Technical Vision
AquaPilot is a single platform that delivers continuous pond intelligence to every farm, regardless of connectivity, infrastructure, or budget. One app, one data platform, multiple deployment models. Sensors feed data to AI that runs where it can: on the farmer's phone, on an edge device at the pond, or in the cloud. The architecture adapts to what's available on-site.
Technical Challenge
Business Challenges
Deployment Decision Tree
Three questions determine the hardware architecture at each pond. The answers define which deployment scenario applies.
Scenario Comparison
| A: Cloud Only | B: Phone AI | C: Low-Spec Edge | D: High-Spec Edge AI | |
|---|---|---|---|---|
| Data source | Camera, forms, GPS | Same as A + Sensors via WiFi | Same as A + Sensors via Edge HW | Same as A + Sensors via Edge HW |
| Edge hardware cost | None | None | Medium | High |
| AI location | Cloud only | On phone | On phone | On edge device |
| On-site diagnostics | No | Yes (on visit) | Yes (on visit) | Yes (on visit & 24/7 with SIM) |
| Real-time alerts | No | No | No | No / Yes (with SIM) |
| Internet required | For AI & sync | For sync only | For sync only | For sync / Yes (with SIM) |
| Best for | Entry-level, validation | WiFi-capable sensors | Cost-sensitive farms | High-value stock, hatcheries |
Data Platform Architecture
Five layers. Every data point accessible via API. The phone is both consumer (AquaPilot app) and producer (camera-based water testing). Extensible to any future data source.
Ingestion Protocols
| Protocol | Source | Pattern |
|---|---|---|
HTTPS POST | AquaPilot phone app (primary) | Phone uploads buffered edge data + camera data on WiFi. Compressed JSON batch. |
BLE | Edge gateway → technician phone | ~10KB/day per device. Full transfer in <1 sec. Clock sync on handshake. |
Phone camera | Test strips, spectrophotometer clip, visual inspection | Image capture, on-device processing, structured reading submitted with edge data. |
Webhook | External services, partner integrations | Event-driven, real-time |
Flat file | Lab results, CSV exports, manual data | Batch upload, validated on arrival |
Edge Gateway
Every pond gets an edge device. No SIM, no internet at pond level. The device buffers locally and syncs to the technician's phone via BLE on daily rounds. The phone uploads to the platform when it reaches WiFi.
Hardware & Software Requirements
Non-negotiable requirements for any edge device we deploy at pond level.
| Category | Requirement |
|---|---|
| Deployment | Proprietary software pre-loaded. Standard provisioning process (warehouse flash). |
| Operation | Headless (no screen, no keyboard). IP65 enclosure for outdoor deployment. |
| Connectivity | BLE to technician phone (primary uplink). Phone uploads on WiFi. Optional SIM/2G fallback. No per-pond SIM required. |
| Clock | Real-time clock (RTC) on device. Every reading timestamped locally. Clock sync on BLE handshake. |
| Security | Full disk encryption. TLS for all communication. Per-device certificates. Remote wipe. |
| Updates | OTA software updates. Signed updates only. Auto-rollback on failure. |
| Management | Remotely accessible. Central fleet control. SSH via VPN only. |
Questions for OxyGuard
Two architecture scenarios under evaluation. The answers below determine which path we take.
Scenario A: Phone as Edge Device (preferred)
The AquaPilot app on the technician's phone connects directly to Pacific units via local WiFi (no internet). Phone pulls sensor data, runs on-device AI, and syncs to the platform later.
| # | Question | Why we need it |
|---|---|---|
| 1 | WiFi connectivity on Pacific units | Does the Pacific expose any data interface (Modbus TCP, HTTP, or other) over WiFi, or only over the wired Ethernet port? The fish transport example (page 8, Example 2) shows an Access Point connected to the Pacific. What protocols are accessible over that wireless link? |
| 2 | Programmatic data access via WiFi | The Pacific Web Interface allows viewing readings in a browser over WiFi. Can a third-party app also pull readings programmatically (Modbus TCP, REST API, raw TCP) over the same WiFi connection? Or is it browser-only? |
| 3 | Modbus TCP over WiFi | If we connect a phone to the same local WiFi network as the Pacific unit, can the phone send Modbus TCP requests to the Pacific's IP address and receive sensor data? Any firewall or protocol restrictions? |
| 4 | Auto-discovery / mDNS | Is there a way to discover Pacific units on the local network automatically (mDNS, SSDP, broadcast)? Or do we need to know the IP address in advance? |
| 5 | Connection limits | How many simultaneous TCP/Modbus connections does a Pacific unit support? Can the Pacific Web Interface and a Modbus TCP client run at the same time? |
Scenario B: Local Edge Device as Bridge
If the Pacific only exposes Modbus TCP on the wired Ethernet port, we need a small device on the wired network that bridges data to the phone via WiFi or BLE.
| # | Question | Why we need it |
|---|---|---|
| 1 | Modbus TCP register map for Pacific Main Module | Required to read sensor values (DO, temp, salinity, pH, ORP) from our bridge device. Without this we cannot integrate. |
| 2 | Register data types and scaling factors | Integer vs float, scaling (e.g. value / 100 = actual reading), byte order (big/little endian). |
| 3 | Polling frequency limits | How often can we poll registers without impacting Pacific performance? Target: every 10-30 seconds. |
| 4 | Master Box aggregation via Modbus | When using a Pacific Master Box (up to 20 probes), are all probe values accessible via the same Modbus TCP connection? |
| 5 | Alarm/relay status registers | Can we read alarm thresholds and relay states via Modbus? |
| 6 | Simultaneous connections | Can our bridge device and a local HMI both connect via Modbus TCP at the same time? |