spec)
|
||
# 和发布、发布前检查同一套口径：子工作流先读好，process.human_in_subgraph 才查得到
|
||
subgraphs = await load_subgraphs(session, spec)
|
||
with with_subgraphs(subgraphs):
|
||
out = apply_fixes(graph, payload.apply, payload.choices, level=payload.level, versions=versions, cards=cards,
|
||
checkers=checkers, providers=providers)
|
||
out["assist"] = None
|
||
# 人在选项里选了「交给 Copilot」（handoff），和点「交给 Copilot」是一回事
|
||
if payload.assist or out["handoff"]:
|
||
if out["ok"]:
|
||
out["assist"] = {"ok": True, "summary": "修复后已没有阻止发布的错误，未再交给助手", "questions": []}
|
||
return out
|
||
left = [i for i in out["remaining"] if i["level"] == "error"]
|
||
if all(i.get("code") in PROVIDER_CODES for i in left):
|
||
# 剩下的只有点名接入的问题：换成哪个接入要人拿主意（助手的自查也不看接入表），不交给助手
|
||
fixable = any(i.get("fix") for i in left)
|
||
out["assist"] = {"ok": False, "summary": PROVIDER_LEFT + (PROVIDER_LEFT_FIX if fixable else ""),
|
||
"questions": []}
|
||
return out
|
||
from app.api.copilot import assist_publish_fix
|
||
|
||
helped = await assist_publish_fix(session, out["graph"], level=payload.level,
|
||
provider=payload.provider, model=payload.model)
|
||
out["assist"] = {"ok": helped["accepted"], "summary": helped["summary"], "questions": helped["questions"],
|
||
"looks": helped["looks"]}
|
||
if helped["accepted"]:
|
||
fixed = helped["graph"]
|
||
out["changes"] += diff_changes(out["graph"], fixed, fix_id="assist", label="助手的修改")
|
||
out["ops"] += helped["ops"]
|
||
out["applied"].append("assist")
|
||
fixed_spec = GraphSpec.model_validate(fixed)
|
||
fixed_checkers = await _sql_checkers(session, fixed_spec)
|
||
fixed_providers = await _providers(session, fixed_spec)
|
||
with with_subgraphs(await load_subgraphs(session, fixed_spec)):
|
||
after = apply_fixes(fixed, [], level=payload.level, versions=versions, cards=cards,
|
||
checkers=fixed_checkers, providers=fixed_providers)
|
||
out.update(graph=fixed, remaining=after["remaining"], fixes=after["fixes"], ok=after["ok"])
|
||
elif helped["reason"]:
|
||
out["rejected"].append({"fix_id": "assist", "reason": helped["reason"]})
|
||
return out
|
||
|
||
|
||
class ValidateIn(BaseModel):
|
||
graph: dict[str, Any]
|
||
|
||
|
||
@router.post("/validate")
|
||
async def validate(payload: ValidateIn, session: AsyncSession = Depends(get_session)) -> dict[str, Any]:
|
||
"""画布上的实时校验。前端每次改动都会调它来标红。
|
||
|
||
点名的模型接入不存在或已停用（只写了模型、提供它的接入已停用也算）只给警告：探索运行仍可发起，执行到那个节点时
|
||
照实报错；发布时才挡住。这时同一节点上「没有指定模型，将使用接入「X」的默认模型」和它矛盾，不再列
|
||
（quiet_unset_model，发布一类在 governance._not_repeated 里同样处理）。"""
|
||
try:
|
||
spec = GraphSpec.model_validate(payload.graph)
|
||
except Exception as e: # noqa: BLE001
|
||
return {"ok": False, "issues": [{"level": "error", "message": f"无法解析工作流结构：{graph_error(e)}"}]}
|
||
result = validate_graph(spec)
|
||
found = await load_issues(session, spec, level="warning")
|
||
result.issues = [*quiet_unset_model(spec, result.issues, found), *found]
|
||
return result.model_dump()
|
||
|
||
|
||
@router.post("/variables")
|
||
async def variables(payload: ValidateIn) -> dict[str, Any]:
|
||
"""这张图里有哪些变量、谁产出、谁引用。
|
||
|
||
纯静态分析，不用跑图也不用有运行记录——用户在编排到一半时最需要它，
|
||
而那时候还没有任何一次运行可看。运行期的实际取值由前端从事件里补上。
|
||
"""
|
||
from app.engine.variables import analyze
|
||
|
||
try:
|
||
spec = GraphSpec.model_validate(payload.graph)
|
||
except Exception as e: # noqa: BLE001
|
||
return {"variables": [], "issues": [{"level": "error", "message": f"无法解析工作流结构：{graph_error(e)}"}]}
|
||
return analyze(spec).model_dump()
|