Authenticate every request with your Browse AI API key, sent as a Bearer token in the Authorization header. Generate your API key.
Still using the deprecated v1 API? See its documentation here.
This tag is used for endpoints that are used to check the status of Browse AI infrastructure
A robot can be trained do almost anything you do manually on the web. For example:
Robots are created either by using Prebuilt Robots or using Browse AI Recorder and its click-and-extract interface. Every robot has a few input parameters (like the webpage address) that you can adjust every time you run it.
Beta. This part of the API may change before it is generally available.
Create and retrain robots from code.
One robot per site. A robot is not tied to one page. To extract data from another page on the same site, run the robot you already have with a different originUrl input parameter. You do not need a new robot for each page.
A robot draft trains a robot on a webpage. Send an originUrl with overall instructions, a schema of the data to extract, or both. The schema is a JSON Schema: each property is a column, and its description holds the column's instructions. The draft starts a preview right away. Poll the draft until its status is done, missingData, or failed, then look at its latestPreview. Approve a done draft to create the robot, or set autoApprove when you create the draft to approve it automatically. There are no webhooks for drafts.
To retrain a robot, create a draft with the robot's robotId. Approving it writes a new version of the same robot.
Approving does not use credits for training. It also starts the robot's first task, which uses task credits like any other task. Training may use credits once your plan's AI budget is used up.
If your workspace already has 6 or more robots on the same site, creating another robot is refused unless you send acknowledgeSameDomainTraining: true. The error lists robots you can run instead.
Drafts stay in the list after they are approved.
Each robot is trained to perform a certain task. Every time you run that robot, it will perform that task and the details, including the extracted data, will be stored under that task.
If you set up a monitoring robot to monitor a webpage for changes daily, it will have to run a task every day or about 30 tasks per month for you.
Each robot on Browse AI can optionally have one or more monitors. Monitoring robots come with one monitor by default.
For example, if you set up a monitoring robot to monitor a category page on an e-commerce site for changes daily, you can set up additional monitors to monitor other category pages on the same site using the same robot.
Each monitor can have different input parameters and schedule.
After a robot finishes a task, its webhooks will be called.
You can run up to 50,000 tasks at once using a robot with different input parameters for each task.