Bring AI into your Niagara Station — query, analyze, automate, and decide, in natural language.
The Niagara AI Agent is a native Niagara 4 module that connects your station to large language models via the OpenAI-compatible API format. Instead of writing BQL queries by hand or clicking through Workbench, you can ask questions in plain English and get structured results written directly into Niagara points and histories.
✅ Tested with: DeepSeek via OpenAI-compatible API
🔧 Compatible with: Any LLM that supports the OpenAI API format (GPT-4o, local ollama, vLLM, LM Studio, etc.) — not individually tested but expected to work.
| You Ask | The AI Agent Does |
|---|---|
| "What's the current temperature on floor 3?" | Runs a BQL query, returns the value to a Niagara point |
| "Compare today's energy usage to last Tuesday" | Pulls history data, runs the comparison, writes result |
| "Alert me if any AHU supply temp exceeds 28°C" | Sets up an alarm condition using natural language |
| "Summarize yesterday's alarms in zone B1" | Queries alarm console, formats a summary, stores it |
| "Generate a daily energy report and email it" | Aggregates metering data, composes report, triggers email |
# 1. Install gline.pem certificate into your station trust store
# (available in this repository under /certs/)
# 2. Deploy the module
# Place AIbotAgent-rt.jar in your Niagara modules/ directory
# 3. Get a trial license
# Send your Host ID → jason.zhang@gline-net.com
# We'll reply with a temporary license within 24 hours
# 4. Configure your AI provider
# Open the AI Agent service in Workbench:
# - Set base URL: https://api.deepseek.com (or your LLM endpoint)
# - Enter your API key (encrypted with included ApiKeyEncryption tool)
# - Choose your model name
# 5. Start asking questions
# Wire the AI Agent input/output to any Niagara ControlPointIf your organization requires a certificate signed with your own key:
# 1. Provide us with your certificate file (.pem) and password
# 2. We recompile and re-sign the JAR with your certificate
# 3. Deploy the custom-signed JAR to your station
# 4. No external certificate installation neededContact us for custom signing requests.
| Provider | Status | Notes |
|---|---|---|
| DeepSeek | ✅ Tested | Recommended for cost efficiency |
| OpenAI (GPT-4o, etc.) | 🔧 Expected to work | OpenAI-compatible format |
| Local LLM (ollama, vLLM) | 🔧 Expected to work | Set your local endpoint as base URL |
| Any OpenAI-compatible API | 🔧 Expected to work | Same format, not individually tested |
- Queries BQL / NEQL and returns structured results
- Writes outputs directly to ControlPoints, histories, and alarm summaries
- Works inside PX pages, Program objects, and Workbench
- Low-latency async design — won't block your station
| Tool | Purpose |
|---|---|
| TokenCalculator | Estimate token usage before calling the API |
| HeapMonitor | Monitor JACE memory impact in real time |
| ApiKeyEncryption | Securely store provider API keys in station database |
| h |
- API keys encrypted at rest in station database
- No external outbound calls except to the configured LLM endpoint
- Optional on-premise / air-gapped LLM support for sensitive sites
- JAR can be re-signed with your own certificate on request
| Tier | License | Price | Best For |
|---|---|---|---|
| 🆓 | Trial | Free | Evaluation, testing, proof-of-concept |
| 🥇 | Single Station | $299 | One JACE / Supervisor |
| 🏢 | Site Pack (5 stations) | $999 | Multi-controller projects |
| 🌐 | Enterprise | Contact us | Large deployments, OEM, white-label |
Trial license is valid for 30 days. No credit card required. All commercial licenses include 12 months of updates and email support.
| Component | Requirement |
|---|---|
| Niagara Version | 4.14 or later (JACE, Edge, Supervisor, Windows) |
| JAR Signing | Pre-signed with gline.pem, or custom-signed with your certificate on request |
| Network | Outbound HTTPS to your LLM provider |
| JACE Memory | 512 MB+ heap recommended (tested on JACE-8000) |
Q: Do I need an internet connection?
A: Yes, to reach the LLM API. For air-gapped deployments, we support on-premise local LLMs via custom endpoint configuration.
Q: Which LLMs have you tested?
A: We have tested DeepSeek (via OpenAI-compatible API). Any LLM that supports the OpenAI API format should work, but has not been individually tested.
Q: Does the AI module impact JACE performance?
A: Each inference call is asynchronous and lightweight (non-blocking). Use the included HeapMonitor to verify on your hardware.
Q: Can I use my own API key?
A: Yes. Configure any OpenAI-compatible endpoint with your own key. All keys are encrypted at rest.
Q: How long does a query take?
A: Typically 1–5 seconds depending on LLM provider and query complexity.
Q: Is the trial fully functional?
A: Yes — full features, no limitations. Only the license duration (30 days) is restricted.
Q: Can I use my own certificate instead of gline.pem?
A: Yes. Provide us with your certificate and password. We will recompile and re-sign the JAR with your certificate at no additional charge.
| Manual | Description |
|---|---|
| AIbot Technical Reference | Complete API reference, configuration guide, and usage examples |
| ApiKey Encryption | How to securely store and manage your LLM API keys |
| StringToFile | Save AI-generated text output directly to station files |
| TokenCalculator | Estimate token usage and API costs before inference |
├── AIbotAgent-rt.jar ← Pre-signed module (gline.pem)
├── certs/
│ └── gline.pem ← Default certificate
├── docs/ ← User manuals (PDF)
├── LICENSE
├── README.md
└── src/ ← Source code (available on request)
- Email: jason.zhang@gline-net.com
- WhatsApp: +86 138 0199 0968
- Web: www.glineweb.com
Shanghai Gline Net Co., Ltd. — Your Partner in Smarter Automation
This module is provided under a proprietary commercial license.
- Free Trial: 30-day evaluation, fully functional
- Commercial: Paid license required for production use
See LICENSE for full terms.
© 2026 Shanghai Gline Net Co., Ltd. All rights reserved.