Access Multiple Previous Steps Output

This example demonstrates **Workflows 2.0** advanced data flow capabilities

This example demonstrates Workflows 2.0 shows how to:

Access outputs from specific named steps (get_step_content())
Aggregate all previous outputs (get_all_previous_content())
Create comprehensive reports by combining multiple research sources

Key Features:

Step Output Access: Retrieve data from any previous step by name or collectively.
Custom Reporting: Combine and analyze outputs from parallel or sequential steps.
Streaming Support: Real-time updates during execution.

1from kern.agent.agent import Agent
2from kern.tools.hackernews import HackerNewsTools
3from kern.tools.yfinance import YFinanceTools
4from kern.workflow.step import Step
5from kern.workflow.types import StepInput, StepOutput
6from kern.workflow.workflow import Workflow
7
8# Define the research agents
9hackernews_agent = Agent(
10    name="HackerNews Researcher",
11    instructions="You are a researcher specializing in finding the latest tech news and discussions from Hacker News. Focus on startup trends, programming topics, and tech industry insights.",
12    tools=[HackerNewsTools()],
13)
14
15finance_agent = Agent(
16    name="Finance Researcher",
17    instructions="You are a finance researcher. Search for stock data, market trends, and financial information to gather detailed insights.",
18    tools=[YFinanceTools()],
19)
20
21reasoning_agent = Agent(
22    name="Reasoning Agent",
23    instructions="You are an expert analyst who creates comprehensive reports by analyzing and synthesizing information from multiple sources. Create well-structured, insightful reports.",
24)
25
26# Create the research steps
27research_hackernews = Step(
28    name="research_hackernews",
29    agent=hackernews_agent,
30    description="Research latest tech trends from Hacker News",
31)
32
33research_finance = Step(
34    name="research_finance",
35    agent=finance_agent,
36    description="Research financial data and market trends",
37)
38
39# Custom function step that has access to ALL previous step outputs
40
41
42def create_comprehensive_report(step_input: StepInput) -> StepOutput:
43    """
44    Custom function that creates a report using data from multiple previous steps.
45    This function has access to ALL previous step outputs and the original workflow message.
46    """
47
48    # Access original workflow input
49    original_topic = step_input.input or ""
50
51    # Access specific step outputs by name
52    hackernews_data = step_input.get_step_content("research_hackernews") or ""
53    finance_data = step_input.get_step_content("research_finance") or ""
54
55    # Or access ALL previous content
56    _ = step_input.get_all_previous_content()
57
58    # Create a comprehensive report combining all sources
59    report = f"""
60        # Comprehensive Research Report: {original_topic}
61
62        ## Executive Summary
63        Based on research from HackerNews and web sources, here's a comprehensive analysis of {original_topic}.
64
65        ## HackerNews Insights
66        {hackernews_data[:500]}...
67
68        ## Finance Research Findings
69        {finance_data[:500]}...
70    """
71
72    return StepOutput(
73        step_name="comprehensive_report", content=report.strip(), success=True
74    )
75
76
77comprehensive_report_step = Step(
78    name="comprehensive_report",
79    executor=create_comprehensive_report,
80    description="Create comprehensive report from all research sources",
81)
82
83# Final reasoning step using reasoning agent
84reasoning_step = Step(
85    name="final_reasoning",
86    agent=reasoning_agent,
87    description="Apply reasoning to create final insights and recommendations",
88)
89
90workflow = Workflow(
91    name="Enhanced Research Workflow",
92    description="Multi-source research with custom data flow and reasoning",
93    steps=[
94        research_hackernews,
95        research_finance,
96        comprehensive_report_step,  # Has access to both previous steps
97        reasoning_step,  # Gets the last step output (comprehensive report)
98    ],
99)
100
101if __name__ == "__main__":
102    workflow.print_response(
103        "Latest developments in artificial intelligence and machine learning",
104        stream=True,
105    )