Adding Providers

Each AI platform is implemented as a provider — a self-contained plugin consisting of an Extractor, a Transformer, and a Provider registration.

Provider structure

Each provider lives in src/providers/{platform}/:

src/providers/myplatform/
├── extractor.ts    # Fetches raw data from the platform
├── transformer.ts  # Converts raw data to HAEVN.Chat format
├── model.ts        # TypeScript types for the raw data
└── provider.ts     # Registers the provider with the system

Step-by-step guide

1. Create the provider directory

   mkdir src/providers/myplatform

2. Define your raw data model (model.ts)

   Define TypeScript types matching the platform’s API or DOM structure.

3. Write the Extractor (extractor.ts)

   Implement the Extractor<TRaw> interface. For API-based platforms, this makes fetch calls using the existing browser session. For DOM-based platforms, this reads from the page DOM.

4. Write the Transformer (transformer.ts)

   Implement the Transformer<TRaw> interface. This converts your raw data into the canonical HAEVN.Chat format. See Data Model for the format spec.

5. Register the provider (provider.ts)

   Export a Provider object with your extractor, transformer, and platform metadata (name, icon URL, host patterns).

6. Register in init.ts

   Add your provider to the provider registry in src/background/init.ts.

7. Add host permissions

   Add the platform’s domain to manifest.json’s host_permissions array.

8. Add platform detection

   Update src/utils/platform.ts to recognize the platform’s URLs and return the correct provider ID.

Tip

The best way to understand the provider pattern is to read an existing simple provider. src/providers/qwen/ is a good starting point.

Testing your provider

pnpm test

Write unit tests for your transformer in tests/providers/myplatform/. Use real (anonymized) API response fixtures to validate the transformation.