Ollama is an open-source platform that allows you to run large language models (LLMs) locally on your own infrastructure. All AI processing happens entirely within your network — no data is sent to any external server. This means sensitive operational data such as device details, alarm information, and performance metrics never leave your controlled environment, making it suitable for organizations with strict data privacy policies, regulatory compliance requirements, or air-gapped networks.
OpManager supports Ollama integration, enabling AI-powered features such as alarm summarization, script generation, and device insights — all powered by a locally hosted model with complete data sovereignty.
For detailed information on installing, configuring, and managing Ollama, refer to the official Ollama documentation:
Before integrating Ollama with OpManager, ensure the following requirements are met.
Ollama must be installed and running on a machine accessible from the OpManager server. Follow the installation instructions on the Ollama website for your operating system. After installation, verify Ollama is running:
curl http://localhost:11434/api/tagsA model must be downloaded before OpManager can use Ollama. To pull a model:
ollama pull <model-name>Verify the model is available:
ollama listEnsure the machine running Ollama meets the minimum hardware requirements for the model you intend to use:
| Tier | Model Size (Parameters) | Minimum RAM (CPU) | Recommended RAM | Recommended VRAM (GPU) |
|---|---|---|---|---|
| Entry-level | 3B — 4B | 4 GB | 8 GB | 3—4 GB |
| Low-mid | 5B — 7B | 8 GB | 16 GB | 5—6 GB |
| Mid-range | 8B — 9B | 12 GB | 16—24 GB | 6—8 GB |
| Upper-mid | 10B — 14B | 16 GB | 24—32 GB | 10—12 GB |
| High-end | 20B — 34B | 32 GB | 48—64 GB | 16—24 GB |
| Enterprise | 65B — 70B+ | 48 GB | 64—128 GB | 40—48 GB+ |
OLLAMA_HOST=0.0.0.0 ollama serve Only for the initial model download (ollama pull). After that, Ollama runs entirely offline. In air-gapped environments, model files can be transferred manually.
AI features in OpManager will return an error message. Core monitoring functionality — alerts, dashboards, reports — is not affected. AI features are supplementary.
Yes, and this is recommended for production environments. Run Ollama on a dedicated machine with adequate resources and configure OpManager with its IP address and port.
Ollama supports multiple models, but OpManager uses the model configured in the settings at any given time. You can change the active model as needed.
Yes. Ollama is open-source and free to use. The only cost is the hardware to run it.
OpManager's AI integration follows a consistent pattern across supported providers — API endpoint configuration, model selection, and AI-generated summaries for alarms, devices, and device groups. For a detailed walkthrough of this pattern, see Integrating OpManager with OpenAI.
OpManager can generate AI-powered summaries for individual alarms, all alarms, device groups, and individual devices. Anomaly data surfaced by OpManager's built-in Zia engine can also be interpreted and summarized through a connected LLM. See Anomaly Detection in OpManager for more detail.
Alert notifications are configured separately from AI features and remain fully functional regardless of whether an LLM is connected. AI-generated summaries are supplementary to — not a replacement for — standard alarm workflows. See Configuring Notifications for setup instructions.
OpManager supports a wide range of integrations across alert management, analytics, help desk, IT automation, and AI. Ollama falls under the AI Integration category. See the OpManager Integrations Overview for a full list.