Skip to content

mef3io-server

A gRPC server for efficient, concurrent access to MEF 3.0 recordings — oriented purely in channels and time.

MEF 3.0 (Multiscale Electrophysiology Format) is a compressed, encryptable format for long-term electrophysiology recordings. mef3io-server puts a network service in front of it: open a file, read its metadata, then request any channels over any [start_uutc, end_uutc) window. Decode is backed by mef3io, served from a shared per-channel tile cache with parallel decode across worker processes and configurable window prefetch.

Why mef3io-server

  • Channels-and-time only — there is no segment grid and no server-side channel selection state; each request is self-contained and self-describing.
  • Shared tile cache — a single byte-budgeted, per-channel cache across all open files, with an idle TTL so finished sessions free memory before the budget is hit.
  • Parallel decode — cold reads and prefetch fan out to worker processes, each holding its own mef3io session, so decode is genuinely parallel across channels.
  • Smooth paging — configurable look-ahead/behind prefetch warms the next and previous window so navigation is a cache hit.
  • Deployable — configured entirely via environment variables, with a prebuilt Docker image published to GHCR.

Quick taste (Python)

from mef3io_server.client import Mef3Client

client = Mef3Client("localhost:50051")

info = client.open_file("/path/to/file.mefd")
t0 = info["start_uutc"]

res = client.get_signal_range(
    "/path/to/file.mefd",
    channels=["Ch1", "Ch2"],           # or None for all channels
    start_uutc=t0,
    end_uutc=t0 + 10_000_000,          # +10 s
)
print("Shape:", res["shape"], "fs:", res["fs"])

client.close_file("/path/to/file.mefd")
client.shutdown()

Where to go
Install pip install, the prebuilt Docker image, building locally
Quick start launch the server and read your first window
Python client the Mef3Client interface end to end
Configuration every environment variable, explained
Docker deployment the /host_root mount and container setup
gRPC API RPCs and message shapes
Performance & benchmarks when the server beats a direct read

Status and scope

Read access (metadata + signal ranges) is complete and released. Decode is handled by mef3io, the C++-backed successor to the legacy pymef/mef_tools stack, imported through its drop-in MefReader compatibility layer.

mef3io-server is developed at the Mayo Clinic BNEL (Bioelectronics Neurophysiology and Engineering Lab).