Skip to content

Latest commit

 

History

History
177 lines (147 loc) · 3.32 KB

README.md

File metadata and controls

177 lines (147 loc) · 3.32 KB
Raw

Quickstart

Prerequisites

  • Docker (with Compose)
  • curl
  • jq

Clone the repository:

git clone [email protected]:openmeterio/openmeter.git
cd openmeter/quickstart

1. Launch OpenMeter

Launch OpenMeter and its dependencies via:

docker compose up -d

2. Ingest usage event(s)

Ingest usage events in CloudEvents format:

curl -X POST http://localhost:8888/api/v1/events \
-H 'Content-Type: application/cloudevents+json' \
--data-raw '
{
  "specversion" : "1.0",
  "type": "request",
  "id": "00001",
  "time": "2023-01-01T00:00:00.001Z",
  "source": "service-0",
  "subject": "customer-1",
  "data": {
    "method": "GET",
    "route": "/hello",
    "duration_ms": 10
  }
}
'

Note how ID is different:

curl -X POST http://localhost:8888/api/v1/events \
-H 'Content-Type: application/cloudevents+json' \
--data-raw '
{
  "specversion" : "1.0",
  "type": "request",
  "id": "00002",
  "time": "2023-01-01T00:00:00.001Z",
  "source": "service-0",
  "subject": "customer-1",
  "data": {
    "method": "GET",
    "route": "/hello",
    "duration_ms": 20
  }
}
'

Note how ID and time are different:

curl -X POST http://localhost:8888/api/v1/events \
-H 'Content-Type: application/cloudevents+json' \
--data-raw '
{
  "specversion" : "1.0",
  "type": "request",
  "id": "00003",
  "time": "2023-01-02T00:00:00.001Z",
  "source": "service-0",
  "subject": "customer-1",
  "data": {
    "method": "GET",
    "route": "/hello",
    "duration_ms": 30
  }
}
'

3. Query Usage

Query the usage hourly:

curl 'http://localhost:8888/api/v1/meters/api_requests_total/query?windowSize=HOUR&groupBy=method&groupBy=route' | jq
{
  "windowSize": "HOUR",
  "data": [
    {
      "value": 2,
      "windowStart": "2023-01-01T00:00:00Z",
      "windowEnd": "2023-01-01T01:00:00Z",
      "subject": null,
      "groupBy": {
        "method": "GET",
        "route": "/hello"
      }
    },
    {
      "value": 1,
      "windowStart": "2023-01-02T00:00:00Z",
      "windowEnd": "2023-01-02T01:00:00Z",
      "subject": null,
      "groupBy": {
        "method": "GET",
        "route": "/hello"
      }
    }
  ]
}

Query the total usage for customer-1:

curl 'http://localhost:8888/api/v1/meters/api_requests_total/query?subject=customer-1' | jq
{
  "data": [
    {
      "value": 3,
      "windowStart": "2023-01-01T00:00:00Z",
      "windowEnd": "2023-01-02T00:01:00Z",
      "subject": "customer-1",
      "groupBy": {}
    }
  ]
}

4. Configure additional meter(s) (optional)

In this example we will meter LLM token usage, groupped by AI model and prompt type. You can think about it how OpenAI charges by tokens for ChatGPT.

Configure how OpenMeter should process your usage events in this new tokens_total meter.

# ...

meters:
  # Sample meter to count LLM Token Usage
  - slug: tokens_total
    description: AI Token Usage
    eventType: prompt               # Filter events by type
    aggregation: SUM
    valueProperty: $.tokens         # JSONPath to parse usage value
    groupBy:
      model: $.model                # AI model used: gpt4-turbo, etc.
      type: $.type                  # Prompt type: input, output, system

Cleanup

Once you are done, stop any running instances:

docker compose down -v