config.yaml.example

# EvoAgentBench configuration
# Copy to config.yaml and edit for your environment
#
# Setup:
#   1. cp config.yaml.example config.yaml
#   2. cp .env.example .env && edit .env (fill in API keys)
#   3. Copy one agent *.yaml.example to its corresponding *.yaml and edit it
#   4. Select one of the four domains below
#
# Relative paths in this file are resolved from this file's directory.

# ── Agent configs ────────────────────────────────────────────────
# Map of agent name -> yaml config path.
# Copy the .yaml.example to .yaml and edit for your environment.
agent_configs:
  nanobot: ./src/agents/nanobot/nanobot.yaml
  openclaw: ./src/agents/openclaw/openclaw.yaml

# Active agent (must be a key in agent_configs above)
agent: nanobot

# ── Benchmark ─────────────────────────────────────────────────────
# REQUIRED: select exactly one paper domain and its matching config:
#   information_retrieval -> ./src/domains/information_retrieval/information_retrieval.yaml
#   code_implementation   -> ./src/domains/code_implementation/code_implementation.yaml
#   software_engineering  -> ./src/domains/software_engineering/software_engineering.yaml
#   knowledge_work        -> ./src/domains/knowledge_work/knowledge_work.yaml
# Each domain config contains:
#   - data paths (dataset_file, split_file, index_path)
#   - judge LLM config (information_retrieval: judge.model, judge.api_base, judge.api_key)
#   - agent_timeout, MCP server settings
domain:
  name: information_retrieval
  config: ./src/domains/information_retrieval/information_retrieval.yaml

# ── Output ────────────────────────────────────────────────────────
job_dir: ./jobs

# ── Run settings ──────────────────────────────────────────────────
trials: 1                            # pass@k: run each task N times
parallel: 1                          # max concurrent tasks
max_retries: 2                       # retry on failure

# Docker settings (optional, for Software Engineering)
# docker: docker
# docker_sock: /var/run/docker.sock