Original routes and ingest

Still supported. The nine original routes below and the ingest routes keep working exactly as before, with the same token and scopes. For anything new, use the tools API: it covers all of these and much more.

Bodies are JSON; the ingest routes also take raw CSV (Content-Type: text/csv).
Open a route for its fields, an example and its response.
Errors: 422 (bad input, with a message), 404 (unknown or someone else’s id), 401 (bad token), 403 (the token lacks the scope), 429 (over 120 requests a minute).

Models & reliability

GET/api/v1/models

List your saved models (life-data and degradation).

Response · 200

{
  "models": [
    { "id": "a1b2…", "name": "Bearing life", "kind": "distribution",
      "distribution": "Weibull", "n": 30, "unit": "hours",
      "created_at": "2026-07-15T02:48:08+00:00", "url": "/modelling/m/a1b2…" }
  ]
}

GET/api/v1/models/{id}

A model's fitted parameters, life metrics, and goodness-of-fit.

Request

id string · path required Model id.

Response · 200

{
  "id": "a1b2…", "name": "Bearing life", "distribution": "Weibull",
  "kind": "distribution", "unit": "hours", "n": 30,
  "params": [ { "name": "alpha", "value": 1435.1, "se": 108.9, "ci": [1220, 1650] },
              { "name": "beta",  "value": 2.5,    "ci": [1.78, 3.21] } ],
  "coefficients": [],          // populated for proportional-hazards models
  "metrics": null,             // {median, mttf, b10} for discrete / non-parametric
  "gof": [ { "id": "aic", "label": "AIC", "value": 466.3 }, … ],
  "validation": { … }          // regression models only: how good is this model?
                               // {available, concordance: {value, rating, reading},
                               //  brier: {model, baseline, improvement, from, to, reading},
                               //  auc: [{time, value}], n_units, sampled, note}
                               // or {available: false, reason}
  "diagnostics": { … }         // regression models only: does the model hold?
                               // {ph_test: {available, verdict, title, reading,
                               //            rows: [{covariate, statistic, df, p}],
                               //            global: {statistic, df, p}},
                               //  robust?, ties?, strata?}   (Cox PH adds the last three)
}

POST/api/v1/models/{id}/reliability

Evaluate the reliability functions. With t you get point values; without it, the full grid.

Request

`t`	number	optional	Time to evaluate at. Omit for the whole curve.
`covariates`	object	optional	Covariate values for a proportional-hazards model, e.g. {"temp_C": 90}.

curl -X POST https://reliafy.com/api/v1/models/MODEL_ID/reliability \
  -H "Authorization: Bearer rlf_..." -H "Content-Type: application/json" \
  -d '{"t": 1000}'

Response · 200

{
  "model": "Bearing life", "unit": "hours",
  "at": { "t": 1000, "reliability": 0.666, "failure": 0.334,
          "hazard": 0.00101, "cumulative_hazard": 0.406, "density": 0.000675 }
}
// without "t":  { "model": …, "unit": …, "curves": { "x": [...], "sf": [...], "ff": [...], "hf": [...], "Hf": [...], "df": [...] } }

Datasets & fitting

POST/api/v1/datasets

Create a dataset from CSV text or column arrays.

Request

`name`	string	required	Dataset name.
`csv`	string	optional	Raw CSV text (with a header row). Provide this or data.
`data`	object	optional	Column arrays, e.g. {"hours": [...], "failed": [...]}.

curl -X POST https://reliafy.com/api/v1/datasets \
  -H "Authorization: Bearer rlf_..." -H "Content-Type: application/json" \
  -d '{"name": "Bearings", "csv": "hours,failed\n120,1\n340,0"}'

Response · 200

{ "id": "d3…", "name": "Bearings", "n_rows": 2,
  "columns": ["hours", "failed"], "url": "/modelling/data/d3…" }

POST/api/v1/fit

Fit and save a model from one of your datasets.

