Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/Xander44-4/traffic_reducer/llms.txt

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

Manual mode lets you test the signal decision algorithm with arbitrary vehicle counts without needing a live camera feed or an active video source. It is the default state of the dashboard — the four range sliders are enabled from the moment the page loads, and every slider movement fires an immediate prediction request.

​
How manual mode works

When the Simulación en vivo toggle is off (the default), all four range sliders in the Manual Control section are enabled. Moving any slider immediately calls POST /predict with the current values from all four inputs. No debounce delay is applied — the request fires on every input event. The sliders and their identifiers:
Slider labelHTML input IDDirection index
Norteinput-north0
Surinput-south1
Esteinput-east2
Oesteinput-west3
Each slider accepts integer values from 0 to 100 (representing vehicle count) with a default of 10. A live numeric readout beneath each slider updates on every change.

​
Visual density feedback

As you drag a slider, the corresponding road arm in the intersection visualization updates its cars density bar opacity in real time using this formula from script.js: 
carsVisuals[direction].style.opacity = 0.05 + (value / 100) * 0.6;
A value of 0 produces opacity 0.05 (nearly invisible); a value of 100 produces opacity 0.65 (fully rendered). This gives immediate visual feedback about relative traffic load across all four directions before the prediction response arrives.

​
Prediction request and response

Every slider interaction sends this request to the server: 
// POST /predict
{
  "norte": 10,
  "sur": 55,
  "este": 10,
  "oeste": 10
}
The server computes the winning direction using np.argmax([norte, sur, este, oeste]) and responds with the zero-based direction index and the echo of the submitted counts: 
{
  "prediction": 1,
  "traffic_data": {
    "norte": 10.0,
    "sur": 55.0,
    "este": 10.0,
    "oeste": 10.0
  }
}
prediction maps to directions as follows: 0 = NORTE, 1 = SUR, 2 = ESTE, 3 = OESTE. On receiving the response the dashboard:
  1. Activates the winning direction’s green traffic light bulb in the intersection graphic.
  2. Sets all other lights to red.
  3. Updates the Prioridad activa status card with the direction name.
  4. Highlights the winning road arm with the .active-road CSS class.

​
Randomized scenarios with /simulate

To quickly generate a realistic but random traffic scenario, call the /simulate endpoint directly: 
curl http://127.0.0.1:5000/simulate

{ "norte": 17, "sur": 8, "este": 42, "oeste": 61 }
The endpoint returns four random integers between 0 and 80, one per direction. The values are not automatically applied to the sliders via the API — use the response data to manually set slider positions, or integrate the endpoint into your own test script.

​
Priority display edge cases

The Prioridad activa card can show values other than a direction name:
  • EMERGENCIA (rendered in #b64400) — appears when traffic_data.emergency is true in the prediction response. This overrides the direction name regardless of which direction won.
  • PEATONES (rendered in #707070) — appears when traffic_data.pedestrians exceeds 5.
In manual mode with the /predict endpoint, emergency and pedestrians are not part of the slider payload. They will only appear in the priority display if the server-side model or camera pipeline injects them into the response.
Set one direction to 100 and leave all others at 0 to verify the green light correctly assigns to that direction. This is the fastest sanity check after a new deployment — if the light does not match the maxed-out slider, confirm the /predict route is reachable and the model loaded successfully at startup.

Build docs developers (and LLMs) love

Get started for freeTalk to us