This project has been archived by its maintainers, and is no longer receiving any updates.
ADK Mimir Memory
Persistent, local, encrypted cross-session memory for Google ADK agents — backed by Mimir.
Why Mimir?
| Backend | Dependencies | Encryption | Hybrid Search | Local |
|---|---|---|---|---|
| InMemoryMemoryService | None | ❌ | ❌ | ✅ |
| VertexAiMemoryBankService | GCP + Gemini | ❌ | Gemini-driven | ❌ |
| VertexAiRagMemoryService | GCP + RAG | ❌ | GCP vector | ❌ |
| MimirMemoryService | Single binary | ✅ AES-256 | ✅ BM25+FTS5+Dense | ✅ |
- Zero cloud dependencies — a single Rust binary, SQLite database, fully local
- AES-256-GCM encryption at rest — your memory data stays private
- Hybrid search — BM25 keyword + FTS5 + dense vector search
- 30+ MCP tools — remember, recall, synthesize, benchmark, federate, and more
- Ebbinghaus confidence decay — memories fade naturally, important ones persist
Installation
pip install adk-mimir-memory
This package requires the mimir binary. Download it from:
https://github.com/Perseus-Computing-LLC/mimir/releases
Or build from source:
cargo install mimir
Quick Start
from google.adk.agents import Agent
from google.adk.runners import Runner
from google.adk.sessions import InMemorySessionService
from adk_mimir_memory import MimirMemoryService
agent = Agent(
name="my_agent",
model="gemini-2.5-flash",
instruction="You are a helpful assistant with persistent memory.",
)
# The memory service is configured on the Runner, not on the Agent.
runner = Runner(
agent=agent,
app_name="my_app",
session_service=InMemorySessionService(),
memory_service=MimirMemoryService(db_path="~/.adk/mimir.db"),
)
That's it. Sessions, events, and explicit memories are now persisted across restarts.
Configuration
# Custom database location
MimirMemoryService(db_path="/data/agent_memory.db")
# Custom mimir binary path (if not on $PATH)
MimirMemoryService(mimir_binary="/usr/local/bin/mimir")
# Both
MimirMemoryService(
db_path="/data/agent_memory.db",
mimir_binary="/usr/local/bin/mimir",
)
Perseus Live Context (Optional)
This package also includes a drop-in agent with live workspace awareness via Perseus:
from google.adk.runners import Runner
from google.adk.sessions import InMemorySessionService
from adk_mimir_memory.perseus_context import perseus_context_agent
# The agent resolves @file, @search, @memory directives at inference time.
# Bind it on the Runner; run_async takes no `agent` argument.
runner = Runner(
agent=perseus_context_agent,
app_name="my_app",
session_service=InMemorySessionService(),
)
runner.run_async(
user_id="user",
session_id="session",
new_message=types.Content(role="user", parts=[types.Part.from_text(
text="What does the README say about deployment?"
)]),
)
pip install adk-mimir-memory[perseus] # installs perseus-ctx
Set directives via session state:
session = await runner.session_service.create_session(
app_name="my_app",
user_id="user",
state={
"_perseus_directives": "@file AGENTS.md @file README.md @memory deployment",
"_perseus_workspace": "/path/to/project",
},
)
How It Works
┌─────────────┐ JSON-RPC (MCP stdio) ┌──────────┐
│ ADK Agent │ ──────────────────────────▶ │ Mimir │
│ (Python) │ ◀────────────────────────── │ (Rust) │
└─────────────┘ └────┬─────┘
│
SQLite + FTS5
(AES-256-GCM)
The MimirMemoryService spawns a mimir subprocess and communicates via JSON-RPC over stdin/stdout (MCP stdio transport). Each add_session_to_memory, add_memory, and search_memory call translates to a Mimir MCP tool invocation.
License
Metadata
Release files for adk-mimir-memory 0.3.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| adk_mimir_memory-0.3.0.tar.gz | 11.9 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| adk_mimir_memory-0.3.0-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 23.3 kB
Release files / adk_mimir_memory-0.3.0.tar.gz
| Download URL | adk_mimir_memory-0.3.0.tar.gz |
|---|---|
| Size | 11.9 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
29c3d1721dca3c4d7bbd69f40189e7c79831b36e683aacb4d55e097c948f7d88
|
|
BLAKE2b-256 checksum How to use checksums |
c8417b0f97f5cd0856456aad028a72a52452bd35cad08b399fee18a654e81bfb
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.12
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jun 28, 2026.
Transparency logRelease files / adk_mimir_memory-0.3.0-py3-none-any.whl
| Download URL | adk_mimir_memory-0.3.0-py3-none-any.whl |
|---|---|
| Size | 11.4 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
9669dee5bea2e1d61b3caecf8d5b9aa15b9d54582853e0d020d828b443e78cb1
|
|
BLAKE2b-256 checksum How to use checksums |
5eaf06e30df3e8524de5f4475c878dc671601887fd85e936d357c51282bda655
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.12
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jun 28, 2026.
Transparency log