agentscope-ai/pawbench

Python

5 environment variables found in .env.example

Set these on your host before deploying. Each platform has an “Environment Variables” section in the deploy flow.

DASHSCOPE_API_KEY secret ── Model provider: DashScope (recommended default) ─────────────────────────── Used by --model dashscope/qwen3.6-plus etc.
OPENAI_API_KEY secret ── Model provider: OpenAI / OpenAI-compatible ──────────────────────────────── Used by --model openai/gpt-4o etc. Also serves as a generic…
OPENAI_BASE_URL secret
JUDGE_API_KEY secret ── LLM judge (required for llm_judge / hybrid graded tasks) ────────────────── Falls back to OPENAI_API_KEY / OPENAI_BASE_URL if unset.
JUDGE_BASE_URL

Best match Fly.io Anything Dockerized — requires the Fly CLI.

macOS

brew install flyctl && fly launch --from https://github.com/agentscope-ai/pawbench

Linux

curl -L https://fly.io/install.sh | sh && fly launch --from https://github.com/agentscope-ai/pawbench

Windows (PowerShell)

iwr https://fly.io/install.ps1 -useb | iex; fly launch --from https://github.com/agentscope-ai/pawbench

Or deploy to a different host