Create a Job
Jobs are configurable workflows that run on your files using a pipeline you choose. Currently, Haystack Enterprise Platform supports a job for batch question answering.
About This Task
A job runs once on all files in your workspace. You can see the results on the job details page and you can easily share them as a formatted HTML page.
Jobs can only use pipelines that return answers, such as RAG or extractive question answering pipelines, or chat.
Jobs that run for more than 12 hours are automatically stopped.
Prerequisites
Create a Job
- In the navigation, go to Jobs > New Job.
- Choose the job type, and click Next.
Currently, only batch question answering is supported. Choose how the queries run on your files:- Run once across all files: Each query searches for an answer across all files. You get one answer per query.
- Repeat queries per file: The queries are applied to each file individually. You get as many answers per query as there are files.
- Upload your query set or choose a previously uploaded one, and move on to the next step. You can also enter queries directly without uploading a file.
- Choose the pipeline you want to use to process the queries. If the pipeline is not indexed, starting the job triggers indexing. Jobs can only use pipelines that return answers, like RAG, chat, or extractive question answering.
- Give your job a name.
- Choose one of the following:
- To start the job now, click Start Job. The job starts running. It continues running even if you leave the page.
- To save the job as a draft, click Save as Draft. You can run it later from the Jobs page.
If a job is interrupted — for example, because a worker stopped — you can restart it without losing progress. The job resumes from the last successfully completed query.
What To Do Next
Click the job name to view its details. You can check the details of each query, such as the answer, the prompts used, the parameters, and more.
You can also generate a link to the results page and share it with others.
Was this page helpful?