AGI｜研究报告还能这样写？揭秘Open Deep Research智能生成全流程

Part1

工作流模式

# 主工作流图构建
builder = StateGraph(ReportState, input=ReportStateInput, output=ReportStateOutput)
# 智能规划节点
builder.add_node("generate_report_plan", generate_report_plan)
# 人工反馈节点
builder.add_node("human_feedback", human_feedback)
# 章节研究子图
builder.add_node("build_section_with_web_research", section_builder.compile())
builder.add_node("gather_completed_sections", gather_completed_sections)
builder.add_node("write_final_sections", write_final_sections)
builder.add_node("compile_final_report", compile_final_report)

async def generate_report_plan(state: ReportState, config: RunnableConfig):
    topic = state["topic"]
    feedback_list = state.get("feedback_on_report_plan", [])
    
    # 获取配置参数
    configurable = Configuration.from_runnable_config(config)
    report_structure = configurable.report_structure
    number_of_queries = configurable.number_of_queries
    
    # 生成规划查询
    structured_llm = writer_model.with_structured_output(Queries)
    system_instructions = report_planner_query_writer_instructions.format(
        topic=topic,
        report_organization=report_structure,
        number_of_queries=number_of_queries,
        today=get_today_str()
    )
    
    # 执行搜索和规划
    results = await structured_llm.ainvoke([
        SystemMessage(content=system_instructions),
        HumanMessage(content="Generate search queries for planning.")
    ])
    
    # 网络搜索获取背景信息
    query_list = [query.search_query for query in results.queries]
    source_str = await select_and_execute_search(search_api, query_list, params_to_pass)
    
    # 生成报告章节结构
    planner_llm = init_chat_model(model=planner_model, model_provider=planner_provider)
    structured_llm = planner_llm.with_structured_output(Sections)
    report_sections = await structured_llm.ainvoke([
        SystemMessage(content=system_instructions_sections),
        HumanMessage(content=planner_message)
    ])
    
    return {"sections": report_sections.sections}

def human_feedback(state: ReportState, config: RunnableConfig) -> Command:
    sections = state['sections']
    sections_str = "\n\n".join(
        f"Section: {section.name}\n"
        f"Description: {section.description}\n"
        f"Research needed: {'Yes' if section.research else 'No'}\n"
        for section in sections
    )
    
    interrupt_message = f"""Please provide feedback on the following report plan:
                        \n\n{sections_str}\n
                        \nDoes the report plan meet your needs?
                        Pass 'true' to approve or provide feedback to regenerate:"""
    
    feedback = interrupt(interrupt_message)
    
    if isinstance(feedback, bool) and feedback isTrue:
        # 批准后启动并行章节研究
        return Command(goto=[
            Send("build_section_with_web_research", {
                "topic": topic, 
                "section": s, 
                "search_iterations": 0
            }) for s in sections if s.research
        ])
    elif isinstance(feedback, str):
        # 反馈后重新规划
        return Command(
            goto="generate_report_plan", 
            update={"feedback_on_report_plan": [feedback]}
        )

# 章节研究子图
section_builder = StateGraph(SectionState, output=SectionOutputState)
section_builder.add_node("generate_queries", generate_queries)
section_builder.add_node("search_web", search_web)
section_builder.add_node("write_section", write_section)

section_builder.add_edge(START, "generate_queries")
section_builder.add_edge("generate_queries", "search_web")
section_builder.add_edge("search_web", "write_section")

async def write_section(state: SectionState, config: RunnableConfig) -> Command:
    # 生成章节内容
    section_content = await writer_model.ainvoke([
        SystemMessage(content=section_writer_instructions),
        HumanMessage(content=section_writer_inputs_formatted)
    ])
    
    section.content = section_content.content
    
    # 质量评估
    reflection_model = init_chat_model(model=planner_model).with_structured_output(Feedback)
    feedback = await reflection_model.ainvoke([
        SystemMessage(content=section_grader_instructions_formatted),
        HumanMessage(content=section_grader_message)
    ])
    
    # 根据评估结果决定下一步
    if feedback.grade == "pass"or state["search_iterations"] >= max_search_depth:
        return Command(update={"completed_sections": [section]}, goto=END)
    else:
        return Command(
            update={"search_queries": feedback.follow_up_queries, "section": section},
            goto="search_web"
        )

