Note: This is an experimental proof-of-concept project exploring alternative syntax for PromQL queries.
Try it out: Interactive Playground
Flow is a query language that transpiles to PromQL (Prometheus Query Language). When writing PromQL queries, users typically start by exploring a metric and its labels, then add transformations like rate calculations and aggregations. However, PromQL's nested function syntax forces you to jump back and forth in the query—adding a new transformation means wrapping the entire expression in yet another function. Flow solves this by providing a left-to-right pipeline syntax using the | operator, allowing you to write queries in the same order you think about them: start with a metric, then apply transformations step by step.
The language supports all PromQL features including metric selection, label matching, aggregations, range functions, and binary operations. It also introduces let expressions for defining reusable query fragments, reducing duplication and improving maintainability. Flow is available both as a Go library and as a WebAssembly module for client-side transpilation.
# Flow
http_requests_total{method="POST"} | rate 5m | sum by (endpoint)
# PromQL
sum by (endpoint) (rate(http_requests_total{method="POST"}[5m]))
# Flow
let
req_get = http_requests_total{method="GET"} | rate 5m | sum by (endpoint),
req_total = http_requests_total | rate 5m | sum by (endpoint)
in req_get / req_total
# PromQL
sum by (endpoint) (rate(http_requests_total{method="GET"}[5m]))
/
sum by (endpoint) (rate(http_requests_total[5m]))
# Flow
let
req_get = http_requests_total{method="GET"} | rate 5m | sum by (endpoint),
req_total = http_requests_total | rate 5m | sum by (endpoint)
in 100 * (req_total - req_get) / req_total
# PromQL
100 * (
sum by (endpoint) (rate(http_requests_total[5m]))
-
sum by (endpoint) (rate(http_requests_total{method="GET"}[5m]))
)
/
sum by (endpoint) (rate(http_requests_total[5m]))
# Flow
let
avail = kube_deployment_status_replicas_available{namespace="monitoring"},
loop_time = looping_time{group_name="realtime"} | rate 2m | sum by () | histogram_sum,
sleep_time = sleeping_time{group_name="realtime"} | rate 2m | sum by () | histogram_sum
in (avail > 1) * loop_time / (loop_time + sleep_time)
# PromQL
(kube_deployment_status_replicas_available{namespace="monitoring"} > 1)
*
histogram_sum(sum(rate(looping_time{group_name="realtime"}[2m])))
/
(
histogram_sum(sum(rate(looping_time{group_name="realtime"}[2m])))
+
histogram_sum(sum(rate(sleeping_time{group_name="realtime"}[2m])))
)