CRM Tools Evaluation #38
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: CRM Tools Evaluation | |
| on: | |
| workflow_dispatch: | |
| inputs: | |
| toolsets: | |
| description: "Toolsets to run" | |
| required: true | |
| type: string | |
| trials: | |
| description: "Number of trials to run" | |
| required: true | |
| type: string | |
| seed: | |
| description: "Random seed (optional)" | |
| required: false | |
| type: string | |
| jobs: | |
| run_tasks: | |
| runs-on: ubuntu-latest | |
| steps: | |
| - name: Checkout repository | |
| uses: actions/checkout@v4 | |
| - name: Set up Python 3.12 | |
| uses: actions/setup-python@v5 | |
| with: | |
| python-version: "3.12" | |
| - name: Install dependencies | |
| run: | | |
| pip install -r requirements.txt | |
| - name: Run Tasks | |
| run: | | |
| # Parse input toolsets and convert to CLI arguments | |
| TOOLSET_ARG="" | |
| if [ -n "${{ inputs.toolsets }}" ]; then | |
| TOOLSET_ARG="--toolsets ${{ inputs.toolsets }}" | |
| fi | |
| # Trials amount | |
| TRIALS_ARG="" | |
| if [ -n "${{ inputs.trials }}" ]; then | |
| TRIALS_ARG="--trials ${{ inputs.trials }}" | |
| fi | |
| # Add seed if provided | |
| SEED_ARG="" | |
| if [ -n "${{ inputs.seed }}" ]; then | |
| SEED_ARG="--seed ${{ inputs.seed }}" | |
| fi | |
| # Create a directory for the results | |
| mkdir -p results | |
| # Run the benchmark | |
| python run.py ${TOOLSET_ARG} ${TRIALS_ARG} ${SEED_ARG} | |
| env: | |
| OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }} | |
| SUPERFACE_API_KEY: ${{ secrets.SUPERFACE_API_KEY }} | |
| COMPOSIO_API_KEY: ${{ secrets.COMPOSIO_API_KEY }} | |
| HUBSPOT_API_KEY: ${{ secrets.HUBSPOT_API_KEY }} | |
| - name: Process results | |
| run: | | |
| # Parse input toolsets and convert to CLI arguments | |
| TOOLSET_ARG="" | |
| if [ -n "${{ inputs.toolsets }}" ]; then | |
| TOOLSET_ARG="--toolsets ${{ inputs.toolsets }}" | |
| fi | |
| python process.py ${TOOLSET_ARG} | |
| - name: Upload results | |
| if: always() | |
| uses: actions/upload-artifact@v4 | |
| with: | |
| name: evaluation-results | |
| path: | | |
| results/*.jsonl | |
| processed/*.json | |
| processed/*.csv |