This is a microservice to accompany blockprint and blockdreamer.
The aim of the gauge is to measure blockprint's accuracy live using synthetic blocks from
blockdreamer. It does this by receiving blocks from blockdreamer via POST /classify, and then
preparing and sending them to blockprint for classification. The classifications are then
compared to the known identities of the nodes, recorded by the gauge and served on the
GET /accuracy endpoint.
A Docker image is available on the GitHub container registry:
docker pull ghcr.io/blockprint-collective/blockgauge
Or you can build from source:
cargo build --release
Blockgauge needs to be pointed at a blockprint API server and a Lighthouse node.
Measure the accuracy of a blockprint instance using synthetic blocks
Usage: blockgauge [OPTIONS] --lighthouse-url <URL> --blockprint-url <URL>
Options:
--lighthouse-url <URL> Lighthouse node to fetch block reward data from
--blockprint-url <URL> Blockprint instance to use for classifying blocks
--listen-address <IP> Address to listen on [default: 127.0.0.1]
--port <N> Port to listen on [default: 8002]
-h, --help Print help
-V, --version Print version
This API endpoint is for submitting blocks for classification. They will be recorded
in-memory by blockgauge, which then serves accuracy data on /accuracy (see below). If you are
using blockdreamer, it automatically sends data in blockgauge's format when using the setting
extra_data = true.
If you would like to make your own requests to this endpoint, the data structure is:
{
"names": [string],
"labels": [string],
"blocks": [BeaconBlock]
}
For example:
{
"names": ["lighthouse-node-1", "teku-node-1"],
"labels": ["Lighthouse", "Teku"],
"blocks": [
{
"slot": "123",
"proposer_index": "0"
},
{
"slot": "123",
"proposer_index": "0"
}
]
}The contents of the BeaconBlock can be mostly garbage, blockprint only really cares about the
contained attestations. However, all BeaconBlock fields must be present.
This corresponds to the ClassifyRequest struct in blockdreamer's code.
Response:
- On success: HTTP 200 OK with JSON data from Lighthouse's
POST /lighthouse/analysis/block_rewardsAPI. You can save this data on disk for training ablockprintclassifier. Blockdreamer comes with a configuration option to do this for you (results_dir = "/path/"). - On failure: error status code.
Return a JSON Summary (see code) containing information about the classified blocks.
See an example response on https://api.blockprint.sigp.io/accuracy