Skip to main content

Lab 4.5 Challenge: Professional Model Configuration, Resiliency & Portability

Goal​

In this lab, you will upgrade your "Support Analyzer" agent to a production-ready state, in place -- keeping its Module 4 structured-output contract (SupportAnalysis, output_schema, output_key) fully working. You will learn how to implement advanced retry logic using a custom Gemini subclass and how to provide a multi-model fallback using LiteLlm, layered on top of the agent you already built.

Prerequisites​

Environment setup: this lab modifies the support_analyzer project from Module 4, in place -- it's not a fresh project. If you're continuing the course in sequence, cd into that project (inside your adk-training directory) and you're all set. If you're starting fresh at this module, you'll need that project first -- Module 4's lab-solution.md has the complete support_analyzer code to recreate it, then come back here. Note that lab-solution.md usually only shows the contents of agent.py; if that's not enough to get a working project (e.g. you still need to scaffold it with uv/adk create), check Module 4's lab.md for those setup steps.

  1. Navigate to your support_analyzer project and install LiteLLM: 
    cd /path/to/your/adk-training/support_analyzer
    uv add "litellm==1.96.0"

    Note: We pin to litellm==1.96.0 here. Newer litellm releases (1.97.0+, as of this writing) have known issues on Python 3.10 — 1.98.0 raises ImportError: cannot import name 'NotRequired' from 'typing' the moment the LiteLlm model path is actually exercised, and 1.97.0 hits an unrelated Pydantic model-definition error. 1.96.0 has been verified to work cleanly with google-adk on Python 3.10.

Lab Tasks​

  1. Create a Resilient Subclass:

    • In your support_analyzer/agent.py, create a class named ResilientGemini that inherits from Gemini.
    • Override the api_client property.
    • Configure it with a HttpRetryOptions policy: max_delay=10, exp_base=2.0, and jitter=0.5.

  2. Implement Multi-Model Fallback:

    • Update your agent's model logic.
    • If the environment variable USE_LOCAL_MODEL is set to "1", use LiteLlm with ollama_chat/mistral.
    • Otherwise, use your new ResilientGemini class.

  3. Keep the Module 4 structured output intact:

    • Your SupportAnalysis Pydantic model, output_schema=SupportAnalysis, and output_key="last_ticket_analysis" from Module 4 should still be set on root_agent -- you're only changing which model powers the agent, not the fact that it returns structured JSON.

  4. Verify the Configuration:

    • Run the agent using uv run adk run support_analyzer.
    • Verify it still returns a schema-conforming JSON response, exactly like it did at the end of Module 4. (Note: You won't "see" the retries unless a network error occurs, but your code is now protected!).

Python Approach (Primary)​

Modify agent.py to use the advanced configuration patterns, on top of the SupportAnalysis schema you already built in Module 4.

import os
from functools import cached_property
from pydantic import BaseModel
from google.adk import Agent
from google.adk.models import Gemini
from google.adk.models.lite_llm import LiteLlm
from google.genai import Client, types

# Carried over from Module 4 -- keep this working!
class SupportAnalysis(BaseModel):
    category: str
    sentiment: str
    summary: str

# TODO: Step 1 - Define the ResilientGemini subclass
class ResilientGemini(Gemini):
    @cached_property
    def api_client(self) -> Client:
        # TODO: Implement Client with HttpRetryOptions
        pass

# TODO: Step 2 - Implement the model selection logic
if os.getenv("USE_LOCAL_MODEL") == "1":
    # Use LiteLLM
    model_to_use = ...
else:
    # Use your ResilientGemini
    model_to_use = ...

root_agent = Agent(
    name="support_analyzer_agent",
    model=model_to_use,
    instruction="Analyze customer support issues.",
    output_schema=SupportAnalysis,  # TODO: Keep this from Module 4
    output_key="last_ticket_analysis",  # TODO: Keep this from Module 4
)

Self-Reflection Questions​

  • Why is "Jitter" important in a retry policy for a high-traffic production application?
  • What are the advantages of centralizing model configuration in a subclass instead of passing parameters to every agent instance?
  • In which scenario would you prefer using the native Gemini class over the LiteLlm abstraction?

🔐 Reference Solution Vault​

🔐

Module 04_5 Reference Solution Vault

Active Comprehension Gate • Zero Guessable URLs • Guided Verification

Step 1/3: Question 1 of 3

What is the primary architectural purpose of integrating LiteLLM with Google ADK?