gbanyan
bc281b8e0a
Implement a new Deduplication Agent that identifies and groups similar transformation descriptions. Supports two deduplication methods: - Embedding: Fast vector similarity comparison using cosine similarity - LLM: Accurate pairwise semantic comparison (slower but more precise) Backend changes: - Add deduplication router with /deduplicate endpoint - Add embedding_service for vector-based similarity - Add llm_deduplication_service for LLM-based comparison - Improve expert_transformation error handling and progress reporting Frontend changes: - Add DeduplicationPanel with interactive group visualization - Add useDeduplication hook for state management - Integrate deduplication tab in main App - Add threshold slider and method selector in sidebar 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
105 lines
3.5 KiB
Python
105 lines
3.5 KiB
Python
"""Expert Transformation Agent 提示詞模組"""
|
||
|
||
from typing import List, Optional
|
||
|
||
|
||
def get_expert_generation_prompt(
|
||
query: str,
|
||
categories: List[str],
|
||
expert_count: int,
|
||
custom_experts: Optional[List[str]] = None
|
||
) -> str:
|
||
"""Step 0: 生成專家團隊（不依賴主題，純隨機多元）"""
|
||
import time
|
||
import random
|
||
|
||
custom_text = ""
|
||
if custom_experts and len(custom_experts) > 0:
|
||
custom_text = f"（已指定：{', '.join(custom_experts[:expert_count])}）"
|
||
|
||
# 加入時間戳和隨機數來增加多樣性
|
||
seed = int(time.time() * 1000) % 10000
|
||
diversity_hints = [
|
||
"冷門、非主流、跨領域",
|
||
"罕見職業、新興領域、邊緣學科",
|
||
"非傳統、創新、小眾專業",
|
||
"未來趨向、實驗性、非常規",
|
||
"跨文化、混合領域、獨特視角"
|
||
]
|
||
hint = random.choice(diversity_hints)
|
||
|
||
return f"""/no_think
|
||
隨機組建 {expert_count} 個來自完全不同領域的專家團隊{custom_text}。
|
||
|
||
【創新要求】（隨機種子:{seed}）
|
||
- 優先選擇{hint}的專家
|
||
- 避免常見職業（如醫生、工程師、教師、律師等）
|
||
- 每個專家必須來自完全不相關的領域
|
||
- 越罕見、越創新越好
|
||
|
||
回傳 JSON：
|
||
{{"experts": [{{"id": "expert-0", "name": "職業", "domain": "領域", "perspective": "角度"}}, ...]}}
|
||
|
||
規則：
|
||
- id 為 expert-0 到 expert-{expert_count - 1}
|
||
- name 填寫職業名稱（非人名），2-5字
|
||
- domain 要具體且獨特，不可重複類型"""
|
||
|
||
|
||
def get_expert_keyword_generation_prompt(
|
||
category: str,
|
||
attribute: str,
|
||
experts: List[dict], # List[ExpertProfile]
|
||
keywords_per_expert: int = 1
|
||
) -> str:
|
||
"""Step 1: 專家視角關鍵字生成"""
|
||
# 建立專家列表，格式更清晰
|
||
experts_list = "\n".join([f"- {exp['id']}: {exp['name']}" for exp in experts])
|
||
|
||
return f"""/no_think
|
||
你需要扮演以下專家，為屬性生成創新關鍵字：
|
||
|
||
【專家名單】
|
||
{experts_list}
|
||
|
||
【任務】
|
||
屬性：「{attribute}」（類別：{category}）
|
||
|
||
請為每位專家：
|
||
1. 先理解該職業的專業背景、知識領域、工作內容
|
||
2. 從該職業的獨特視角思考「{attribute}」
|
||
3. 生成 {keywords_per_expert} 個與該專業相關的創新關鍵字（2-6字）
|
||
|
||
關鍵字必須反映該專家的專業思維方式，例如：
|
||
- 會計師 看「移動」→「資金流動」「成本效益」
|
||
- 建築師 看「移動」→「動線設計」「空間流動」
|
||
- 心理師 看「移動」→「行為動機」「情緒轉變」
|
||
|
||
回傳 JSON：
|
||
{{"keywords": [{{"keyword": "詞彙", "expert_id": "expert-X", "expert_name": "名稱"}}, ...]}}
|
||
|
||
共需 {len(experts) * keywords_per_expert} 個關鍵字，每個關鍵字必須明顯與對應專家的專業領域相關。"""
|
||
|
||
|
||
def get_single_description_prompt(
|
||
query: str,
|
||
keyword: str,
|
||
expert_id: str,
|
||
expert_name: str,
|
||
expert_domain: str
|
||
) -> str:
|
||
"""Step 2: 為單一關鍵字生成描述"""
|
||
# 如果 domain 是通用的，就只用職業名稱
|
||
domain_text = f"（{expert_domain}領域）" if expert_domain and expert_domain != "Professional Field" else ""
|
||
|
||
return f"""/no_think
|
||
你是一位{expert_name}{domain_text}。
|
||
|
||
任務：為「{query}」生成一段創新應用描述。
|
||
關鍵字：{keyword}
|
||
|
||
從你的專業視角，說明如何將「{keyword}」的概念應用到「{query}」上。描述要具體、有創意，15-30字。
|
||
|
||
只回傳 JSON，不要其他文字：
|
||
{{"description": "你的創新應用描述"}}"""
|