Agent/insert_knowledge_direct.py

import os
import sys
import json
import uuid
from datetime import datetime, timezone
from pathlib import Path

# 添加项目路径
sys.path.insert(0, '/root/Agent')

os.environ['OPEN_ROUTER_API_KEY'] = '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'
os.environ['QWEN_API_KEY'] = 'sk-9453c827b9e14108b53d2b30ef7c75fe'
os.environ['KNOWHUB_DB'] = 'gp-t4n72471pkmt4b9q7o-master.gpdbmaster.singapore.rds.aliyuncs.com'
os.environ['KNOWHUB_PORT'] = '5432'
os.environ['KNOWHUB_USER'] = 'aiddit_aigc'
os.environ['KNOWHUB_PASSWORD'] = '%a&&yqNxg^V1$toJ*WOa^-b^X=QJ'
os.environ['KNOWHUB_DB_NAME'] = 'knowhub'

from knowhub.knowhub_db.pg_store import PostgreSQLStore
from knowhub.embeddings import get_embedding

# 知识数据
knowledge_data = {
    "task": "人物近景肖像",
    "content": "生成人物近景半身或胸部以上画面、突出面部表情和情绪、背景虚化的核心工序要素包括：\n\n1. **近景/半身构图**：使用 tight medium close-up、extreme close-up、中近景半身等提示词；推荐 85mm 镜头视角；构图比例 2:3 或 9:16（竖版）、3:4（职业照）\n\n2. **面部表情细节**：使用锁脸技术（nano-banana 锁脸功能）、微表情描述（slightly parted lips、calm introspective expression）、多表情生成（3×3 网格生成 9 种表情）；关键提示词：direct steady eye contact、subtle catchlights in eyes\n\n3. **背景虚化效果**：推荐参数 85mm 镜头 + f/1.4-f/1.8 光圈；提示词：shallow depth of field、extremely shallow depth of field、soft bokeh blur、blurring out the neck, ear, and background completely；后期工具：Photoshop Lens Blur、DxO Photolab 9 Bokeh\n\n4. **皮肤质感真实**：避免塑料感的关键提示词：realistic skin texture with visible pores and natural imperfections、no plastic smoothing、dewy smooth texture；修复工具：Enhancor.ai Skin Realism\n\n5. **眼神互动感**：direct eye contact、catchlights in eyes、gentle and engaging eyes、looking directly at the camera",
    "types": ["strategy"],
    "tags": {
        "domain": "AI 生图",
        "source": "多渠调研汇总",
        "task": "人物近景肖像"
    },
    "score": 5,
    "source": {"category": "research"}
}

print("正在连接数据库...")
store = PostgreSQLStore()

# 生成 ID
now = datetime.now(timezone.utc)
knowledge_id = f"knowledge-{now.strftime('%Y%m%d')}-{uuid.uuid4().hex[:8]}"
message_id = f"msg-{uuid.uuid4().hex[:12]}"

# 生成 embedding
print("正在生成 embedding...")
task_embedding = get_embedding(knowledge_data["task"])
content_embedding = get_embedding(knowledge_data["content"])

# 构建知识记录
knowledge = {
    "id": knowledge_id,
    "message_id": message_id,
    "task": knowledge_data["task"],
    "content": knowledge_data["content"],
    "types": knowledge_data["types"],
    "tags": knowledge_data["tags"],
    "tag_keys": list(knowledge_data["tags"].keys()),
    "scopes": [],
    "owner": "librarian",
    "source": knowledge_data["source"],
    "eval": {"score": knowledge_data["score"]},
    "created_at": now.isoformat(),
    "updated_at": now.isoformat(),
    "status": "approved",
    "task_embedding": task_embedding,
    "content_embedding": content_embedding,
    "requirement_ids": [],
    "capability_ids": [],
    "tool_ids": [],
    "resource_ids": [],
}

print(f"正在插入知识：{knowledge_id}")
print(f"  Task: {knowledge['task']}")
print(f"  Types: {knowledge['types']}")
print(f"  Tags: {knowledge['tags']}")

store.insert(knowledge)

print(f"\n✓ 知识已成功入库!")
print(f"  ID: {knowledge_id}")
print(f"  Message ID: {message_id}")

# 验证插入
retrieved = store.get_by_id(knowledge_id)
if retrieved:
    print(f"\n✓ 验证成功：已检索到刚插入的知识")
    print(f"  Task: {retrieved['task']}")
    print(f"  Content 长度：{len(retrieved['content'])} 字符")
else:
    print(f"\n✗ 验证失败：无法检索到刚插入的知识")