Design Philosophy
Less debug with AI — deliver platform capability to the project site. AI understands, answers accurately; users just use it.
The most expensive parts of an IoT project are rarely cloud VMs. They are two things:
- People — days spent matching points, on-site experts, night triage;
- Data — no dedicated owners, no governance team; the longer you run, the less you trust reports and decisions.
Lingdian AIoT is designed to cut repetitive human days, put capability in the field, and keep data governable and traceable so trust grows with time.
Why buy: what the field is burning
| Pain | Cost today | How Lingdian cuts it |
|---|---|---|
| Point tables three times | Manuals, registers, models rewritten; one wrong address burns a night | AI reads tables / payloads; generate models — configure by review |
| Every device needs a one-off script | Protocol change means core patches and downtime | Hot-plug protocol plugins; brownfield via templates |
| Demo looks great; field needs experts | Laptops and Wireshark on site | Gateway + admin / mini program; dual-track logs remotely |
| The phone never stops | Offline, garbled data, OTA failures escalate | Context-aware AI Q&A; common issues self-serve |
| Nobody owns the data; trust decays | No governance team; same name different meaning, mixed units, dirty data into business; reports become unusable | Standard parameters / thing models, quality and lineage by platform policy; raw + normalized dual track; dirty data does not flow straight into business |
Buy logic: fewer on-site days and support calls — plus less of the silent debt where “three years later nobody can explain the numbers.” Human days get accepted at delivery; trustworthy data keeps the operation alive.
Design purpose
| Purpose | On a real project |
|---|---|
| Eliminate repetitive action | Automate point matching, register copying, FAQ triage |
| Reduce waste of life | Do not burn engineers’ nights on capture and guesswork |
| More humane experience | Field and business users operate in business words — just use it |
Design thinking
| Thinking | How we apply it |
|---|---|
| Common-sense design | Admit what burns money first; no empty empowerment |
| Simple methods for complex problems | AI less debug, plugins for protocols, one-way layers |
| Higher abstraction | Business speaks device / property / service / event |
| Standard components | Standard models, parameters, hot-plug plugins; reuse the base |
Data governance follows the same path: when there is no dedicated team, standards and platform rules hold the definitions, instead of hoping every project hires its own governance staff.
Design principles
Concepts & naming · MECE
Mutually exclusive, collectively exhaustive — clear structure. Protocol / thing model / rules / alarms / governance boundaries do not overlap or leave gaps (otherwise the field will fight over “whose point is this?”).
Openness
Every parallel capability must be open and extensible (wherever multiple options exist). New protocols, industries, and governance rules ship as modules — no core rewrite, no locked option set.
Minimization
| Principle | Meaning |
|---|---|
| Fewest clicks | Short paths for onboard and daily work |
| Least code | Extend via plugins and config, not core forks |
| Least mental load | Default path works; AI answers accurately |
| Delete if possible | Drop steps and intermediate states; no feature piles |
Consistency
- Backward compatible — upgrades do not break existing onboard.
- Do not break habits — business vocabulary stays stable.
- Consistent UX — admin and mini program share expectations.
Consistency applies to data too: same point, same unit, same coding rules across projects — or trust decays the longer you run.
Three roles, what each gets
| Who | Friction | Value |
|---|---|---|
| PM / integrator | Schedule slip, on-site overrun, unexplained data after acceptance | Faster onboard; reusable projects; governable definitions |
| Commissioning / ops | Registers, night calls, dirty data with no source | Less debug; remote observability; dual-track trace |
| Business / end users | Black-box platform; reports they dare not trust | Business words; data gets more trustworthy — just use it |
How to verify (in a demo)
- Table / payload → thing model — reviewable, not another night of typing.
- Field path to online — basic onboard without an expert laptop.
- Remote clarity — raw and normalized searchable; Trace to UI.
- Accurate answers — this device / point / log.
- Definitions hold — standards exist; synonym and unit chaos can be caught; dirty data does not enter business unchecked.
Claims vs. returns
| Claim | You save | You gain |
|---|---|---|
| Less debug with AI | Point days, rework, night overtime | Controllable schedule |
| Capability to the site | Resident experts, return visits | Field can progress; repeatable delivery |
| Understands, answers, just use | Tickets, training, doc diving | Business users run it; experts on hard cases |
| Governable, trustworthy data | Silent data debt, bad decisions | Trust without a full-time governance team |
Less debug is not less capability — complexity stays in the platform; simplicity reaches the field; standards stay in the platform; trust accrues over time.
Technical landing
Protect today’s savings when you scale. Prefer plugins over core changes.
Device → Go (data plane) → Node (control plane) → Vue (presentation)
No cross-layer shortcuts. Plugins for protocols. Dual-track raw + normalized. Standard parameters / models / quality / lineage so definitions hold without a dedicated team. Same boundaries from one host to cluster.
Summary
| Promise | One line |
|---|---|
| Less debug · to the site | Cut human days; capability reaches the field |
| Just use | Accurate answers; business users run it |
| Trustworthy data | Govern without a full-time team; trust grows |
| Purpose | Kill repetition, save life, humane UX |
| Principles | MECE · open · minimize · consistent |
| Verifiable · scalable | Demo proves it; same boundary to cluster |
Next: System Architecture; Features.