Build Your First AI Agent with Eino ADK: Excel Agent End-to-End

// Full code: https://github.com/cloudwego/eino/blob/main/adk/prebuilt/planexecute/plan_execute.go

// NewPlanner creates a new planner agent based on the provided configuration.
func NewPlanner(_ context.Context, cfg *PlannerConfig) (adk.Agent, error)

// NewExecutor creates a new executor agent.
func NewExecutor(ctx context.Context, cfg *ExecutorConfig) (adk.Agent, error)

// NewReplanner creates a new replanner agent.
func NewReplanner(_ context.Context, cfg *ReplannerConfig) (adk.Agent, error)

// New creates a new plan-execute-replan agent with the given configuration.
func New(ctx context.Context, cfg *Config) (adk.Agent, error)

type Plan interface {
    // FirstStep returns the first step to be executed in the plan.
    FirstStep() string
    // Marshaler serializes the Plan into JSON.
    // The resulting JSON can be used in prompt templates.
    json.Marshaler
    // Unmarshaler deserializes JSON content into the Plan.
    // This processes output from structured chat models or tool calls into the Plan structure.
    json.Unmarshaler
}

### Task Plan
- [x] 1. Read the contents of '模拟出题.csv' from the working directory into a pandas DataFrame.
- [x] 2. Identify the question type (e.g., multiple-choice, short-answer) for each row in the DataFrame.
- [x] 3. For non-short-answer questions, restructure the data to place question, answer, explanation, and options in the same row.
- [x] 4. For short-answer questions, merge the answer content into the explanation column and ensure question and merged explanation are in the same row.
- [x] 5. Verify that all processed rows have question, answer (where applicable), explanation, and options (where applicable) in a single row with consistent formatting.
- [x] 6. Generate a cleaned report presenting the formatted questions with all relevant components (question, answer, explanation, options) in unified rows.

import github.com/cloudwego/eino/adk

// Execute plan → search → write once
sequential := adk.NewSequentialAgent(ctx, &adk.SequentialAgentConfig{
    Name: "research_pipeline",
    SubAgents: []adk.Agent{
        planAgent,    // plan research
        searchAgent,  // search materials
        writeAgent,   // write report
    },
})

import github.com/cloudwego/eino/adk

// Loop 5 times: analyze → improve → validate
loop := adk.NewLoopAgent(ctx, &adk.LoopAgentConfig{
    Name: "iterative_optimization",
    SubAgents: []adk.Agent{
        analyzeAgent,  // analyze current status
        improveAgent,  // propose improvements
        validateAgent, // validate improvements
    },
    MaxIterations: 5,
})

import github.com/cloudwego/eino/adk

// Concurrent: sentiment + keywords + summary
parallel := adk.NewParallelAgent(ctx, &adk.ParallelAgentConfig{
    Name: "multi_analysis",
    SubAgents: []adk.Agent{
        sentimentAgent,  // sentiment analysis
        keywordAgent,    // keyword extraction
        summaryAgent,    // content summarization
    },
})

type Agent interface {
    Name(ctx context.Context) string
    Description(ctx context.Context) string
    Run(ctx context.Context, input *AgentInput, options ...AgentRunOption) *AsyncIterator[*AgentEvent]
}

type AgentInput struct {
    Messages        []Message
    EnableStreaming bool
}

type Message = *schema.Message

type AgentEvent struct {
    AgentName string          // Agent name (auto-filled by framework)
    RunPath   []RunStep       // Full run path to reach current Agent (auto-filled)
    Output    *AgentOutput    // Agent output message content
    Action    *AgentAction    // Agent action event content
    Err       error           // Agent error
}

type AgentOutput struct {
    MessageOutput  *MessageVariant // Model message output
    CustomizedOutput any           // Customized output content
}

type MessageVariant struct {
    IsStreaming bool            // Whether streaming output
    Message       Message       // Non-streaming message output
    MessageStream MessageStream // Streaming message output
    Role schema.RoleType        // Message role
    ToolName string             // Tool name
}

