Aerix provides a Web Data Extraction API that turns any web page into a structured, validated JSON endpoint. Users supply a URL and a JSON Schema, and the service uses a headless browser and LLM to extract data, automatically healing broken endpoints and caching results. Features include scheduled runs, smart webhooks, batch processing of up to 50 endpoints, and self‑correcting extraction without writing scraping code.
Funding
Funding not disclosed
Founders
Product
Problem
Extracting data from websites typically requires writing custom scraping scripts, handling dynamic content, and continuously maintaining code as page structures change, which is time‑consuming and error‑prone.
Solution
Aerix offers a Web Data Extraction API that converts any web page into a structured JSON endpoint without the need for custom scraper development. Users provide a target URL and a JSON Schema describing the desired output; the service uses a headless browser and a two‑pass LLM process to retrieve, parse, and format the data, then validates it against the schema. Results are cached and can be accessed instantly, while built‑in self‑healing mechanisms detect failures, regenerate extraction prompts, and automatically adjust to page layout changes. The platform also supports scheduled runs, conditional smart webhooks, and batch processing of up to 50 endpoints per request, enabling reliable, automated data pipelines.
Target Audience
Primary users are developers, data engineers, and product teams that need reliable, low‑maintenance web data feeds for use cases such as sports scores, price monitoring, and news aggregation.
Features
- Headless browser fetches pages, handling JavaScript and dynamic content
- Two‑pass LLM extraction: raw data extraction followed by schema‑compliant formatting
- JSON Schema validation with automatic retry on validation errors
- Self‑healing endpoints that regenerate extraction logic after consecutive failures
- Smart webhooks configurable for change detection, failure alerts, or custom data conditions
- Redis caching with configurable TTL for instant repeat responses (X‑Cache: HIT)
- Batch fan‑out allowing up to 50 concurrent endpoint executions while preserving input order
- Endpoint wizard that generates prompts, schemas, and schedules from plain‑English descriptions