feat(monitor): 抖音 Web 接口客户端 —— 绕开爬虫子进程,直接发 HTTP
Deploy VitePress site to Pages / build (push) Canceled after 0s
Deploy VitePress site to Pages / Deploy (push) Canceled after 0s

移植自 mac-agent-os 的 mediacrawler_adapter:不起子进程、不开页面,用浏览器里那份
登录态直接调抖音 Web 接口。产物键名照抄 store/douyin,所以 ingest 那条链路一个字不用改。

**目前能用的(真环境实测,非推断)**:

    profile/other : 200, 7075 字节  —— 博主主页指标(粉丝/获赞/作品数/昵称)
    aweme/detail  : 200, 45425 字节 —— 单条作品详情(含点赞/评论/收藏/分享)

**目前不能用的:作品列表 `aweme/post`。** 两个互相独立的原因:
  1. 这个接口被抖音单独升级成了真校验:不带 x-tt-argus 回 403「Uifid Not Found」,
     带上 dummy 值回 200 + **空 body**。也就是说「头在不在」骗得过,「真校验」过不了。
     同一套头打 profile/other 和 aweme/detail 都是通的 —— 抖音是挑着接口加保护的,
     挑中的恰好是「批量拉作品列表」这个最敏感的动作。
  2. 改走页面截获也不行:CDP 浏览器打开博主主页会落到「验证码中间页」(当天大量探测的
     代价,过几小时要重测)。

  所以现在的边界是:**已知作品的指标刷新能做,自动发现新作品做不了**。

**排查中控住变量后得到的两条事实**(都写进注释了):
  · `Accept` / `Accept-Language` / `Referer` 才是主页接口能返回真数据的原因 —— 只有
    UA+client hints+Cookie 时是 200 但仅 121 字节的空壳,补上这三个头变 7074 字节。
    (我先前猜的 sec-ch-ua 不是关键。)
  · 因此 UA 与 client hints 必须**成套地取自同一个浏览器**,所以 BrowserIdentity 一次
    从 CDP 取齐 cookie + UA + hints,而不是各自写死。

「200 + 空 body 必须当场报错」也是刻意写死的:放过去它会在下游变成「这个博主没作品」,
把一次失败伪装成一条正常结果 —— 爬虫那条路正是这么栽的,还被翻译成「账号被封」。

