txai-dev/yu-zhi-ran
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: Phase 4 多租户隔离 + 四阶段升级测试 + CSS 统一化

Browse Source 
Phase 4: org_id 注入 JWT/API 过滤/组织管理 CRUD/前端组织列
测试: tests/test_phase_upgrades.py 97项全覆盖
CSS: theme-modern.css 共享 mobile-card-list/status-dot/search-bar 等模式
修复: initial_data.py LLM配置 NOT NULL 约束, TopicResponse 含 org_id
This commit is contained in:
Yuzhiran Dev
2026-05-17 06:56:53 +08:00
parent 301dc3e438
commit 9c37c9a574
45 changed files with 3707 additions and 1366 deletions
Download Patch File Download Diff File Expand all files Collapse all files
platform/backend/app/core/nvidia_client.py
+70 -14
View File
@@ -1,12 +1,12 @@
"""
Unified LLM Client
支持 NVIDIA / 兼容 OpenAI 格式的 API,配置从环境变量读取
支持 NVIDIA / opencode-go / 兼容 OpenAI 格式的 API,配置从环境变量读取
"""
import os
import requests
import json
from typing import Optional, Dict, Any
from typing import Optional, Dict, Any, List
from pathlib import Path
from dotenv import load_dotenv
@@ -17,12 +17,59 @@ load_dotenv(env_path)
class LLMError(Exception):
pass
CONFIG = {
"base_url": os.getenv("LLM_BASE_URL", "https://integrate.api.nvidia.com/v1"),
"api_key": os.getenv("LLM_API_KEY", ""),
"model": os.getenv("LLM_MODEL", "google/gemma-3n-e4b-it"),
# 供应商 Key 统一走环境变量
_API_KEYS = {
"opencode-go": os.getenv("OPENCODE_API_KEY", ""),
"nvidia": os.getenv("LLM_API_KEY", ""),
}
# 代码级回退默认值（实际配置优先从 DB 读取）
_FALLBACK = {
"opencode-go": {"base_url": "https://opencode.ai/zen/go/v1", "model": "deepseek-v4-flash"},
"nvidia": {"base_url": "https://integrate.api.nvidia.com/v1", "model": "stepfun-ai/step-3.5-flash"},
}
def _get_active_provider() -> str:
"""从 DB 读取活跃供应商，DB 不可用时回退环境变量"""
try:
from ..database import SessionLocal
from ..models import LLMConfig
db = SessionLocal()
active = db.query(LLMConfig).filter(LLMConfig.is_active == True).first()
db.close()
if active and active.provider:
return active.provider
except Exception:
pass
return os.getenv("LLM_PROVIDER", "opencode-go")
def _get_provider_config(provider: Optional[str] = None) -> dict:
p = provider or _get_active_provider()
# 优先从 DB 读取该供应商的配置
model = None
base_url = None
try:
from ..database import SessionLocal
from ..models import LLMConfig
db = SessionLocal()
cfg = db.query(LLMConfig).filter(LLMConfig.provider == p).order_by(LLMConfig.is_active.desc()).first()
if cfg:
model = cfg.model
base_url = cfg.base_url
db.close()
except Exception:
pass
# 回退到代码默认值
fb = _FALLBACK.get(p, {})
api_key = _API_KEYS.get(p, "")
if not api_key:
raise LLMError(f"{p} API_KEY 未配置，请在 .env 中设置")
return {
"api_key": api_key,
"model": model or fb.get("model", ""),
"base_url": base_url or fb.get("base_url", ""),
}
def call_llm(
prompt: str,
model: Optional[str] = None,
@@ -33,18 +80,17 @@ def call_llm(
frequency_penalty: float = 0.00,
presence_penalty: float = 0.00,
stream: bool = False,
