Qwen / DashScope

TapPass supports Qwen models via Alibaba Cloud’s DashScope API, powered by LiteLLM.

Configuration

# Server-side (TapPass stores the key in its encrypted vault)
DASHSCOPE_API_KEY=your-dashscope-api-key

Get your API key from the DashScope console.

Available models

Model	Context Window	Best for
dashscope/qwen-turbo	131K	Fast, cost-effective tasks
dashscope/qwen-plus	131K	Balanced performance and cost
dashscope/qwen-max	32K	Most capable, complex tasks
dashscope/qwen-max-longcontext	1M	Long document processing
dashscope/qwen-vl-max	32K	Multimodal (vision + language)
dashscope/qwen-vl-plus	32K	Multimodal, cost-effective

Usage

from tappass import Agent

agent = Agent("http://localhost:9620", "tp_...")

# Qwen Turbo (fast, cheap)
response = agent.chat("Hello", model="dashscope/qwen-turbo")

# Qwen Max (most capable)
response = agent.chat("Analyze this contract", model="dashscope/qwen-max")

# Qwen Plus (balanced)
response = agent.chat("Summarize this report", model="dashscope/qwen-plus")

Zero-code

export OPENAI_BASE_URL=http://localhost:9620/v1
export OPENAI_API_KEY=tp_...
# Any OpenAI-compatible tool now routes through TapPass

LLM Judge

Use Qwen as the governance pipeline’s LLM judge:

TAPPASS_LLM_JUDGE_MODEL=dashscope/qwen-turbo
# or
TAPPASS_LLM_JUDGE_MODEL=dashscope/qwen-plus
# or
TAPPASS_LLM_JUDGE_MODEL=dashscope/qwen-max

Data residency

DashScope endpoints are hosted in mainland China. If you require EU data residency (TAPPASS_EU_DATA_RESIDENCY=true), use Qwen models via Ollama (self-hosted) instead:

TAPPASS_LLM_JUDGE_MODEL=ollama/qwen2.5

All requests pass through the full governance pipeline regardless of provider.