Traces Observer Service

Overview

Traces Observer Service is a lightweight backend that provides an HTTP API for querying traces collected in the observability plane. It communicates with OpenSearch to retrieve and aggregate tracing data produced by agents(components) across the platform, and exposes simple endpoints for the console. It is planned to migrate this service and functionality to the OpenChoreo Observability Plane Observer service later.

Features

• Query traces and span documents stored in OpenSearch
• Support time-range filtering and pagination
• Provide a health endpoint for readiness/liveness checks
• Serve as the backend for the console traces UI

How it works

1. The console sends a tracing query to the HTTP API exposed by this service.
2. The service validates and translates the request into an OpenSearch query.
3. The OpenSearch client executes the query against the configured cluster/indices.
4. Results are formatted and returned to the caller (with paging and basic aggregation info where applicable).

Architecture:

HTTP API --> Request Handler --> Service/Query Layer --> OpenSearch Client --> OpenSearch Cluster

Configuration

This service is configured via environment variables. The following are supported and commonly used:

# Server Configuration
TRACES_OBSERVER_PORT=9098

# OpenSearch Configuration
OPENSEARCH_ADDRESS=http://localhost:9200
OPENSEARCH_USERNAME=admin
OPENSEARCH_PASSWORD=admin
OPENSEARCH_TRACE_INDEX=custom-otel-span-index

Set the environment Variables

Build and run — local (Go)

Prerequisites: Go 1.20+ and a running OpenSearch instance reachable by the service.

Build the binary:

go build -o traces-observer-service ./...

Set the enviornment variables:

export OPENSEARCH_ADDRESS=https://localhost:9200 OPENSEARCH_USERNAME=admin OPENSEARCH_PASSWORD=ThisIsTheOpenSearchPassword1 TRACES_OBSERVER_PORT=9098

Run with environment variables set (example):

OPENSEARCH_ADDRESS=http://localhost:9200 OPENSEARCH_USERNAME=admin OPENSEARCH_PASSWORD=ThisIsTheOpenSearchPassword1  TRACES_OBSERVER_PORT=9098 \
./traces-observer-service

Or run directly:

go run . # or `go run .` from the service root, depending on project layout

Docker

Build the image:

docker build -t traces-observer-service .

Run the container (example):

docker run --rm -p 9098:9098 \
	-e OPENSEARCH_ADDRESS=http://host.docker.internal:9200 \
	-e OPENSEARCH_USERNAME=admin \
	-e OPENSEARCH_PASSWORD=ThisIsTheOpenSearchPassword1 \
    -e TRACES_OBSERVER_PORT=9098 \
	traces-observer-service

API — Query Parameter Examples

All APIs use standard GET requests with query parameters

1. List traces - GET /api/v1/traces

Retrieves a list of trace overviews for a specific service with optional time filtering and pagination.

Query Parameters:

• serviceName (required) - Name of the service to query traces for
• startTime (optional) - Start time in RFC3339 format (e.g., 2025-11-03T00:00:00Z)
• endTime (optional) - End time in RFC3339 format (e.g., 2025-11-08T23:59:59Z)
• limit (optional) - Maximum number of traces to return (default: 10)
• offset (optional) - Number of traces to skip for pagination (default: 0)
• sortOrder (optional) - Sort order: asc or desc (default: desc - newest first)

Example request:

curl --location 'http://localhost:9098/api/v1/traces?serviceName=sample-app&startTime=2025-11-03T00:00:00Z&endTime=2025-11-08T23:59:59Z&limit=10&offset=0'

Response (200):

{
  "traces": [
    {
      "traceId": "5974d036b3d7709f2fc9f2b48461c176",
      "rootSpanId": "58f16238f09ae1b2",
      "rootSpanName": "LangGraph.workflow",
      "startTime": "2025-11-07T06:23:24.035086494Z",
      "endTime": "2025-11-07T06:23:27.545584559Z",
      "spanCount": 8
    }
  ],
  "totalCount": 1
}

2. Get trace spans - GET /api/v1/trace

Retrieves all spans for a specific trace ID and service.

Query Parameters:

• traceId (required) - The trace ID to retrieve spans for
• serviceName (required) - Name of the service
• sortOrder (optional) - Sort order for spans: asc or desc (default: asc - chronological)
• limit (optional) - Maximum number of spans to return (default: 100)

Example request:

curl --location 'http://localhost:9098/api/v1/trace?traceId=21a29d5d24837ca724b8751494e70a95&serviceName=langchain-docker-app&sortOrder=asc&limit=100'

Response (200):

{
  "spans": [
    {
      "traceId": "21a29d5d24837ca724b8751494e70a95",
      "spanId": "e2c22d3d4b7736bd",
      "name": "LangGraph.workflow",
      "service": "langchain-docker-app",
      "startTime": "2025-11-03T11:42:18.329535246Z",
      "endTime": "2025-11-03T11:42:20.55201954Z",
      "duration": 2222484,
      "kind": "SPAN_KIND_INTERNAL",
      "status": "0",
      "attributes": {
        "traceloop.entity.input": "{...}",
        "traceloop.entity.output": "{...}",
        "traceloop.workflow.name": "LangGraph"
      }
    },
    {
      "traceId": "21a29d5d24837ca724b8751494e70a95",
      "spanId": "c189ec26ae2a0bb5",
      "parentSpanId": "e2c22d3d4b7736bd",
      "name": "agent.task",
      "service": "langchain-docker-app",
      "startTime": "2025-11-03T11:42:18.331008193Z",
      "endTime": "2025-11-03T11:42:20.55134794Z",
      "duration": 2220339,
      "kind": "SPAN_KIND_INTERNAL",
      "status": "0",
      "attributes": {
        "traceloop.entity.name": "agent",
        "traceloop.workflow.name": "LangGraph"
      }
    }
  ],
  "totalCount": 2
}

3. Health check - GET /health

curl http://localhost:9098/health

Response (200):

{
  "status": "healthy",
  "timestamp": "2025-11-09T12:34:56Z"
}

Error responses

All endpoints return appropriate HTTP status codes:

• 200 OK - Success
• 400 Bad Request - Invalid parameters (missing required fields, invalid format)
• 500 Internal Server Error - Server/OpenSearch errors