Skip to content

linux.md

Latest commit

 

History

History
63 lines (43 loc) · 1.11 KB

linux.md

File metadata and controls

63 lines (43 loc) · 1.11 KB

Linux/macOS Guide

Use this guide when running large-scale experiments on Linux/macOS.

1) Prerequisites

  • Python 3.10+
  • uv
  • openclaw CLI available in PATH
  • Running OpenClaw/EcoClaw runtime

2) Environment

From repo root:

cp .env.example .env

Set at least:

ECOCLAW_API_KEY=sk-xxx
ECOCLAW_BASE_URL=https://concept.dica.cc/llm
ECOCLAW_MODEL=claude-haiku-4.5
ECOCLAW_JUDGE=gpt-5-nano

If PinchBench skill is not at ../skill or ~/skill, set:

ECOCLAW_SKILL_DIR=/absolute/path/to/skill

3) Run

chmod +x experiments/scripts/*.sh

Single-task sanity check:

./experiments/scripts/run_pinchbench_baseline.sh --suite task_00_sanity --runs 1
./experiments/scripts/run_pinchbench_ecoclaw.sh --suite task_00_sanity --runs 1

Full automated suite:

./experiments/scripts/run_pinchbench_baseline.sh --suite automated-only --runs 1
./experiments/scripts/run_pinchbench_ecoclaw.sh --suite automated-only --runs 1

Compare:

./experiments/scripts/compare_pinchbench_results.sh

Report output:

  • results/reports/pinchbench_comparison.json