Request

`name`	string	required	Model name.
`dataset_id`	string	required	Id of the dataset to fit.
`distribution`	string	required	e.g. weibull, lognormal, weibull_ph, discrete_weibull, best. Every id is listed in the model reference: bounded (beta, uniform), discretized, royston_parmar, the semi-parametric additive_hazards / proportional_odds / buckley_james, and shared frailty (weibull_frailty, cox_frailty, …).
`mapping`	object	required	Role → column, e.g. {"x": "hours", "c": "failed"}. Covariates that change over time (regression models): add "i", the item column, with each row's start and stop as "xl" / "xr" (or a timeline's times as "x"). A shared-frailty model also takes "group": the site or batch column.
`unit`	string	optional	Unit of the time axis.
`covariates`	string[]	optional	Covariate columns for a PH model.
`formula`	string	optional	Formulaic covariate formula (alternative to covariates).

curl -X POST https://reliafy.com/api/v1/fit \
  -H "Authorization: Bearer rlf_..." -H "Content-Type: application/json" \
  -d '{"name": "Bearing life", "dataset_id": "d3…", "distribution": "weibull",
       "mapping": {"x": "hours", "c": "failed"}, "unit": "hours"}'

Response · 200

{ "id": "a1b2…", "name": "Bearing life", "kind": "distribution",
  "distribution": "Weibull", "n": 2, "unit": "hours",
  "created_at": "…", "url": "/modelling/m/a1b2…" }

Fleet forecasts

GET/api/v1/fleets/{id}/forecast

The live failure forecast — expected failures per period. Id is in /fleet/forecasts/<id>.

Request

id string · path required Fleet id.

Response · 200

{
  "fleet": { "id": "f1…", "name": "Delivery trucks", "model_id": "a1b2…" },
  "forecast": { "status": "ok", "method": "renewals", "periods": 12,
                "period_label": "months", "expected_failures": [0.4, 0.9, …],
                "total_expected": 11.7 }
}

Strategy calculators

POST/api/v1/strategy/optimal-replacement

Cost-optimal preventive-replacement interval.

Request

`distribution_id`	string	required	e.g. weibull.
`params`	number[] \| {name,value}[]	required	Distribution parameters.
`planned_cost`	number	required	Cost of a planned replacement.
`unplanned_cost`	number	required	Cost of a failure.
`unit`	string	optional	Time unit for display.

Response · 200

{ "distribution": "Weibull", "unit": "hours", "mttf": 1272.4,
  "beneficial": true, "optimal_time": 580.5, "optimal_cost_rate": 0.51,
  "planned_cost": 200, "unplanned_cost": 1500 }

POST/api/v1/strategy/failure-finding

Inspection interval that keeps a hidden (protective) function available.

Request

`distribution_id`	string	required	e.g. exponential.
`params`	number[] \| {name,value}[]	required	Distribution parameters.
`target_availability`	number	required	Target availability in (0,1), e.g. 0.99.
`unit`	string	optional	Time unit for display.

Response · 200

{ "distribution": "Exponential", "unit": "hours", "mttf": 8760,
  "target_availability": 0.99, "interval": 176.4, "method": "…", "note": "…" }

POST/api/v1/strategy/demonstration-test

Plan a reliability demonstration test (success run, binomial, Weibayes or MTBF).

Request

`reliability`	number	required	Reliability to show over one mission, in (0,1) (attribute tests).
`confidence`	number	optional	Confidence in (0,1), default 0.95.
`mission_time`	number	optional	The mission time; omit to plan in missions.
`failures`	int	optional	Failures the test allows, default 0.
`test_multiple`	number	optional	Test length per unit in missions (needs shape when not 1).
`shape`	number	optional	Weibull shape β of the lifetime, assumed known.
`units`	int	optional	Units available: solve for the test time per unit instead.
`method / mtbf`	string / number	optional	method="mtbf" with an MTBF plans a constant-rate (chi-squared) test.
`design_reliability`	number	optional	A design's true reliability: its chance of passing.
`b_life`	number	optional	A B-life requirement instead of reliability: 10 for “B10 ≥ mission_time”.
`shape_interval`	number[2]	optional	β’s interval [lower, upper] (e.g. a fit’s 95% interval): adds shape_sensitivity, with a warning when β is too poorly known for the plan.
`unit`	string	optional	Time unit for display.