type AgentAction struct {
    Exit bool                               // Agent exit
    Interrupted *InterruptInfo              // Agent interrupted
    TransferToAgent *TransferToAgentAction  // Agent transfer
    CustomizedAction any                    // Customized Agent action
}

iter := myAgent.Run(ctx, "hello")
for {
    event, ok := iter.Next()
    if !ok { break }
    // handle event
}

func GetSessionValues(ctx context.Context) map[string]any
func GetSessionValue(ctx context.Context, key string) (any, bool)
func AddSessionValue(ctx context.Context, key string, value any)
func AddSessionValues(ctx context.Context, kvs map[string]any)
func WithSessionValues(v map[string]any) AgentRunOption

func SetSubAgents(ctx context.Context, agent Agent, subAgents []Agent) (Agent, error)
func NewTransferToAgentAction(destAgentName string) *AgentAction

func NewAgentTool(_ context.Context, agent Agent, options ...AgentToolOption) tool.BaseTool

import pandas as pd

df = pd.read_csv('questions.csv')
first_column = df.iloc[:, _0_]
first_column.to_csv('extracted_first_column.csv', index=_False_)

{
  "is_success": true,
  "result": "Successfully extracted the first column from questions.csv and saved it to first_column.csv.",
  "files": [
    {
      "path": "/User/user/go/src/github.com/cloudwego/eino-examples/adk/multiagent/integration-excel-agent/playground/00f118af-4bd8-42f7-8d11-71f2801218bd/first_column.csv",
      "desc": "A CSV file containing only the first column data from the original questions.csv."
    }
  ]
}

name: Planner
answer: {
    "steps": [
        { "index": 1, "desc": "Read the 'questions.csv' file into a pandas DataFrame." },
        { "index": 2, "desc": "Extract the first column from the DataFrame." },
        { "index": 3, "desc": "Save the extracted first column to a new CSV file." }
    ]
}

name: Executor
tool name: CodeAgent
arguments: {"request":"Read the 'questions.csv' file into a pandas DataFrame using pandas. Use the pandas.read_csv function and store the result in a variable named df."}

name: Executor
tool name: PythonRunner
arguments: {"code":"```python\nfirst_column = df.iloc[:, 0]\n```"}

name: Executor
tool response: Traceback (most recent call last):
  File "/User/user/go/src/github.com/cloudwego/eino-examples/adk/multiagent/integration-excel-agent/playground/00f118af-4bd8-42f7-8d11-71f2801218bd/00f118af-4bd8-42f7-8d11-71f2801218bd.py", line 1, in <module>
    first_column = df.iloc[:, 0]
                   ^^
NameError: name 'df' is not defined

  name: Executor
  answer: The error occurs because the DataFrame `df` is not defined. We need to first load the data from the existing CSV file `questions.csv` into `df`. Here's the corrected code:
  tool name: PythonRunner
  arguments: {"code":"```python\nimport pandas as pd\ndf = pd.read_csv('questions.csv')\nfirst_column = df.iloc[:, 0]\nprint(first_column.head())  # Verify the result\n```"}

name: Replanner
answer: { "is_success": true, "result": "Successfully extracted the first column from 'questions.csv' into the new CSV file 'extracted_first_column.csv'.", "files": [ { "desc": "A new CSV file containing the first column from the original table", "path": "extracted_first_column.csv" } ] }

name: Report
tool name: SubmitResult
arguments: { "is_success": true, "result": "Successfully extracted the first column from questions.csv and saved it to first_column.csv.", "files": [ { "path": "/User/user/go/src/github.com/cloudwego/eino-examples/adk/multiagent/integration-excel-agent/playground/00f118af-4bd8-42f7-8d11-71f2801218bd/first_column.csv", "desc": "A CSV file containing only the first column data from the original questions.csv." } ] }