> ## Documentation Index
> Fetch the complete documentation index at: https://docs.canaryedge.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Baseline

> Create or retrieve a healthy baseline for a machine

## Create Baseline

`POST /v1/baseline`

Create or update a healthy baseline for a specific machine. This computes energy statistics from the provided time series and fine-tunes a per-machine predictor (\~462K params, approximately 10-15 seconds on CPU).

<ParamField body="machine_id" type="string" required>
  Unique machine identifier within your organization.
</ParamField>

<ParamField body="series" type="array" required>
  Array of **healthy operating** data points, each with `timestamp` (ISO 8601) and `value` (number). Minimum 100 points. Do not include known anomalies.
</ParamField>

<ParamField body="granularity" type="string" default="none">
  Data frequency hint for model selection.
</ParamField>

<ParamField body="label" type="string">
  Human-readable label for the baseline (e.g., "Normal operating conditions, January 2024").
</ParamField>

<ResponseField name="machine_id" type="string">
  The machine identifier.
</ResponseField>

<ResponseField name="baseline_id" type="string">
  Baseline identifier (format: `bl_` + machine\_id).
</ResponseField>

<ResponseField name="created_at" type="string">
  ISO 8601 creation timestamp.
</ResponseField>

<ResponseField name="stats" type="object">
  Baseline energy statistics.

  * `energy_mean` (number): Mean energy during healthy operation.
  * `energy_std` (number): Standard deviation of healthy energy.
  * `energy_p99` (number): 99th percentile of healthy energy.
  * `num_windows` (integer): Number of sliding windows computed.
  * `series_length` (integer): Length of the input series.
</ResponseField>

<ResponseField name="finetune_stats" type="object">
  Fine-tuning results.

  * `status` (string): `completed` or `skipped`.
  * `final_loss` (number): Best predictor loss achieved.
  * `epochs_run` (integer): Actual epochs trained.
  * `epochs_max` (integer): Maximum allowed epochs.
  * `early_stopped` (boolean): Whether early stopping triggered.
  * `windows` (integer): Training windows used.
  * `duration_seconds` (number): Fine-tuning wall time.
</ResponseField>

<ResponseField name="status" type="string">
  Baseline status (`active`).
</ResponseField>

<RequestExample>
  ```bash Create Baseline theme={null}
  curl -X POST https://api.canaryedge.com/v1/baseline \
    -H "Content-Type: application/json" \
    -H "X-Canary-Key: YOUR_API_KEY" \
    -d '{
      "machine_id": "pump-47",
      "series": [
        {"timestamp": "2024-01-01T00:00:00Z", "value": 1.0},
        {"timestamp": "2024-01-01T00:01:00Z", "value": 1.02},
        {"timestamp": "2024-01-01T00:02:00Z", "value": 0.98}
      ],
      "granularity": "minutely",
      "label": "Normal operating conditions, January 2024"
    }'
  ```

  ```bash Get Baseline theme={null}
  curl "https://api.canaryedge.com/v1/baseline?machine_id=pump-47" \
    -H "X-Canary-Key: YOUR_API_KEY"
  ```
</RequestExample>

<ResponseExample>
  ```json 200 theme={null}
  {
    "machine_id": "pump-47",
    "baseline_id": "bl_pump-47",
    "created_at": "2026-03-30T12:00:00Z",
    "stats": {
      "energy_mean": 0.023,
      "energy_std": 0.008,
      "energy_p99": 0.051,
      "num_windows": 1984,
      "series_length": 2048
    },
    "finetune_stats": {
      "status": "completed",
      "final_loss": 0.000342,
      "epochs_run": 35,
      "epochs_max": 50,
      "early_stopped": true,
      "windows": 1984,
      "duration_seconds": 12.3
    },
    "status": "active"
  }
  ```
</ResponseExample>

***

## Get Baseline

`GET /v1/baseline?machine_id={machine_id}`

Retrieve the current baseline for a machine. Returns the same response format as the create endpoint.