测试 +11:cookie 解析、请求头成套性(含 uifid 缺失/回退)、产物键名与 store 对齐、
以及 _get 的三条失败路径(空 body / 403 带网关原话 / 正常返回)。
This commit is contained in:
2026-10-10 17:12:25 +08:00
parent 21ff01b894
commit 2112c1a870
1601 changed files with 367425 additions and 0 deletions
@@ -0,0 +1,306 @@
"""
comment_workbench.py — 评论工作台 API(v3 角色化评论)
路由前缀: /api/comment-workbench
"""
import logging
import sys
from pathlib import Path
from fastapi import APIRouter, HTTPException
logger = logging.getLogger("dashboard.comment_workbench")
_THIS_DIR = Path(__file__).resolve().parent.parent
SCRIPTS_DIR = _THIS_DIR.parent / "07_matrix" / "scripts"
sys.path.insert(0, str(SCRIPTS_DIR))
router = APIRouter(prefix="/api/comment-workbench", tags=["comment_workbench"])
ROLE_LABELS = {
"sharer": "分享型", "questioner": "提问型", "sympathizer": "共情型",
"skeptic": "质疑型", "sufferer": "患者型", "filler": "灌水型",
"expert_ref": "推荐型", "answerer": "解答型",
}
DEFAULT_ROLES = {
"filler": 0.30, "questioner": 0.17, "sharer": 0.17,
"sympathizer": 0.13, "sufferer": 0.10, "skeptic": 0.07,
"expert_ref": 0.06, "answerer": 0.12,
}
def _get_corpus_mgr():
from mc.corpus import CorpusManager
return CorpusManager()
@router.get("/roles")
def api_roles(platform: str = "douyin"):
"""获取可用角色列表及默认比例"""
mgr = _get_corpus_mgr()
roles_raw = mgr.get_roles(platform)
result = []
for role_id, info in roles_raw.items():
label = ROLE_LABELS.get(role_id, role_id)
default_pct = DEFAULT_ROLES.get(role_id, 0)
result.append({
"id": role_id,
"label": label,
"count": info["count"],
"examples": info["examples"][:3],
"default_pct": default_pct,
})
return {"roles": result}
@router.post("/generate")
def api_generate_comments(data: dict):
"""按角色生成评论
Body:
video_title: str — 视频标题
video_tags: [str] — 视频标签(可选)
platform: str — 平台,默认 douyin
role_distribution: dict — 角色比例,空则用默认(比例模式)
role_counts: dict — 角色精确条数 {"filler": 2, ...}(数字模式,优先于比例)
total: int — 比例模式下生成多少条,默认 30
ai_enhance: bool — 是否 AI 改写
long_ratio: float — 长评占比 0~1
"""
video_title = data.get("video_title", "")
video_tags = data.get("video_tags", []) or []
video_industry = data.get("video_industry") or None
guide_points = data.get("guide_points", "") or ""
content_type = data.get("content_type", "") or ""
direction = data.get("direction", "auto")
platform = data.get("platform", "douyin")
role_dist = data.get("role_distribution", {}) or DEFAULT_ROLES
role_counts = data.get("role_counts") or None
total = data.get("total", 30)
ai_enhance = data.get("ai_enhance", False)
long_ratio = data.get("long_ratio", 0.0)
guide_ratio = min(max(float(data.get("guide_ratio", 1.0)), 0.0), 1.0) # 引导结合比例 0~1
if not video_title:
raise HTTPException(400, detail="video_title 必填")
mgr = _get_corpus_mgr()
# 行业:前端指定 > 自动识别 > None(通用)
if video_industry:
industry = video_industry
else:
industry = mgr._classify_video(video_title)
if not industry:
for tag in video_tags:
industry = mgr._classify_video(tag)
if industry:
break
# 视频标签转为逗号字符串
tags_str = ", ".join(video_tags) if isinstance(video_tags, list) else str(video_tags)
logger.info(" 📋 视频行业: %s | 内容类型: %s | 引导: %.30s", industry or "通用", content_type or "-", guide_points or "-")
gen_kwargs = dict(
platform=platform,
video_title=video_title,
video_industry=industry,
guide_points=guide_points,
content_type=content_type,
video_tags=tags_str,
ai_enhance=ai_enhance,
long_ratio=long_ratio,
guide_ratio=guide_ratio,
)
if role_counts:
# 数字模式:精确按角色条数取,total 由后端自动求和
comments = mgr.batch_get_comments_by_roles(
role_counts=role_counts,
**gen_kwargs,
)
else:
comments = mgr.batch_get_comments_by_roles(
role_distribution=role_dist,
total=total,
**gen_kwargs,
)
return {"comments": comments, "total": len(comments)}
@router.post("/generate-discussion")
def api_generate_discussion(data: dict):
"""多人讨论模式:一次生成完整讨论剧本(单次 AI 调用)→ turns 列表
Body:
video_title: str — 视频标题
topic: str — 讨论主题(用户主导概念)
guide_items: [str]|str — 引导要素(多个推荐对象,每行一个)
outline: str — 剧本走向(自由文本,如"先质疑→过来人澄清→带出医生")
total: int — 总条数(3~40,默认 15)
bystander_ratio: float — 路人打酱油占比(0~0.5,默认 0.2)
"""
video_title = (data.get("video_title") or "").strip()
topic = (data.get("topic") or "").strip()
outline = (data.get("outline") or "").strip()
guide_path = (data.get("guide_path") or "").strip()
total = max(3, min(60, int(data.get("total", 15))))
bystander_ratio = min(max(float(data.get("bystander_ratio", 0.2)), 0.0), 0.5)
role_counts = data.get("role_counts") or {}
if not isinstance(role_counts, dict):
role_counts = {}
long_comment_count = max(0, min(3, int(data.get("long_comment_count", 2))))
guide_count = max(0, int(data.get("guide_count", 0))) # 0 = 自动按 30%
