AI Assistant
How to get an API Key for AI Assistant? Which LLM providers are supported?
You need to provide your own LLM API Key:
- DeepSeek: Register at platform.deepseek.com, supports
deepseek-v4-flashanddeepseek-v4-promodels - OpenRouter: Register at openrouter.ai, supports x-ai, OpenAI, Google, Anthropic and more
Configure the provider and API Key in the AI Profile settings page.
How to choose and configure models other than DeepSeek?
The documentation primarily demonstrates using DeepSeek, but the AI Assistant also supports other models through OpenRouter.
Base Models tab quick-select:
- DeepSeek:
deepseek-v4-flash/deepseek-v4-pro - OpenRouter:
x-ai,z-ai,openai,google,anthropicand more
To use models not listed in the quick-select, switch to the Custom Mode tab for custom configuration (custom API endpoints, model names, temperature, etc.).
Compatible with large language models that support the OpenAI API format. If your model provider is compatible with the OpenAI API format (such as vLLM, Ollama, etc.), you can fill in the corresponding API endpoint in Custom Mode.
Local small-parameter models are not recommended. The AI Assistant requires a large context window to accommodate topology information, conversation history, and tool call records. Small-parameter models have insufficient context windows, which may cause functional issues or severely degraded response quality.
Is my API Key stored securely in GNS3?
Yes. API Keys are protected by:
- Encryption using Python's
cryptography.fernetmodule (AES-128-CBC algorithm) - Stored encrypted in the database; all API query endpoints return
api_key: null - Only decrypted internally when the AI Assistant actually calls the LLM
Does AI Assistant require internet access?
Yes. The AI Assistant needs to connect to the configured LLM provider's API endpoint. In air-gapped environments, ensure access to the LLM API is available or deploy a local LLM service.
What device types does AI Assistant support?
The following device types have been tested and verified:
| Device Type | device_type tag | Notes |
|---|---|---|
| Cisco IOS | cisco_ios_telnet | Configuration command generation is most mature and smooth |
| Huawei CE | gns3_huawei_telnet_ce | Requires accurate skills configuration |
| Ruijie | gns3_ruijie_telnet | Requires accurate skills configuration |
| VPCS | gns3_vpcs_telnet | Skills built-in |
Devices must have the correct device_type tag configured. The AI Assistant uses this tag to identify device types and call appropriate configuration skills. For more supported Telnet device types, refer to GNS3 Server - Netmiko Supported Devices.
Currently, testing has primarily focused on Cisco IOS images without a console login password set.
What's the difference between Teaching Assistant and Lab Automation Assistant modes?
| Mode | Tool Scope | Use Case |
|---|---|---|
| Teaching Assistant | Read-only diagnostics + packet analysis + node management | Teaching, fault diagnosis guidance |
| Lab Automation Assistant | All tools (including configuration changes) | Lab automation, batch device configuration |
Teaching Assistant only executes read-only commands (view config, diagnostics), suitable for learning and troubleshooting. Lab Automation Assistant can execute configuration changes, suitable for automated deployment.