hyunsoo's blog

agent piplinefull codeYouTube Shorts Maker - Complete Code CollectionThis document contains all Python code files from the project, organized by file path for easy copy-paste.youtube_shorts_maker/__init__.pyfrom . import agentyoutube_shorts_maker/agent.pyfrom google.genai import typesfrom google.adk.agents import Agentfrom google.adk.tools.agent_tool import AgentToolfrom google.adk.models.lite_l..

Step 6: Save Report (Optional)6-1. Main Agent LLM Call #12 (Report Generation)Response received by Google ADK from OpenAI GPT-4o API (LLM Response #12):JSON { "tool_calls": [ { "id": "call_401", "function": { "name": "save_advice_report", "arguments": "{\"summary\": \"[Final Advice Text]\", \"ticker\": \"AAPL\"}" } } ]}Generator: OpenAI GPT-4o LLM (for Main A..

System OverviewThis system is a financial advisory system utilizing a multi-agent architecture. The main agent (Financial Advisor) coordinates three specialized sub-agents to provide comprehensive investment analysis.Financial Advisor (Main Agent)├── Data Analyst (Basic Data Collection)├── Financial Analyst (Financial Statement Analysis)├── News Analyst (News Search and Analysis)└── save_advice_..

agent.pyfrom google.genai import typesfrom google.adk.tools import ToolContextfrom google.adk.agents import Agentfrom google.adk.tools.agent_tool import AgentToolfrom google.adk.models.lite_llm import LiteLlmfrom .sub_agents.data_analyst import data_analystfrom .sub_agents.financial_analyst import financial_analystfrom .sub_agents.news_analyst import news_analystfrom .prompt import PROMPTMODEL =..

Now we learn about google adk. By using this wonderful tool, you don't have to code frontend ui when your goal is just debugging and checking if code is alright as developer.First of all, you set up files like this. And pyproject.toml file is [project] name = "financial-analyst" version = "0.1.0" description = "Add your description here" readme = "README.md" requires-python = ">=3.12" dependenc..

1. Purpose and Concept of Output GuardrailThe Output Guardrail validates the response generated by the agent to check if it contains inappropriate content. Specifically, it ensures the Technical Support Agent does not include content unrelated to technical support (such as billing, orders, account management, etc.).Core Concepts:Automated Verification: Checks immediately after response generatio..

Tool System Integration Explanation1. Tool Definition - @function_tool Decorator@function_tooldef run_diagnostic_check( context: UserAccountContext, product_name: str, issue_description: str) -> str: """ Run a diagnostic check on the customer's product to identify potential issues. Args: product_name: Name of the product experiencing issues issue_description: Descriptio..

1. Handoff Data Model Definitionclass HandoffData(BaseModel): to_agent_name: str issue_type: str issue_description: str reason: strHandoffData: Pydantic model containing handoff informationFields:to_agent_name: Name of the target agent to hand off toissue_type: Type of issueissue_description: Description of the issuereason: Reason for the handoff2. Handoff Handler Functiondef handle_..

Context and Input Guardrail System1. Context Model Definitionmodels.py Lines 5-9class UserAccountContext(BaseModel): customer_id: int name: str tier: str = "basic" email: Optional[str] = None # premium enterpriseUserAccountContext: Pydantic model containing customer account informationFields:customer_id: Customer IDname: Customer nametier: Tier (default "basic", can be "premium", "e..

Now we make customer support ai agent. we want cooperating multiagent and input&output gaurdrail. First, you should setup files like this. 1)main.pyimport dotenvdotenv.load_dotenv()from openai import OpenAIimport asyncioimport streamlit as stfrom agents import Runner, SQLiteSession, InputGuardrailTripwireTriggered, OutputGuardrailTripwireTriggeredfrom models import UserAccountContextfrom my_agen..