relay

Codegen for the Relay artifact.

A relay is a GitHub Action that pulls metrics from a source vendor and pushes them to a destination vendor. The compiler emits a self-contained set of files that a target repository drops in to start relaying.

Today the only supported pairing is Langfuse → Datadog, producing four files:

.github/workflows/<name>.yml — workflow that runs the runner on a schedule, manages a cursor in a repo Variable, and dispatches manually.
relay/gleam.toml — runner project manifest.
relay/src/relay.gleam — runner that queries Langfuse v2 metrics for [from, to) and submits one Datadog metric per scorer.
relay/src/manifest.gleam — static declaration of which metrics the relay emits. The runner imports it for its scorer allowlist; later language tooling will import it to type-check measurement references.

File contents come from templates under priv/templates/relay/<pairing>/. Each template is a real .yml / .toml / .gleam file with {{PLACEHOLDER}} tokens — so editors and reviewers get the right syntax highlighting and the diffs read naturally.

Types

LangfuseDatadogRelay

</>

Configuration for a Langfuse → Datadog relay. These fields are the placeholder shape until the language constructs land — at that point they will be populated from a parsed relay declaration rather than constructed directly.

pub type LangfuseDatadogRelay {
  LangfuseDatadogRelay(
    name: String,
    metric_prefix: String,
    schedule_cron: String,
    scorers: List(Scorer),
  )
}

Constructors

```
LangfuseDatadogRelay(
  name: String,
  metric_prefix: String,
  schedule_cron: String,
  scorers: List(Scorer),
)
```
Arguments

name

Identifier for the relay; used as the workflow filename stem and as the workflow’s name field.

metric_prefix

Prefix for Datadog metric names — caller-supplied, no default. The relay emits <metric_prefix>.count (count metric, one submission per scorer per window) and <metric_prefix>.value (gauge, one submission per numeric scorer with the window’s avg value). Choosing the prefix at relay-declaration time keeps the user’s Datadog namespace under their own control.

schedule_cron

Cron expression for the workflow’s schedule trigger.

scorers

Allowlist of scorers to emit metrics for. Rows from Langfuse whose name isn’t in this list are dropped before submission. The same list drives the generated manifest module.

RelayFile

</>

One file in a relay artifact’s output: a path relative to the target repository root, plus its full contents.

pub type RelayFile {
  RelayFile(path: String, content: String)
}

Constructors

RelayFile(path: String, content: String)

Scorer

</>

A scorer the relay should emit metrics for. The user declares the set up-front; the relay filters Langfuse rows to this allowlist before submitting to Datadog.

pub type Scorer {
  Scorer(name: String, data_type: ScorerDataType)
}

Constructors

Scorer(name: String, data_type: ScorerDataType)

ScorerDataType

</>

Which family of Langfuse scores a scorer produces. Numeric and Boolean scores both get count + value metrics; Categorical scores only get counts (the v2 metrics API rejects value measure on categorical).

pub type ScorerDataType {
  Numeric
  Boolean
  Categorical
}

Constructors

```
Numeric
```
```
Boolean
```
```
Categorical
```

Values

generate_langfuse_to_datadog

</>

pub fn generate_langfuse_to_datadog(
  relay: LangfuseDatadogRelay,
) -> List(RelayFile)

Generate the file set for a Langfuse → Datadog relay.

Constructors

Arguments

Constructors

Constructors

Constructors