Response · 200

{ "method": "attribute", "solve_for": "units", "units": 59,
  "test_time_per_unit": 1000, "total_test_time": 59000, "failures": 0,
  "demonstrated_reliability": 0.9505, "consumer_risk": 0.0485,
  "summary": "Test 59 units for 1,000 hours each with no failures to show …",
  "assumptions": ["…"], "tradeoff": { "failures": [0,1,2,3], "rows": [{ "label": "1,000 hours (1×)", "values": [59,93,124,153] }] } }

Push operational data (ingest)

Append data to existing artifacts — JSON or raw CSV, idempotent, atomic (a bad row is a 422 with no changes applied).

POST/api/ingest/fleets/{fleet_id}/usage

Update forecast items' current use, then recompute the forecast and check the fleet's alerts.

Request

`name / id`	csv col	required	Item name (or id).
`current_use`	csv col	required	Accumulated use so far (the meter reading).
`rate`	csv col	optional	Optional manual per-period usage rate.
`read_at`	csv col	optional	When the meter was read, ISO 8601 (e.g. 2026-09-27T08:00:00+10:00; no offset = UTC). Default: the request time. Readings older than the item's latest are ignored.

curl -X POST https://reliafy.com/api/ingest/fleets/FLEET_ID/usage \
  -H "Authorization: Bearer rlf_..." -H "Content-Type: text/csv" \
  --data-binary @meter_readings.csv          # name,current_use,read_at

Response · 200

{ "fleet": "…", "updated_items": 8, "forecast": { … }, "alerts_fired": 0 }

Usage-rate estimate. Each reading at least a day after the previous one updates the item's estimated usage per period: Δuse ÷ elapsed periods, smoothed (EWMA, α = 0.3). Elapsed time is converted with the fleet's period label — days, weeks, months (30.44 d), quarters or years; any other label gets no estimate. A reading lower than the last one (meter reset or replaced unit) restarts the baseline without estimating.
rate_source (fleet setting, "manual" by default): set it to "estimated" on the fleet page to forecast with each item's estimated rate, falling back to its manual rate, then the fleet default. The forecast's per_item rows report rate_used and rate_basis.
Alerts. Ingesting usage evaluates the fleet's alert rules (set on the fleet page) against the recomputed forecast and emails their owners — one email per crossing. alerts_fired counts the rules that fired; an alert problem never fails the ingest.

POST/api/ingest/tracking/{fleet_id}/measurements

Append (item, time, value) readings to a tracked fleet and re-predict remaining life.

Request

measurements object[] required [{item, time, value}, …] (JSON), or CSV columns item,time,value.

-d '{"measurements": [{"item": "Pump 7", "time": 5300, "value": 6.1}]}'

Response · 200

{ "updated": 1, "items": [ { … } ] }

POST/api/ingest/datasets/{dataset_id}/lives

Append failure rows to a dataset and refit its models in place.

Request

`refit`	bool · query	optional	Default true; ?refit=false to skip refitting.
`(rows)`	csv / json	required	Rows whose columns match the dataset.

Response · 200

{ "appended": 12, "n_rows": 142, "refit": [ { "id": "…", "name": "…" } ] }

Reliafy Maintenance users also have an ingest route for automated meter and condition readings; it’s described in the app.

Model import

POST/api/import/models

Create a model from a fit done elsewhere (what reliafy.push wraps).

Request

`name`	string	required	Model name.
`distribution`	string	required	SurPyval name or Reliafy id.
`unit`	string	optional	Time unit.
`data`	object	optional	{x, c?, n?} arrays — refit into a full model. Provide this or params.
`params`	{name,value}[]	optional	Parameters for a params-only model.
`options / extras`	object	optional	offset/zi/lfp (data) or gamma/p/f0 (params).

Response · 200

{ "id": "a1b2…", "name": "Pump bearings", "distribution": "Weibull",
  "params_only": false, "url": "/modelling/m/a1b2…" }