batchwizard 0.4.0

BatchWizard: Manage OpenAI batch processing jobs with ease

BatchWizard

BatchWizard is a powerful CLI tool for managing OpenAI batch processing jobs with ease. It provides functionalities to upload files, create batch jobs, check their status, and download the results. The tool uses asynchronous processing to efficiently handle multiple jobs concurrently.

Table of Contents

• Installation
• Usage
• Configuration
• Commands
• Features
• Contributing
• License

Installation

You can install BatchWizard using pipx for an isolated environment or directly via pip.

Using pipx (recommended)

pipx install batchwizard

Using pip

pip install batchwizard

Ensure you have pipx or pip installed on your system. For pipx, you can follow the installation instructions here.

Usage

BatchWizard provides a command-line interface (CLI) for managing batch jobs. Here are some example commands:

Process Batch Jobs

To process input files or directories:

batchwizard process <input_paths>... [--output-directory OUTPUT_DIR] [--max-concurrent-jobs NUM] [--check-interval SECONDS]

You can provide multiple input paths, which can be individual JSONL files or directories containing JSONL files.

Example with Sample Input

Let's say you have a file named batchinput.jsonl with the following content:

{"custom_id": "request-1", "method": "POST", "url": "/v1/chat/completions", "body": {"model": "gpt-4o-mini", "messages": [{"role": "system", "content": "You are a helpful assistant."},{"role": "user", "content": "Hello world!"}],"max_tokens": 1000}}
{"custom_id": "request-2", "method": "POST", "url": "/v1/chat/completions", "body": {"model": "gpt-4o-mini", "messages": [{"role": "system", "content": "You are an unhelpful assistant."},{"role": "user", "content": "Hello world!"}],"max_tokens": 1000}}

To process this file using BatchWizard:

1. First, ensure your OpenAI API key is set:

   batchwizard configure --set-key YOUR_API_KEY
   

2. Then, run the process command:

   batchwizard process /path/to/batchinput.jsonl --output-directory /path/to/output
   

   This command will:
   • Upload the batchinput.jsonl file to OpenAI
   • Create a batch job
   • Monitor the job status
   • Download the results to the specified output directory when complete

You can also process multiple files or directories:

batchwizard process /path/to/file1.jsonl /path/to/directory_with_jsonl_files /path/to/file2.jsonl

Submit Without Blocking (fire and forget)

process blocks until every batch completes — which can take up to 24 hours. To submit and get your terminal back:

batchwizard submit /path/to/inputs/           # or: batchwizard process ... --submit-only

Submitted jobs are recorded in a local manifest (SQLite, in the BatchWizard config directory). Later — even after a reboot — reattach with:

batchwizard watch [--output-directory OUTPUT_DIR]

watch picks up every pending job from the manifest, polls until completion, and downloads the results.

Check Tracked Jobs

batchwizard status [--all]

Shows pending jobs from the local manifest (--all includes finished ones), with any failure reasons.

Failure Reasons and Error Files

When a batch fails, BatchWizard surfaces the provider's actual error (e.g. insufficient_funds, invalid_request with the offending line). When individual requests inside a batch fail, the per-request error file is downloaded alongside the results as <batch_id>_errors.jsonl.

Batch Endpoints

By default requests go to /v1/chat/completions. Use --endpoint on process/submit for other batch-capable endpoints such as /v1/responses or /v1/embeddings.

List Recent Jobs

To list recent batch jobs from the provider:

batchwizard list-jobs [--limit NUM]

Cancel a Job

To cancel a specific batch job:

batchwizard cancel <job_id>

Download Job Results

To download results for a completed batch job:

batchwizard download <job_id> [--output-directory OUTPUT_DIR]

This downloads the results file and, if any requests failed, the per-request error file.

Configuration

Setting up the OpenAI API Key

To set the OpenAI API key:

batchwizard configure --set-key YOUR_API_KEY

Show Current Configuration

To show the current configuration:

batchwizard configure --show

Reset Configuration

To reset the configuration to default values:

batchwizard configure --reset

Commands

BatchWizard supports the following commands:

• process: Submit batch jobs and wait for completion (add --submit-only to return immediately).
• submit: Submit batch jobs and exit; jobs are tracked in the local manifest.
• watch: Reattach to pending jobs, poll, and download results.
• status: Show jobs tracked in the local manifest.
• configure: Manage BatchWizard configuration.
• list-jobs: List recent batch jobs from the provider.
• cancel: Cancel a specific batch job.
• download: Download results (and error file) for a batch job.

For detailed information on each command, use the --help option:

batchwizard <command> --help

Features

• Flexible Input: Process individual JSONL files or entire directories containing JSONL files.
• Asynchronous Processing: Efficiently handle multiple batch jobs concurrently.
• Rich UI: Display progress and job status using a rich, interactive interface.
• Flexible Configuration: Easily manage API keys and other settings.
• Job Management: List, cancel, and download results for batch jobs.
• Error Handling: Robust error handling and informative error messages.

Contributing

We welcome contributions to BatchWizard! To contribute, follow these steps:

1. Fork the repository.
2. Create a new branch: git checkout -b feature/your-feature-name.
3. Make your changes and commit them: git commit -m 'Add some feature'.
4. Push to the branch: git push origin feature/your-feature-name.
5. Open a pull request.

Running Tests

To run tests, use pytest:

uv run pytest tests/

Ensure your code passes all tests and meets the coding standards before opening a pull request.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contact

For any questions or feedback, feel free to open an issue on the GitHub repository.