Monitor API Usage with Predexy Analytics

Track call volume, error rates, and response latency for each API key using aggregated usage statistics and paginated raw request logs.

Predexy provides two endpoints for understanding how your API keys are being used: an aggregated statistics endpoint that groups calls into time periods, and a raw log endpoint that lets you inspect individual requests. Both require your console session token and the UUID of the API key you want to inspect. You can find the UUID in the list keys response.

Usage statistics

GET /api/v1/console/stats returns call counts and latency metrics grouped by a time period you choose. Use this to spot trends — a sudden spike in error calls, a latency regression, or unusual traffic patterns.

Query parameters:

Parameter	Required	Default	Notes
api_key_id	Yes	—	UUID of the API key to fetch stats for
granularity	No	daily	One of hourly, daily, or monthly
limit	No	30	Number of periods to return. Minimum 1, maximum 90

curl "https://api.predexy.com/api/v1/console/stats?\
api_key_id=3fa85f64-5717-4562-b3fc-2c963f66afa6\
&granularity=daily\
&limit=7" \
  -H "Authorization: Bearer <your-access-token>"

Response:

{
  "data": [
    {
      "period_start": "2026-04-25T00:00:00Z",
      "total_calls": 1523,
      "successful_calls": 1500,
      "error_calls": 23,
      "avg_latency_ms": 42.5,
      "p95_latency_ms": 120.0,
      "unique_endpoints": 3
    },
    {
      "period_start": "2026-04-24T00:00:00Z",
      "total_calls": 891,
      "successful_calls": 891,
      "error_calls": 0,
      "avg_latency_ms": 38.1,
      "p95_latency_ms": 95.0,
      "unique_endpoints": 2
    }
  ],
  "meta": {
    "granularity": "daily",
    "count": 2
  }
}

Response fields (UsageStatPoint):

Field	Description
period_start	Start of the aggregation period (ISO 8601)
total_calls	All API calls made in this period
successful_calls	Calls that returned a 2xx status
error_calls	Calls that returned a 4xx or 5xx status
avg_latency_ms	Mean server-side response time in milliseconds
p95_latency_ms	95th percentile response time — useful for catching tail latency
unique_endpoints	Number of distinct endpoint paths called in the period

Calculate your error rate as error_calls / total_calls. If this ratio rises above your baseline, check the raw logs to identify which endpoint is failing and what status codes are returned.

Raw request logs

GET /api/v1/console/logs returns paginated individual request records for a key. Use this to debug a specific failure, audit what your application called, or investigate an unexpected 401.

Query parameters:

Parameter	Required	Default	Notes
api_key_id	Yes	—	UUID of the API key
limit	No	50	Items per page. Maximum 200
offset	No	0	Number of items to skip

curl "https://api.predexy.com/api/v1/console/logs?\
api_key_id=3fa85f64-5717-4562-b3fc-2c963f66afa6\
&limit=20\
&offset=0" \
  -H "Authorization: Bearer <your-access-token>"

Response:

{
  "data": [
    {
      "endpoint": "/api/v1/external/arbitrage/opportunities",
      "method": "GET",
      "status_code": 200,
      "response_time_ms": 45,
      "created_at": "2026-04-25T14:32:01Z"
    },
    {
      "endpoint": "/api/v1/external/markets",
      "method": "GET",
      "status_code": 429,
      "response_time_ms": 12,
      "created_at": "2026-04-25T14:31:58Z"
    }
  ],
  "meta": {
    "count": 20,
    "offset": 0,
    "limit": 20,
    "total": 1523
  }
}

Response fields (UsageLogEntry):

Field	Description
endpoint	API path called (e.g. /api/v1/external/arbitrage/opportunities)
method	HTTP method (GET, POST, etc.)
status_code	HTTP status code returned by the server
response_time_ms	Server-side response time in milliseconds
created_at	Timestamp of the request (ISO 8601)

To page through results, increment offset by your limit value on each request until offset >= meta.total.

How to find your API key UUID

The api_key_id parameter is the UUID assigned to your key at creation — not the key itself. To retrieve it, call the list keys endpoint:

curl https://api.predexy.com/api/v1/console/keys \
  -H "Authorization: Bearer <your-access-token>"

The id field in each object is the UUID you pass to the stats and logs endpoints.

Practical tips

Track your error rate over time. Use the daily stats with limit=30 to build a 30-day view of error_calls / total_calls. A sudden increase often signals a breaking change in your integration or a rate limit being hit (429 in the raw logs).

Identify slow endpoints. The unique_endpoints field in stats tells you how many different paths were called in a period. When p95_latency_ms spikes, drill into raw logs filtered around that time window to see which endpoint is responsible.

Monitor quota usage. Compare total_calls against your key’s rate_limit (visible from the list keys response, in requests per minute). If you are regularly hitting your limit — indicated by 429 status codes in the logs — consider requesting a higher rate limit when creating a new key.

Use hourly granularity for active debugging. When investigating an incident from the past few hours, switch to granularity=hourly and limit=24 to see per-hour breakdowns instead of daily aggregates.