guide_raw = data.get("guide_items") or []
if isinstance(guide_raw, str):
guide_items = [l.strip() for l in guide_raw.split("\n") if l.strip()]
else:
guide_items = [str(g).strip() for g in guide_raw if str(g).strip()]
if not video_title and not topic:
raise HTTPException(400, detail="video_title 或 topic 至少填一个")
from mc.corpus import AIGenerator
ai = AIGenerator()
if not ai.available:
raise HTTPException(503, detail="AI 不可用(检查 config/ai.yaml 或 agent-local 配置)")
import asyncio as _asyncio
try:
loop = _asyncio.get_event_loop()
except RuntimeError:
loop = _asyncio.new_event_loop()
turns = loop.run_until_complete(ai.generate_discussion_script(
video_title=video_title,
topic=topic,
guide_items=guide_items,
outline=outline,
total=total,
bystander_ratio=bystander_ratio,
guide_path=guide_path,
role_counts=role_counts,
long_comment_count=long_comment_count,
guide_count=guide_count,
))
logger.info(" 💬 讨论剧本生成: %d 条 (主题=%s 要素=%d)", len(turns), topic[:20], len(guide_items))
warning = ""
# 期望条数:配比模式 = 配比总和;自由模式 = total
expected = total
if role_counts:
try:
expected = sum(int(v) for v in role_counts.values() if str(v).strip().isdigit())
except Exception:
expected = total
if expected and len(turns) < expected:
warning = f"AI 少输出 {expected - len(turns)} 条(要求 {expected},实际 {len(turns)}),可重新生成或手动补"
return {"status": "ok", "turns": turns, "total": len(turns), "requested": expected, "warning": warning}
@router.post("/extract-topic")
def api_extract_topic(data: dict):
"""从视频标题提炼「讨论主题」(简短核心词,供讨论模式作起点)
Body: { video_title: str }
"""
video_title = (data.get("video_title") or "").strip()
if not video_title:
raise HTTPException(400, detail="video_title 必填")
from mc.corpus import AIGenerator
ai = AIGenerator()
if not ai.available:
raise HTTPException(503, detail="AI 不可用(检查 config/ai.yaml 或 agent-local 配置)")
import asyncio as _asyncio
try:
loop = _asyncio.get_event_loop()
except RuntimeError:
loop = _asyncio.new_event_loop()
topic = loop.run_until_complete(ai.extract_topic(video_title))
logger.info(" 🎯 主题提炼: %s → %s", video_title[:24], topic)
return {"status": "ok", "topic": topic}
@router.post("/generate-path")
def api_generate_path(data: dict):
"""生成「引导路径」:起点(主题) → 终点(引导要素),步数自适应 max(3, 2+要素数) 封顶 6
Body: { topic: str, guide_items: [str]|str, outline?: str }
"""
topic = (data.get("topic") or "").strip()
outline = (data.get("outline") or "").strip()
if not topic:
raise HTTPException(400, detail="topic 必填")
guide_raw = data.get("guide_items") or []
if isinstance(guide_raw, str):
guide_items = [l.strip() for l in guide_raw.split("\n") if l.strip()]
else:
guide_items = [str(g).strip() for g in guide_raw if str(g).strip()]
from mc.corpus import AIGenerator
ai = AIGenerator()
if not ai.available:
raise HTTPException(503, detail="AI 不可用(检查 config/ai.yaml 或 agent-local 配置)")
import asyncio as _asyncio
try:
loop = _asyncio.get_event_loop()
except RuntimeError:
loop = _asyncio.new_event_loop()
path = loop.run_until_complete(ai.generate_guide_path(topic, guide_items, outline))
steps = len([l for l in (path or "").split("\n") if l.strip()])
logger.info(" 🧭 引导路径: %s + %d 要素 → %d 步", topic[:16], len(guide_items), steps)
return {"status": "ok", "path": path, "steps": steps}
@router.post("/parse-outline")
def api_parse_outline(data: dict):
"""把用户的自然语言想法拆解成结构化「讨论走向」(供讨论模式生成用)
Body: { raw_idea: str } — 用户原始描述(可能跳跃/口语)
"""
raw_idea = (data.get("raw_idea") or "").strip()
if not raw_idea:
raise HTTPException(400, detail="raw_idea 必填")
from mc.corpus import AIGenerator
ai = AIGenerator()
if not ai.available:
raise HTTPException(503, detail="AI 不可用(检查 config/ai.yaml 或 agent-local 配置)")
import asyncio as _asyncio
try:
loop = _asyncio.get_event_loop()
except RuntimeError:
loop = _asyncio.new_event_loop()
outline = loop.run_until_complete(ai.parse_discussion_outline(raw_idea))
logger.info(" 🧭 走向拆解: %s → %d 字", raw_idea[:24], len(outline or ""))
return {"status": "ok", "outline": outline}
@router.post("/save-comments")
def api_save_comments(data: dict):
"""将精选评论保存到语料库
Body: {
"comments": [{"text": "...", "role": "...", "category": "..."}],
"platform": "douyin"
}
"""
comments = data.get("comments", [])
platform = data.get("platform", "douyin")
if not comments:
raise HTTPException(400, detail="comments 必填")
mgr = _get_corpus_mgr()
saved = 0
for c in comments:
text = c.get("text", "").strip()
role = c.get("role", "filler")
cat = c.get("category", "") or "入库评论"
if not text:
continue
# 入库到指定分类(不存在则自动创建)
mgr.add_comment(cat, text, platform)
saved += 1
logger.info(" ✅ 入库 %d 条评论 → %s/%s", saved, platform, cat)
return {"status": "ok", "saved": saved}