mirror of
https://github.com/ThearchyHelios/AgentLab.git
synced 2026-11-15 14:31:58 +00:00
ThearchyHelios
e15e94f93e
- 新增口径结构化语义定义、分批解析与 SQL 生成管道（semantic / semantic_import / semantic_sql） - 支持起草断点续跑恢复、指标自动匹配对齐与存量口径语义迁移面板（MigrationPanel / SemanticPanel） - 口径草稿与变更流程接入候选本体边提议、审阅确认与一键修改保护，同步更新可移植 SQLite 基线校验
77 lines
4.6 KiB
Python
77 lines
4.6 KiB
Python
"""Conservative graph append for independently reviewed missing-definition candidates."""
|
|
from __future__ import annotations
|
|
|
|
import copy
|
|
import re
|
|
from typing import Any
|
|
|
|
|
|
def append_candidate(original: dict[str, Any], candidate: dict[str, Any], *, prefix: str,
|
|
allowed_definitions: set[str]) -> dict[str, Any]:
|
|
from app.engine.schema import GraphSpec
|
|
|
|
graph, extra = copy.deepcopy(original), copy.deepcopy(candidate)
|
|
def one(nodes, kind):
|
|
found = [n for n in nodes if n.get('type') == kind]
|
|
if len(found) != 1:
|
|
raise ValueError('图中入口或口径卡不唯一，请在画布中核对合并；候选已保留')
|
|
return found[0]
|
|
nodes, added = graph['nodes'], extra['nodes']
|
|
start, card = one(nodes, 'input'), one(nodes, 'metrics')
|
|
incoming_start, incoming_card = one(added, 'input'), one(added, 'metrics')
|
|
def config(node):
|
|
return node['data']['config']
|
|
old_metrics, new_metrics = config(card)['metrics'], config(incoming_card)['metrics']
|
|
ids = [m.get('source_definition_id') for m in new_metrics]
|
|
new_names = [m.get('name') for m in new_metrics]
|
|
if not ids or None in ids or not set(ids) <= allowed_definitions:
|
|
raise ValueError('候选包含未请求的指标定义，请核对；候选已保留')
|
|
if not all(new_names) or len(new_names) != len(set(new_names)):
|
|
raise ValueError('候选包含同名重复指标，请核对；候选已保留')
|
|
if set(ids) & {m.get('source_definition_id') for m in old_metrics}:
|
|
raise ValueError('原稿已有这些指标定义，不会覆盖现有实现；候选已保留')
|
|
if set(new_names) & {m.get('name') for m in old_metrics}:
|
|
raise ValueError('存在同名指标，请确认口径后在画布中合并；候选已保留')
|
|
fields = config(start).setdefault('fields', [])
|
|
by_name = {f['name']: f for f in fields}
|
|
for field in config(incoming_start).get('fields', []):
|
|
if field['name'] in by_name and field != by_name[field['name']]:
|
|
raise ValueError(f"入口参数「{field['name']}」定义冲突，不会覆盖原参数；候选已保留")
|
|
if field['name'] not in by_name:
|
|
fields.append(field)
|
|
tools = [n for n in added if n.get('type') == 'tool']
|
|
if not tools or any(n.get('type') not in ('input', 'tool', 'metrics', 'report', 'output') for n in added):
|
|
raise ValueError('候选取数结构已变化，请在画布中核对合并；候选已保留')
|
|
rename = {n['id']: f"{prefix}_{n['id']}" for n in tools}
|
|
rename[incoming_start['id']], rename[incoming_card['id']] = start['id'], card['id']
|
|
if set(rename[n['id']] for n in tools) & {n['id'] for n in nodes}:
|
|
raise ValueError('候选节点已存在，请刷新后核对')
|
|
# Generated SQL uses input parameters, not upstream node references. Refuse edited dependencies.
|
|
if any('nodes.' in str(config(n)) for n in tools):
|
|
raise ValueError('候选查询依赖其他节点，请在画布中核对合并；候选已保留')
|
|
edges = [e for e in extra.get('edges', []) if e['source'] in rename and e['target'] in rename]
|
|
expected_edges = {(a['id'], b['id']) for a, b in zip([incoming_start, *tools], [*tools, incoming_card])}
|
|
if {(e['source'], e['target']) for e in edges} != expected_edges:
|
|
raise ValueError('候选取数链路已变化，请在画布中核对合并；候选已保留')
|
|
for metric in new_metrics:
|
|
expression = metric.get('expression', '')
|
|
refs = set(re.findall(r'\bnodes\.([A-Za-z0-9_]+)', expression))
|
|
if not refs or not refs <= {n['id'] for n in tools}:
|
|
raise ValueError('候选指标依赖未合并的节点，请核对；候选已保留')
|
|
metric['expression'] = re.sub(r'\bnodes\.([A-Za-z0-9_]+)', lambda m: 'nodes.' + rename[m[1]], expression)
|
|
metric['id'] = f"{prefix}_{metric['id']}"
|
|
if {m['id'] for m in new_metrics} & {m['id'] for m in old_metrics}:
|
|
raise ValueError('指标编号冲突，请核对；候选已保留')
|
|
old_metrics.extend(new_metrics)
|
|
for node in tools:
|
|
node['id'] = rename[node['id']]
|
|
nodes.append(node)
|
|
graph['edges'].extend({**e, 'source': rename[e['source']], 'target': rename[e['target']],
|
|
'id': f'{prefix}_edge_{index}'} for index, e in enumerate(edges))
|
|
for node in nodes:
|
|
contract = config(node).get('contract')
|
|
if node.get('type') == 'output' and isinstance(contract, dict) and contract.get('metrics_from') == card['id']:
|
|
contract['required'] = [*contract.get('required', []), *[m['id'] for m in new_metrics]]
|
|
GraphSpec.model_validate(graph)
|
|
return graph
|