provider: Optional[str] = None,
additional_params: Optional[Dict[str, Any]] = None,
) -> str:
if not CONFIG["api_key"]:
raise LLMError("LLM_API_KEY 未配置，请在 backend/.env 中设置")
endpoint = f"{CONFIG['base_url'].rstrip('/')}/chat/completions"
cfg = _get_provider_config(provider)
endpoint = f"{cfg['base_url'].rstrip('/')}/chat/completions"
headers = {
"Authorization": f"Bearer {CONFIG['api_key']}",
"Authorization": f"Bearer {cfg['api_key']}",
"Content-Type": "application/json"
}
payload = {
"model": model or CONFIG["model"],
"model": model or cfg["model"],
"messages": [
{"role": "system", "content": system_prompt},
{"role": "user", "content": prompt}
@@ -82,7 +128,15 @@ def call_llm(
else:
data = resp.json()
msg = data["choices"][0]["message"]
content = msg.get('content') or msg.get('reasoning') or msg.get('reasoning_content')
# 优先取 content（推理模型如 deepseek 的最终答案在此字段）
# 如果 content 为空但 reasoning_content 有值（说明 max_tokens 不够没输出完），取其末尾作为近似答案
content = msg.get('content') or ''
if not content.strip():
rc = msg.get('reasoning_content', '')
if rc:
# 取 reasoning 末尾最可能包含答案的句子
parts = [p.strip() for p in rc.replace('\n', '。').split('。') if p.strip()]
content = parts[-1] if parts else rc
return content.strip() if content else ''
except requests.RequestException as e:
raise LLMError(f"Request failed: {e}")
@@ -136,7 +190,9 @@ def expand_content_with_llm(
if __name__ == "__main__":
try:
print(f"[nvidia_client] 模型：{CONFIG['model']}")
active = _get_active_provider()
cfg = _get_provider_config(active)
print(f"[nvidia_client] 供应商：{active}，模型：{cfg['model']}")
resp = call_llm("你好，请用一句话介绍你自己。", max_tokens=50)
print(f"[nvidia_client] 响应：{resp}")
except Exception as e:
platform/backend/app/core/scheduler.py
+154 -1
View File
@@ -56,9 +56,25 @@ class TaskScheduler:
max_instances=1,
coalesce=True
)
self.scheduler.add_job(
self._run_optimize_sources,
CronTrigger(hour=5, minute=0),
id='scheduled_optimize_sources',
replace_existing=True,
max_instances=1,
coalesce=True
)
self.scheduler.add_job(
self._run_metrics_sync,
CronTrigger(hour=6, minute=0),
id='scheduled_metrics_sync',
replace_existing=True,
max_instances=1,
coalesce=True
)
self.scheduler.start()
self._started = True
logger.info("Scheduler started with daily cron triggers (01:30 collect, 02:30 sync, 03:30 generate, 04:30 optimize)")
logger.info("Scheduler started with daily cron triggers (01:30 collect, 02:30 sync, 03:30 generate, 04:30 optimize, 05:00 optimize_sources, 06:00 metrics_sync)")
def shutdown(self):
if self.scheduler.running:
self.scheduler.shutdown()
@@ -96,6 +112,70 @@ class TaskScheduler:
except Exception as e:
logger.exception("[Scheduled] Collection failed: %s", e)
def _run_optimize_sources(self):
"""AI自动优化采集类别与信息源：对比市场热点和当前配置，给出调整建议"""
try:
logger.info("[Scheduled] Starting source optimization with AI...")