Part2

多智能体模式

class ReportState(MessagesState):
    sections: list[str] # 报告章节列表
    completed_sections: Annotated[list[Section], operator.add] # 已完成章节
    final_report: str # 最终报告
    question_asked: bool # 是否已提问
    source_str: Annotated[str, operator.add] # 搜索来源

async def supervisor(state: ReportState, config: RunnableConfig):
    messages = state["messages"]
    configurable = Configuration.from_runnable_config(config)
    supervisor_model = get_config_value(configurable.supervisor_model)
    
    llm = init_chat_model(model=supervisor_model)
    
    # 如果章节研究完成但报告未完成，启动引言和结论撰写
    if state.get("completed_sections") andnot state.get("final_report"):
        research_complete_message = {
            "role": "user", 
            "content": "Research is complete. Now write introduction and conclusion..."
        }
        messages = messages + [research_complete_message]
    
    # 获取可用工具
    supervisor_tool_list = get_supervisor_tools(config)
    
    # 如果已经提问过，移除Question工具
    if state.get("question_asked", False):
        supervisor_tool_list = [tool for tool in supervisor_tool_list 
                               if tool.name != "Question"]
    
    llm_with_tools = llm.bind_tools(supervisor_tool_list, tool_choice="any")
    
    return {
        "messages": [
            await llm_with_tools.ainvoke([
                {"role": "system", "content": SUPERVISOR_INSTRUCTIONS.format(today=get_today_str())}
            ] + messages)
        ]
    }

async def research_Agent(state: SectionState, config: RunnableConfig):
    configurable = Configuration.from_runnable_config(config)
    researcher_model = get_config_value(configurable.researcher_model)
    
    llm = init_chat_model(model=researcher_model)
    research_tool_list = await get_research_tools(config)
    
    system_prompt = RESEARCH_INSTRUCTIONS.format(
        section_description=state["section"],
        number_of_queries=configurable.number_of_queries,
        today=get_today_str(),
    )
    
    # 添加MCP提示（如果配置了）
    if configurable.mcp_prompt:
        system_prompt += f"\n\n{configurable.mcp_prompt}"
    
    return {
        "messages": [
            await llm.bind_tools(research_tool_list, tool_choice="any").ainvoke([
                {"role": "system", "content": system_prompt}
            ] + state["messages"])
        ]
    }

def get_search_tool(config: RunnableConfig):
    configurable = Configuration.from_runnable_config(config)
    search_api = get_config_value(configurable.search_api)
    
    if search_api.lower() == "none":
        returnNone
    elif search_api.lower() == "tavily":
        search_tool = tavily_search
    elif search_api.lower() == "duckduckgo":
        search_tool = duckduckgo_search
    else:
        raise NotImplementedError(f"Search API '{search_api}' not supported")
    
    # 添加工具元数据
    tool_metadata = {**(search_tool.metadata or {}), "type": "search"}
    search_tool.metadata = tool_metadata
    return search_tool

asyncdef get_research_tools(config: RunnableConfig) -> list[BaseTool]:
    search_tool = get_search_tool(config)
    tools = [tool(Section), tool(FinishResearch)]
    
    if search_tool isnotNone:
        tools.append(search_tool)
    
    # 加载MCP工具
    existing_tool_names = {cast(BaseTool, tool).name for tool in tools}
    mcp_tools = await _load_mcp_tools(config, existing_tool_names)
    tools.extend(mcp_tools)
    
    return tools

Part3

使用示例

[
  {
    "name": "Introduction",
    "description": "Brief overview of the investor's perspective on Tesla's valuation changes, highlighting factors such as financial data, market performance, and future opportunities.",
    "research": false,
    "content": ""
  },
  {
    "name": "Analysis of Tesla's Financial Data",
    "description": "Detailed examination of Tesla's financial performance, including revenue, margins, profit, assets, and liabilities over recent years.",
    "research": true,
    "content": ""
  },
  {
    "name": "Market Performance of Tesla",
    "description": "Evaluation of Tesla's market performance, analyzing stock price trends, comparisons with benchmarks and peers, and investor sentiments.",
    "research": true,
    "content": ""
  },
  {
    "name": "Future Opportunities for Tesla",
    "description": "Exploration of potential future opportunities for Tesla, including new product launches, technological advancements, and strategic market expansions.",
    "research": true,
    "content": ""
  },
  {
    "name": "Conclusion",
    "description": "Summary of the financial performance, market performance, and future prospects of Tesla, including key takeaways for investors.",
    "research": false,
    "content": ""
  }
]

