Doubleword

What This Skill Does

The Doubleword skill for OpenClaw provides a robust interface for managing asynchronous batch inference workloads. It allows developers and users to offload high-volume AI tasks to the Doubleword API (api.doubleword.ai), enabling cost-effective, scalable processing. By utilizing the batch endpoint, users can queue large numbers of AI requests—such as tool calls, structured output generation, or bulk analysis—without waiting for real-time responses. This skill handles the end-to-end lifecycle, including uploading JSONL request files, monitoring job progress, and retrieving final results. It is specifically optimized for models like the Qwen3-VL-235B, significantly reducing costs for non-latency-sensitive workloads by up to 83%.

Installation

To integrate this skill into your environment, run the following command in your terminal:

clawhub install openclaw/skills/skills/pjb157/doubleword

Ensure you have an account at https://app.doubleword.ai/ and have generated an API key within the dashboard settings. You must also have sufficient account credits to trigger inference jobs.

Use Cases

• Mass Data Processing: Analyzing thousands of documents, images, or code snippets that can be processed in a background queue.
• Cost Optimization: Migrating non-time-critical inference tasks from expensive real-time endpoints to cheaper 24-hour SLA batch windows.
• Structured Data Extraction: Generating large-scale JSON outputs from unstructured text for database ingestion.
• Automated Testing: Running massive suites of prompt evaluations or synthetic data generation tasks.

Example Prompts

1. "Submit my 'batch_data.jsonl' file to Doubleword for processing using the Qwen3-VL-235B model with a 24-hour completion window."
2. "Check the status of my active batch job with ID 'batch_12345' and notify me when it is finished."
3. "Download the results from my latest batch job and save them to a file named 'processed_results.jsonl'."

Tips & Limitations

• Pricing: Always check the Doubleword Console to preview costs before submitting large batches; different SLAs (realtime vs. 24h) offer different cost savings.
• Format: Ensure your input files are in strictly formatted JSONL; any invalid lines will cause the batch upload to fail.
• Throughput: While batches are highly efficient, they are not intended for tasks requiring sub-second response times; plan your workflow to handle the async delay.