Your First Agent

“Hello World” Agent Development Based on Dora

English | 简体中文

Step 1: Install the Development and Runtime Environment

Refer to the Installation Guide to complete the setup of the development and runtime environment.

Step 2: Obtain the Intelligent Agent Template

  1. Go to the Dora Intelligent Agent Template Repository.
  2. Select the simplest Reasoner Template.
  3. Copy the template to your development directory.
  4. Review the template description: README.

Step 3: Configuration File Setup

Configuration File Overview

Create or edit the reasoner_agent.yml file:

AGENT:
  ROLE: Knowledgeable Assistant
  BACKSTORY: <Your background description>
  TASK: null  # Specific task

RAG:
  RAG_ENABLE: false
  MODULE_PATH: null
  RAG_MODEL_NAME: text-embedding-3-small
  COLLECTION_NAME: mofa
  IS_UPLOAD_FILE: true
  CHROMA_PATH: ./data/output/chroma_store
  FILES_PATH:
    - ./data/output/arxiv_papers
  ENCODING: utf-8
  CHUNK_SIZE: 256
  RAG_SEARCH_NUM: 2

WEB:
  WEB_ENABLE: false
  SERPER_API_KEY: <Your Serper API key>
  SEARCH_NUM: 20
  SEARCH_ENGINE_TIMEOUT: 5

MODEL:
  MODEL_API_KEY: <Your model API key>
  MODEL_NAME: gpt-4o-mini
  MODEL_MAX_TOKENS: 2048

ENV:
  PROXY_URL: null
  AGENT_TYPE: reasoner

LOG:
  LOG_PATH: ./data/output/log/log.md
  LOG_TYPE: markdown
  LOG_STEP_NAME: reasoner_result
  CHECK_LOG_PROMPT: true

Configuration Details

1. AGENT Module

  • ROLE: Name of the assistant role.
  • BACKSTORY: Background description of the assistant.
  • TASK: Specific task (default is null).

2. RAG Module

  • RAG_ENABLE: Enable (true) or disable (false) RAG.
  • Other Parameters: Configure knowledge retrieval enhancement features.

3. WEB Module

  • WEB_ENABLE: Enable (true) or disable (false) web search.
  • SERPER_API_KEY: Serper search API key.

4. MODEL Module

  • MODEL_API_KEY: API key for the model service.
  • MODEL_NAME: Model name to use (e.g., gpt-4o-mini).
  • MODEL_MAX_TOKENS: Maximum number of tokens the model can generate.

5. ENV Module

  • PROXY_URL: Proxy server URL (set to null if no proxy is needed).
  • AGENT_TYPE: Agent type, e.g., reasoner.

6. LOG Module

  • LOG_PATH: Path to the log file.
  • LOG_TYPE: Log format (e.g., markdown).
  • LOG_STEP_NAME: Log step name.
  • CHECK_LOG_PROMPT: Enable log prompt checking (true or false).

Step 4: Configure Dora Operator

Create a reasoner_agent.py script:

import os
from dora import DoraStatus
import pyarrow as pa
from mofa.kernel.utils.util import load_agent_config, create_agent_output
from mofa.run.run_agent import run_dspy_or_crewai_agent
from mofa.utils.files.dir import get_relative_path
from mofa.utils.log.agent import record_agent_result_log


class Operator:
    """
    Dora-rs Operator for handling INPUT events, loading configurations, running the agent, logging results, and sending outputs.
    """

    def on_event(self, dora_event, send_output) -> DoraStatus:
        if dora_event.get("type") == "INPUT":
            agent_inputs = ['data', 'task']
            event_id = dora_event.get("id")

            if event_id in agent_inputs:
                task = dora_event["value"][0].as_py()

                yaml_file_path = get_relative_path(
                    current_file=__file__,
                    sibling_directory_name='configs',
                    target_file_name='reasoner_agent.yml'
                )

                inputs = load_agent_config(yaml_file_path)
                inputs["task"] = task

                agent_result = run_dspy_or_crewai_agent(agent_config=inputs)

                log_step_name = inputs.get('log_step_name', "Step_one")
                record_agent_result_log(
                    agent_config=inputs,
                    agent_result={f"1, {log_step_name}": {task: agent_result}}
                )

                output_data = create_agent_output(
                    agent_name='keyword_results',
                    agent_result=agent_result,
                    dataflow_status=os.getenv('IS_DATAFLOW_END', True)
                )

                send_output(
                    "reasoner_result",
                    pa.array([output_data]),
                    dora_event.get('metadata', {})
                )

                print('reasoner_results:', agent_result)

        return DoraStatus.CONTINUE

Step 5: Configure Dora Dataflow

Create or edit the reasoner_dataflow.yml file:

nodes:

  - id: terminal-input
    build: pip install -e ../../../node-hub/terminal-input
    path: dynamic
    outputs:
      - data
    inputs:
      reasoner_results: reasoner-agent/reasoner_results

  - id: reasoner-agent
    operator:
      python: scripts/reasoner_agent.py
      inputs:
        task: terminal-input/data
      outputs:
        - reasoner_results

Node Descriptions

  • terminal-input:
    • Function: Handles initial input.
    • Action: Installs the terminal-input module.
    • Output: Generates data, passing it to reasoner-agent.
    • Input: Receives reasoner_results.
  • reasoner-agent:
    • Function: Processes tasks and generates results.
    • Action: Runs the reasoner_agent.py script.
    • Input: Receives data from terminal-input as task.
    • Output: Generates reasoner_results, sending them back to terminal-input.

Step 6: Run Dora Dataflow

Start the Dataflow Using Dora-RS CLI

  1. Start the coordinator
dora up
  1. Build the dataflow
dora build reasoner_dataflow.yml
  1. Start the dataflow with auto-attach
dora start reasoner_dataflow.yml --attach
  1. In another terminal, input tasks to the agent
terminal-input

Then type your queries to interact with the agent.

Summary

This “Hello World” example demonstrates:

  • How to set up a basic MoFA agent using templates
  • Configuration file structure and parameters
  • Integration with Dora-RS dataflow
  • Running and interacting with the agent

This serves as the foundation for building more complex agents and understanding the MoFA framework.