# 从投资者视角解析特斯拉估值变化：财务数据、市场表现与未来机遇

Tesla's valuation changes are pivotal from an investor’s standpoint, influenced by financial data, market performance, and future opportunities. The financial analysis reveals remarkable revenue growth and asset expansion, contrasted by recent profitability challenges. Market performance underscores Tesla's stock appreciation and dominance over benchmarks and peers, highlighting its competitive market stance. Future opportunities further emphasize expansion through innovative product launches and technological advancements, poised to drive sustainable growth. This report examines these dimensions to provide an integrated view of Tesla’s evolving valuation landscape.

## Analysis of Tesla's Financial Data

Tesla's financial performance over recent years indicates significant growth and profitability. Tesla's revenue has consistently increased from $31.5 billion in2020 to $97.7 billion in2024, marking a substantial rise and demonstrating strong market demand and expansion capabilities [1,2]. Despite fluctuations, Tesla maintained robust gross profits, achieving $17.4 billion in2024, which is slightly lower than the $20.8 billion in2023 but still indicative of a strong operational performance [3].

Concerning margins, Tesla’s operating income reached $7.1 billion in2024, while the net income stood at $7.13 billion, a notable drop from the $15 billion in2023, primarily linked to increased operating expenses and reduced income from non-operating sources [1]. The decline in profitability suggests pressures from higher expenses including research and development (R&D) that rose to $4.54 billion in2024 from $3.97 billion the previous year [3].

Tesla’s total assets expanded significantly, from $95.34 billion in2023 to $104.3 billion in2024, reflective of robust investments in infrastructure and expansion projects [4]. Additionally, Tesla’s liabilities also grew, indicating increasing financial obligations possibly due to strategic leveraging for expansion [5].

Analyzing ratios, Tesla shows robust liquidity with a current ratio of 2.00and a quick ratio of 1.38, as of June 2025, indicating healthy short-term financial stability [6]. The debt-to-equity ratio remained low at 0.17, exhibiting strong equity positioning and manageable debt levels [1]. 

Collectively, Tesla’s financial health and strategic investments position it well for future opportunities despite recent profitability challenges.

### Sources
[1] Tesla Financial Statements 2009-2025| TSLA - Macrotrends: https://www.macrotrends.net/stocks/charts/TSLA/tesla/financial-statements
[2] Tesla (TSLA) Financials - Income Statement - Stock Analysis: https://stockanalysis.com/stocks/tsla/financials/
[3] TSLA | Tesla Inc. Financial Statements - WSJ:https://www.wsj.com/market-data/quotes/TSLA/financials
[4] Tesla (TSLA) Financials 2025 - Income Statement and ... - MarketBeat:https://www.marketbeat.com/stocks/NASDAQ/TSLA/financials/
[5] Tesla, Inc. Annual Report:https://www.annualreports.com/Company/tesla-inc
[6] Tesla (TSLA) Financial Ratios - Stock Analysis:https://stockanalysis.com/stocks/tsla/financials/ratios/

## Market Performance of Tesla

Between 2023and2025, Tesla's stock price showed impressive growth, with closing prices surging from $281.95 to $348.68, reflecting a significant increase of 194.3% during this period [1]. Despite fluctuations, Tesla’s overall trend has been upward, underpinned by strong fundamentals and market sentiments.

Comparatively, Tesla has outperformed major indices such as the Dow Jones and NASDAQ. For instance, in 2024, Tesla’s stock returns were +62.52%, significantly higher than both Dow Jones (+12.88%) and NASDAQ (+28.64%) [2].

Tesla’s performance relative to its peers in the electric vehicle (EV) sector is also noteworthy. Over the past five years, it has consistently delivered higher average returns (+178.48%) compared to competitors like Toyota (+8.79%), BYD (+93.67%), and Ferrari (+23.54%) [2].

