Skip to main content

airgap_qwen

What it is​

  • A module that defines a preconfigured ChatModel instance for the Qwen3 chat model running in airgap mode via a local Docker Model Runner endpoint.

Public API​

  • Module constants
    • MODEL_ID: "ai/qwen3"
    • PROVIDER: "qwen"
    • NAME: "qwen3-airgap"
    • DESCRIPTION: Human-readable description of the model setup
    • IMAGE: URL to a representative image
    • CONTEXT_WINDOW: 8192
  • Module instance
    • model: ChatModel: A naas_abi_core.models.Model.ChatModel wrapping a langchain_openai.ChatOpenAI client configured to talk to a local engine.

Configuration/Dependencies​

  • Python dependencies
    • langchain_openai.ChatOpenAI
    • naas_abi_core.models.Model.ChatModel
  • Runtime configuration (hardcoded in module)
    • base_url="http://localhost:12434/engines/v1" (expects a local service listening there)
    • model="ai/qwen3"
    • temperature=0.7

Usage​

from naas_abi.models.airgap_qwen import model

# Use `model` with the APIs expected by `ChatModel` in your environment.
# Example access to metadata:
print(model.model_id, model.provider, model.name)

Caveats​

  • Requires a local engine reachable at http://localhost:12434/engines/v1; otherwise requests made through the underlying client will fail.
  • This module only defines configuration and instantiates objects; it does not expose helper functions for invoking the model.