Zubax CANFace CF3D CAN bus data loggers collect CAN frames into on-device persistent storage, like a black box flight recorder, and store them until wireless connectivity to a predefined network becomes available (e.g., during vehicle service or maintenance). As soon as connectivity is available, the CAN frames collected during the offline period are uploaded in FIFO order to the data collection server. Every frame that is confirmed to be received by the server is erased from the device, enabling completely unattended continuous persistent data collection across the entire fleet of vehicles. The on-board storage is large enough to store at least multiple days or weeks worth of data with highly robust FEC encoding to avoid loss or damage.
This project implements the data collection server of this system. It is a simple Python application that exposes a REST API that is used by dataloggers to upload data when within reach of the wireless network, and by clients to review and fetch stored data of interest (with rich filters). This system enables advanced diagnostics, postmortem analysis, predictive maintenance, and data mining at scale across the entire fleet, without the need to manually collect data per vehicle, as long as each vehicle appears within the range of a predefined wireless network at least intermittently.
The server can be deployed in airgapped networks, ensuring 100% locality and security of the collected datasets. Ingest scripts are available to bulk upload data from datalogger memory cards directly in case of damage (e.g., following accidents where the vehicle is damaged or destroyed). The data is stored in a simple SQLite database that can also be accessed directly for more advanced usage.
graph TD
CAN["🔌 CAN bus<br/>every frame from every ECU"] --> CF3D
subgraph FLEET["✈️ 🚗 🚢 Vehicles"]
CF3D["CF3D Black Box Recorder<br/><i>on-device storage<br/>weeks of data</i>"]
end
CF3D == "📡 auto-upload over Wi-Fi<br/>during service/maintenance<br/>fully unattended" ==> COMMIT
subgraph NESTOR["💾 Nestor Server"]
COMMIT["Ingest<br/>FEC repair records"] --> DB[("SQLite<br/>append-only · idempotent<br/>airgap-safe")]
DB --> API["REST API<br/>devices · boots · records<br/>filtering · long-polling"]
end
COMMIT -. "☑️ ACK<br/>device erases confirmed records" .-> CF3D
API ==> D["Diagnostics"]
API ==> P["Postmortem<br/>Analysis"]
API ==> M["Predictive<br/>Maintenance"]
API ==> F["Fleet-Scale<br/>Data Mining"]
classDef can fill:#e2e8f0,stroke:#475569,color:#1e293b
classDef recorder fill:#880000,stroke:#FF0000,color:#fff,font-weight:bold
classDef ingest fill:#d1fae5,stroke:#059669,color:#064e3b
classDef db fill:#059669,stroke:#047857,color:#fff,font-weight:bold
classDef api fill:#d1fae5,stroke:#059669,color:#064e3b
classDef outcome fill:#fef08a,stroke:#ca8a04,color:#713f12,font-weight:bold
class CAN can
class CF3D recorder
class COMMIT ingest
class DB db
class API api
class D,P,M,F outcome
For details, please reach out to [email protected].
Install and run the server at the default endpoint http://0.0.0.0:8000
pip install zubax-nestor # For development, use `pip install -e .` for editable installation
nestor serve # See --help for extra info.Open the API docs at the /docs endpoint in your browser, e.g., http://localhost:8000/docs.
The server includes a web-based GUI for browsing CAN data. If the gui/dist/ directory exists, it will be served at the root path (/).
cd gui && npm install && npm run build
cd ..
nestor serve --db data/
# Open http://localhost:8000Both the GUI and API are served from the same server — no separate frontend process needed.
For active frontend development with hot reload, run two processes:
# Terminal 1: API server
nestor serve --db data/dev.db
# Terminal 2: Vite dev server (proxies /cf3d/api/v1 to localhost:8000)
cd gui && npm run dev
# Open http://localhost:5173- Device selector dropdown
- Boot session browser with timestamps
- CAN records table with filtering
- Mobile-responsive layout
Run behind a local gateway via Unix socket:
nestor serve --uds /run/nestor/nestor.sockEvery runtime option can be configured with env vars like NESTOR_HOST/.._PORT, NESTOR_DB_PATH, etc.
CLI arguments override environment values.
Configure Zubax CANFace CF3D devices to use the endpoint where the server is running.
Existing dumps from CF3D memory cards can also be uploaded manually using nestor_ingest.py.
You can run the ASGI app through an external process manager/gateway using the app factory:
export NESTOR_DB_PATH=/var/lib/nestor/nestor.db
export NESTOR_LOG_FILE=/var/log/nestor/server.log
export NESTOR_LOG_LEVEL=INFO
uvicorn --factory nestor.server:create_app_from_envPOST /cf3d/api/v1/commit - Upload one or more binary CF3D records; the first HTTP response line is the cumulative ACK (last_seqno).
This is the endpoint that is invoked by dataloggers to transfer previously recorded CAN frames from their internal storage.
GET /cf3d/api/v1/devices - List known devices with their latest heartbeat time (last_heard_ts) and last seen hardware UID (last_uid).
Use this to see which dataloggers have committed their data and when each was last seen online.
GET /cf3d/api/v1/boots - List boot sessions for a device, including first/last record per boot, with optional commit-time window filtering.
GET /cf3d/api/v1/records - Fetch records for a device and one or more boot IDs, with optional seqno bounds and long-polling for real-time streaming.
The REST API is very simple and can be exercised manually using wget or similar tools. Examples (jq is added for nicer formatting):
wget -qO- "http://localhost:8000/cf3d/api/v1/devices" | jqwget -qO- "http://localhost:8000/cf3d/api/v1/boots?device=my+device" | jqwget -qO- "http://localhost:8000/cf3d/api/v1/records?device=my+device&boot_id=1" | jq