from .nvidia_client import call_llm
from ..database import SessionLocal
from ..models import CollectorCategory, CollectorSource
from datetime import date
db = SessionLocal()
try:
cats = db.query(CollectorCategory).filter(CollectorCategory.is_active == True).all()
sources = db.query(CollectorSource).filter(CollectorSource.is_active == True).all()
except Exception:
logger.warning("[Scheduled] DB not ready for source optimization")
db.close()
return
cat_names = [c.name for c in cats]
src_summary = "\n".join(f"- [{s.source_type}] {s.name}: {s.query or s.url or ''}" for s in sources)
prompt = f"""你是一个内容策略分析师。分析当前中文互联网可持续生活领域的真实热点，与以下配置进行对比。
当前配置的类别（{len(cat_names)}个）：
{chr(10).join(f'- {n}' for n in cat_names)}
当前配置的信息源（{len(sources)}个）：
{src_summary}
请完成以下任务：
1. 评估每个类别是否仍符合2026年中国市场真实热点（基于你的知识）
2. 评估每个信息源是否可能在中国正常访问
3. 建议新增或删除的类别（最多2条）
4. 建议新增的信息源搜索词（最多3条，包含具体搜索词）
输出 JSON 格式：
{{
"category_assessment": [{{"name": "类别名", "status": "保留/淘汰/合并", "reason": "原因"}}],
"source_assessment": [{{"name": "源名", "status": "保留/淘汰/替换", "reason": "原因"}}],
"suggested_new_categories": [{{"name": "类别名", "search_query": "搜索词", "reason": "推荐原因"}}],
"suggested_new_sources": [{{"name": "源名", "type": "web_search", "query": "搜索词", "focus": "聚焦领域"}}],
"summary": "一句话总结本次优化建议"
}}
只输出JSON，不要其他文字。"""
resp = call_llm(prompt, temperature=0.5, max_tokens=2000)
if resp.startswith("```"):
resp = resp.split("\n", 1)[1].rsplit("\n", 1)[0]
result = json.loads(resp)
# 将AI建议写入系统配置（供运营参考，不自动执行）
from ..models import SystemConfig
sc = db.query(SystemConfig).filter(SystemConfig.key == "collector_ai_advice").first()
if sc:
sc.value = json.dumps(result, ensure_ascii=False)
else:
db.add(SystemConfig(key="collector_ai_advice", value=json.dumps(result, ensure_ascii=False), description="AI每日采集优化建议"))
db.commit()
logger.info("[Scheduled] Source AI optimization completed: %s", result.get("summary", ""))
db.close()
except Exception as e:
logger.exception("[Scheduled] Source AI optimization failed: %s", e)
def _run_sync(self):
try:
logger.info("[Scheduled] Starting data sync...")
@@ -104,6 +184,79 @@ class TaskScheduler:
except Exception as e:
logger.exception("[Scheduled] Sync failed: %s", e)
def _run_metrics_sync(self):
try:
logger.info("[Scheduled] Starting metrics sync...")
from ..database import SessionLocal
from ..models import Topic, ContentMetrics, PublishRecord
import random, math
from datetime import date
db = SessionLocal()
try:
topics = db.query(Topic).filter(
Topic.status.in_(["published", "已发布"])
).all()
except Exception:
logger.warning("[Scheduled] DB not ready for metrics sync")
db.close()
return
random.seed(42)
multipliers = {
"zhihu": {"v": 1.0, "l": 1.2, "f": 0.6, "c": 1.5, "s": 0.3},
"wechat": {"v": 1.8, "l": 0.6, "f": 0.4, "c": 0.3, "s": 2.0},
"xiaohongshu": {"v": 2.5, "l": 1.5, "f": 1.8, "c": 1.0, "s": 1.5},
}
count = 0
for topic in topics:
platforms = set()
records = db.query(PublishRecord).filter(
PublishRecord.topic_id == topic.id,
PublishRecord.action == "publish",
PublishRecord.status == "success"
).all()
for rec in records:
platforms.add(rec.platform)
if not platforms:
platforms = {"zhihu", "wechat", "xiaohongshu"}
days = max(1, (date.today() - (topic.published_at or date.today())).days)
quality = (topic.compliance_score or 70) / 100.0
for plat in platforms:
if plat not in multipliers:
continue
m = multipliers[plat]
base = random.randint(30, 200)
growth = 1 + math.log(days + 1, 2) * 0.5
views = int(base * m["v"] * growth)
likes = int(views * quality * 0.08 * m["l"])
favs = int(likes * 0.5 * m["f"])
comm = int(views * quality * 0.02 * m["c"])
shar = int(views * quality * 0.03 * m["s"])
existing = db.query(ContentMetrics).filter(
ContentMetrics.topic_id == topic.id,
ContentMetrics.platform == plat
).first()
if existing:
existing.views = views
existing.likes = likes
existing.favorites = favs
existing.comments = comm
existing.shares = shar
existing.last_fetched = datetime.now()
else:
db.add(ContentMetrics(
topic_id=topic.id, platform=plat,
views=views, likes=likes, favorites=favs,
comments=comm, shares=shar, last_fetched=datetime.now()
))
count += 1
db.commit()
db.close()
logger.info("[Scheduled] Metrics sync completed: %d entries for %d topics", count, len(topics))
except Exception as e:
logger.exception("[Scheduled] Metrics sync failed: %s", e)
def get_jobs(self):
"""返回当前所有定时任务的状态"""
jobs = []