104 lines
3.1 KiB
Python
104 lines
3.1 KiB
Python
"""Researcher Agent — domain-aware, bilingual research."""
|
||
|
||
from __future__ import annotations
|
||
|
||
from typing import Any
|
||
|
||
from .base import BaseAgent
|
||
from app.config import settings
|
||
|
||
SYSTEM_EN = """\
|
||
You are a senior industry analyst at a top-tier consulting firm.
|
||
Your task is to produce a thorough research brief based on the given instructions.
|
||
|
||
Requirements:
|
||
1. Be specific — cite concrete data points, market sizes, growth rates, company names
|
||
2. Be structured — organize findings with clear headings and logical flow
|
||
3. Be analytical — don't just list facts, provide insights and implications
|
||
4. Flag data gaps — explicitly note where data is uncertain or unavailable
|
||
|
||
Output (JSON):
|
||
{
|
||
"title": "Research brief title",
|
||
"executive_summary": "2-3 sentence summary of key findings",
|
||
"sections": [
|
||
{
|
||
"heading": "Section heading",
|
||
"content": "Detailed findings (Markdown)",
|
||
"data_points": ["key data points extracted"],
|
||
"sources_quality": "high|medium|low — how confident are you in the data"
|
||
}
|
||
],
|
||
"data_gaps": ["areas where data is insufficient or uncertain"],
|
||
"key_insights": ["top 3-5 non-obvious insights"]
|
||
}"""
|
||
|
||
SYSTEM_ZH = """\
|
||
你是一位顶级咨询公司的资深行业分析师。
|
||
你的任务是根据给定的指令，输出一份深度研究简报。
|
||
|
||
要求：
|
||
1. 具体——引用具体的数据点、市场规模、增长率、企业名称
|
||
2. 结构化——用清晰的标题和逻辑流组织发现
|
||
3. 有分析深度——不要只罗列事实，要提供洞察和含义
|
||
4. 标注数据缺口——明确指出数据不确定或不可获取的地方
|
||
|
||
输出（JSON）：
|
||
{
|
||
"title": "研究简报标题",
|
||
"executive_summary": "核心发现的2-3句总结",
|
||
"sections": [
|
||
{
|
||
"heading": "章节标题",
|
||
"content": "详细发现（Markdown格式）",
|
||
"data_points": ["提取的关键数据点"],
|
||
"sources_quality": "high|medium|low — 对数据的置信度"
|
||
}
|
||
],
|
||
"data_gaps": ["数据不充分或不确定的领域"],
|
||
"key_insights": ["3-5条非显而易见的洞察"]
|
||
}"""
|
||
|
||
|
||
class ResearcherAgent(BaseAgent):
|
||
name = "researcher"
|
||
description = "域感知研究 — 根据领域选择最优模型和语言"
|
||
|
||
def __init__(self, model: str | None = None, language: str = "en"):
|
||
super().__init__(model=model)
|
||
self.language = language
|
||
self.system_prompt = SYSTEM_ZH if language == "zh" else SYSTEM_EN
|
||
|
||
async def run(self, context: dict[str, Any]) -> dict[str, Any]:
|
||
requirement = context["requirement"]
|
||
report_type = context.get("report_type", "")
|
||
extra_data = context.get("extra_data", "")
|
||
|
||
if self.language == "zh":
|
||
prompt = f"""\
|
||
## 研究指令
|
||
{requirement}
|
||
|
||
## 研究方向
|
||
{report_type}
|
||
|
||
## 补充数据
|
||
{extra_data if extra_data else "（无）"}
|
||
|
||
请输出研究简报 JSON。"""
|
||
else:
|
||
prompt = f"""\
|
||
## Research instructions
|
||
{requirement}
|
||
|
||
## Research focus
|
||
{report_type}
|
||
|
||
## Additional data
|
||
{extra_data if extra_data else "(none)"}
|
||
|
||
Output the research brief as JSON."""
|
||
|
||
result = await self.call_llm_json(prompt, max_tokens=6144)
|
||
return {"research": result}
|