Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/luiss811/Backend-Airguide/llms.txt

Use this file to discover all available pages before exploring further.

AirGuide includes a TensorFlow.js neural network that predicts route congestion based on the current hour and a route’s historical usage patterns (contador_usos). The model produces a score between 0 and 1 — a score above 0.7 is considered congested. Three endpoints expose this functionality: one for checking a single route, one for generating a full heatmap overlay, and one admin endpoint for retraining the model when usage data has grown significantly.

​
POST /api/rutas/check-congestion

Predicts whether a specific route is congested at the current hour. No authentication is required.
A score above 0.7 is classified as congested. The model uses the server’s current local hour (0–23) as its primary input alongside learned usage patterns for the route.

​
Request body

​
id_ruta
integer
required
The id_ruta of the route to evaluate.

​
Response — 200 OK

​
congested
boolean
true if the predicted congestion score exceeds 0.7, false otherwise.
​
score
number
Raw AI prediction score between 0 and 1. Higher values indicate heavier predicted congestion.
​
hour
integer
The current hour (0–23) used as input to the model. Useful for debugging or logging.

​
Error responses

StatusBodyCause
400{ "error": "Falta id_ruta" }id_ruta was not included in the request body.
500{ "error": "Error analizando congestión" }Model inference failed.

​
Example

curl -X POST http://localhost:3001/api/rutas/check-congestion \
  -H "Content-Type: application/json" \
  -d '{"id_ruta": 1}'

{
  "congested": true,
  "score": 0.84,
  "hour": 10
}

​
GET /api/rutas/heatmap

Returns congestion scores and waypoint paths for all active internal routes. Intended for rendering a color-coded congestion heatmap overlay on the campus map. No authentication is required. The server evaluates the AI model for every active interna route concurrently using the current hour, then returns the scores alongside each route’s waypoints formatted for direct use with mapping libraries.

​
Response — 200 OK

Array of objects, one per active internal route.
​
id_ruta
integer
Route identifier.
​
score
number
AI congestion score between 0 and 1 for this route at the current hour.
​
path
object[]
Ordered array of waypoints derived from the route’s RutaDetalle records.

​
Example

curl http://localhost:3001/api/rutas/heatmap

[
  {
    "id_ruta": 1,
    "score": 0.84,
    "path": [
      { "lat": 19.43260, "lng": -99.13320 },
      { "lat": 19.43280, "lng": -99.13300 },
      { "lat": 19.43290, "lng": -99.13285 }
    ]
  },
  {
    "id_ruta": 4,
    "score": 0.21,
    "path": [
      { "lat": 19.43310, "lng": -99.13350 },
      { "lat": 19.43330, "lng": -99.13370 }
    ]
  }
]

​
POST /api/analytics/train-congestion

Triggers retraining of the TensorFlow.js congestion model using all current route usage data. Admin access is required.
Retrain the model after a significant volume of new route usage data has accumulated — for example, after a new semester begins or after deploying new routes. Retraining on sparse data produces an unreliable model. The returned loss value indicates training accuracy; lower is better.
Requires a valid admin Bearer token. This endpoint is on the analytics service, not the navigation service.

​
Request headers

​
Authorization
string
required
Bearer token for an admin account: Bearer <token>.

​
Response — 200 OK

​
message
string
Confirmation string: "Modelo de congestión entrenado con éxito".
​
loss
number
Final training loss value. Lower values indicate a better-fitted model.

​
Example

curl --request POST \
  --url http://localhost:3001/api/analytics/train-congestion \
  --header 'Authorization: Bearer <admin-token>'

{
  "message": "Modelo de congestión entrenado con éxito",
  "loss": 0.0412
}

Build docs developers (and LLMs) love

Get started for freeTalk to us