Google ADK

Use mycontext cognitive patterns as the instruction for Google Agent Development Kit (ADK) agents. The GoogleADKHelper creates fully-configured ADK Agent objects from any Context.

pip install mycontext-ai google-adk

Quick Start

from mycontext.templates.free.reasoning import StepByStepReasoner
from mycontext.integrations import GoogleADKHelper

# Build cognitive context
ctx = StepByStepReasoner().build_context(
    problem="Design a scalable notification system for 10M users",
    domain="software architecture",
)

# Create Google ADK agent
agent = GoogleADKHelper.create_agent(
    context=ctx,
    name="architect",
    model="gemini-2.0-flash",
    description="Systems architect specializing in scalable infrastructure",
)

# Use the agent
from google.adk.runners import Runner
runner = Runner(agent=agent, app_name="architecture_advisor")
response = runner.run("What are the key considerations for the notification system?")
print(response)

ctx.to_instruction() / GoogleADKHelper.to_instruction()

Both methods return the assembled context as a plain string for the ADK instruction field:

# Via helper
from mycontext.integrations import GoogleADKHelper
instruction = GoogleADKHelper.to_instruction(ctx)

# Via Context method
instruction = ctx.assemble()

GoogleADKHelper Methods

create_agent(context, name, model, description, tools, **kwargs)

agent = GoogleADKHelper.create_agent(
    context=ctx,
    name="risk_analyst",
    model="gemini-2.0-flash",        # Default
    description="Risk analysis specialist",
    tools=[search_tool, calculator],  # Optional ADK tools
)

Parameters:

Parameter	Type	Default	Description
context	Context	required	mycontext Context
name	str	"agent"	Agent name
model	str	"gemini-2.0-flash"	Gemini model
description	str | None	None	Auto-derived from directive
tools	list | None	[]	ADK tool list

to_instruction(context)

instruction = GoogleADKHelper.to_instruction(ctx)
# Returns the full assembled context string

Multi-Agent ADK Pipeline

from mycontext.templates.free.reasoning import RootCauseAnalyzer
from mycontext.templates.free.planning import ScenarioPlanner
from mycontext.templates.free.specialized import RiskAssessor
from mycontext.integrations import GoogleADKHelper

# Each agent powered by a different cognitive framework
diagnostician = GoogleADKHelper.create_agent(
    RootCauseAnalyzer().build_context(
        problem="Mobile app retention dropped 35%",
        depth="comprehensive",
    ),
    name="diagnostician",
    model="gemini-2.0-flash",
)

planner = GoogleADKHelper.create_agent(
    ScenarioPlanner().build_context(
        topic="Mobile retention recovery strategies",
        timeframe="6 months",
    ),
    name="planner",
    model="gemini-2.0-flash",
)

risk_agent = GoogleADKHelper.create_agent(
    RiskAssessor().build_context(
        decision="Invest in onboarding redesign vs. feature expansion",
        depth="thorough",
    ),
    name="risk_analyst",
    model="gemini-2.0-flash",
)

Direct Context Export

If you need more control, export the context string and build the ADK agent manually:

from mycontext.templates.free.analysis import DataAnalyzer
from google.adk.agents import Agent

ctx = DataAnalyzer().build_context(
    data_description="User engagement metrics: DAU, session length, feature usage",
    goal="Identify leading indicators of premium conversion",
)

agent = Agent(
    model="gemini-2.0-flash",
    name="conversion_analyst",
    instruction=ctx.assemble(),     # mycontext cognitive framework
    description="Analyzes engagement data to predict premium conversion",
    tools=[bigquery_tool],
)

API Reference

GoogleADKHelper

Method	Returns	Description
to_instruction(context)	str	Assembled context as ADK instruction
create_agent(context, name, model, description, tools, **kwargs)	Agent	Google ADK Agent