How this network is built, run, and governed.
Each document is dated and revision-controlled. They describe an independent student-operated pilot — independent student-run studio · not a government body.
What the network is made of and how a detection travels from the roadside to an operator.
The contract a node must satisfy: endpoint, schema, authentication, error codes, and retry behaviour.
What the network collects, how long it is kept, and who can reach it.
How incidents are triaged, escalated, and closed, and how nodes are maintained.
Which specifications the design references — and, plainly, what has not been certified.
What the network does not detect, where it degrades, and what a human still has to do.
A supervised 10-node trial on a ~500 m Delhi stretch: layout, what we measure, cost and what we need.
Dated record of material changes to the system and this register.
Who is responsible for what.
A named party answers for every part of the system.
| Role | Who | Scope |
|---|---|---|
| Operator of record | Kavy | Platform, ingest API, incident triage, publication of this site. |
| Hardware maintainer | Saksham | Node assembly, enclosure, field installation, sensor calibration. |
| Alert recipients | On-duty operator | Acknowledges and resolves incidents; escalates per the matrix. |
| Data controller | The project operators jointly | Retention, access requests, and deletion. |
What happens, and how fast.
Response targets are commitments the operators hold themselves to, not guarantees to third parties.
| Severity | Examples | Respond within | Action |
|---|---|---|---|
| Critical | Possible collision, sustained smoke or gas | 10 minutes | Operator acknowledges, verifies against the node feed, notifies emergency services if confirmed. |
| Warning | Lane obstruction, wet surface, node unreachable | 2 hours | Operator acknowledges and raises a maintenance task if the condition persists. |
| Info | Threshold brushes, recovered links, routine notices | Next working day | Reviewed in the daily log; no action unless a pattern appears. |
The numbers we publish, with their basis.
Each figure is labelled by how it was arrived at.
| Service level | Target | Basis | Provenance |
|---|---|---|---|
| Detect → alert latency | ≤ 2 s | From on-device detection to an incident row being written. | target |
| Node reporting uptime | ≥ 98 % / 30 days | Share of expected telemetry windows in which a packet arrived. | measured |
| Critical acknowledgement | ≤ 10 min | Time from a critical incident opening to an operator acknowledging it. | target |
| Offline detection | ≤ 15 min | A silent node is flagged by the sweeper within one window. | measured |