managed_memory_quickstart

Managed Redis Agent Memory Quickstart

Minimal counterpart to simple_redis_memory. This example uses the managed redis-agent-memory backend (the library default) with no self-hosted Agent Memory Server or Docker setup.

What it demonstrates

• RedisSessionMemoryService for session persistence
• RedisLongTermMemoryService for semantic memory search
• ADK built-in preload_memory and load_memory tools

This example intentionally avoids opensource-only features such as auto-summarization, extraction strategies, recency-boosted search, and MCP.

Prerequisites

• Python 3.10+
• A Redis Agent Memory store (REDIS_AGENT_MEMORY_STORE_ID + API key)
• Google API key for Gemini

Setup

1. Install dependencies

From the repository root (editable install for local PR testing):

uv venv
uv pip install -e ".[all,examples]"

Or use make dev if you also want test/lint tooling.

Note: Even memory-only examples import adk_redis, whose top-level __init__.py currently pulls in search tools that require redisvl. The [all] extra installs that dependency. [memory] alone is not enough today.

2. Configure environment

cd examples/managed_memory_quickstart
cp .env.example .env
# Edit .env with your credentials

3. Start the agent

uv run python main.py

Open http://localhost:8080 in your browser.

Try it

Session 1 — share a few facts:

Hi! My name is Alex.
My favorite color is teal.
I have a cat named Sammy.

Wait for a reply after each message.

Session 2 — click New Session (keep the same userId, e.g. user) and ask:

What do you remember about me?
What is my favorite color?

Cross-session recall uses long-term memory (preload_memory / load_memory). Facts are stored automatically after each turn via the agent's after_agent callback — there is no separate seed step.

Configuration

Variable	Required	Default	Description
GOOGLE_API_KEY	Yes	—	Gemini API key
REDIS_AGENT_MEMORY_API_BASE_URL	Yes	—	Managed API base URL
REDIS_AGENT_MEMORY_API_KEY	Yes	—	Managed API key
REDIS_AGENT_MEMORY_STORE_ID	Yes	—	Memory store ID
REDIS_MEMORY_NAMESPACE	No	managed_memory_quickstart	Namespace for isolation
REDIS_MEMORY_SEARCH_TOP_K	No	5	Results returned by search_memory
PORT	No	8080	ADK web server port

How to run

This example uses a custom main.py with get_fast_api_app because ADK session and memory services must be registered through get_service_registry(). The adk web runner cannot register those services.

For a tools-only example that runs with adk web, see travel_agent_memory_tools.

Related examples

Example	Backend	Runner
This quickstart	Managed (redis-agent-memory)	python main.py
simple_redis_memory	Self-hosted (default)	python main.py
travel_agent_memory_tools	Either (env switch)	adk web .

See examples/README.md for the full matrix.