Market analyst sentiments towards Tesla are mixed. While Cathie Wood's Ark Invest targets an optimistic price of $2,600 by 2029, some analysts like Seth Goldstein of Morningstar maintain a more conservative fair value estimate of $210, albeit seeing long-term potential [3]. Overall, Tesla’s stock continues to be an attractive proposition for investors considering its robust growth prospects and competitive market position.

### Sources
[1] Tesla Stock Price 2023 To 2025| StatMuse Money: https://www.statmuse.com/money/ask/tesla-stock-price-2023-to-2025
[2] Tesla Stock Price Forecast 2025 - 2050 with Complete Analysis: https://futurevaluejournal.com/tesla-stock-price-forecast/
[3] Is Tesla Stock a Buy in 2025? 3-Year Performance Analysis and Forecast: https://www.techi.com/tesla-stock-buy-2025-analysis-forecast/

## Future Opportunities for Tesla

Tesla is poised to capitalize on several promising opportunities in the coming years, focusing on new product launches, technological advancements, and strategic market expansions. 

Regarding new product launches, Tesla plans to unveil the Model 2, an affordable electric vehicle tailored for a broader customer base. This vehicle will likely gain traction in urban areas and emerging economies, helping Tesla penetrate new markets with its cost-effective solutions [1,2]. Other notable products include the Tesla Van, which targets commercial and family use, and a compact, solar-powered Tiny House designed for sustainable living [1,2].

Technological advancements are central to Tesla’s strategy. The development of the 4680 battery cells, which offer higher energy density and lower production costs, will enhance vehicle range and performance, supporting Tesla’s drive to produce more efficient electric cars [2]. Tesla’s Full Self-Driving (FSD) software continues to evolve, aiming to achieve complete autonomy, which will significantly impact the market once regulatory hurdles are cleared [2,6].

Strategic market expansions are also in Tesla's roadmap. The company is accelerating its global expansion by establishing new manufacturing facilities, such as the Gigafactory in Mexico and potential sites in India and the Netherlands [6]. These plants will support increased production capacity, meeting global demand and reducing costs through localized production [6]. Additionally, Tesla's entry into Saudi Arabia aligns with economic diversification efforts, showcasing its commitment to sustainable technologies [6].

These initiatives position Tesla for sustained growth by diversifying its product offerings and leveraging advanced technologies to enhance performance and customer experience.

### Sources
[1] Tesla’s 2025 Innovation Lineup: Affordable Models and New Ventures: https://ilovetesla.com/teslas-2025-innovation-lineup-affordable-models-and-new-ventures/
[2] Elon Musk Confirms 3 NEW Teslas Models For 2025: https://elonbuzz.com/elon-musk-confirms-3-new-teslas-models-for-2025/
[6] Tesla’s Global Expansion: New Factories & Market Strategies For 2025: https://autotimesnews.com/teslas-global-expansion-new-factories-market-strategies-for-2025-2/

## Conclusion

Tesla has demonstrated substantial financial growth with revenues almost tripling from $31.5 billion in 2020 to $97.7 billion in 2024, showing robust market demand. Despite increased R&D expenses impacting net income, the company maintains solid liquidity and a low debt-to-equity ratio, ensuring financial stability. In the stock market, Tesla's performance from 2023 to 2025 outpaced major indices like Dow Jones and NASDAQ, amid a 194.3% rise in the stock price. Future opportunities are promising, with plans for affordable models like the Model 2, advances in battery technology, and global expansion through new Gigafactories.

### Key Financial Metrics and Market Performance

| Metric/Performance | 2023 |2024                        | Future Opportunities |
|------------------------|-----------------------------|-----------------------------|-----------------------------------|
| Revenue | $75.4 billion | $97.7 billion | New models & affordable EVs |
| Net Income | $15 billion | $7.13 billion | Advanced battery technologies |
| Stock Price | $281.95 | $348.68                     | Global expansion plans |
| Debt-to-Equity Ratio |0.17                        | 0.17 |                                    |
| Operating Income | $9.2 billion | $7.1 billion | |

Investors should note Tesla's aggressive expansion and innovative product pipeline, which position it for continued market leadership and growth.

Part4

总结与展望