The platform provides 4 mainstream integration methods, which developers can choose flexibly according to the technology stack:
- Native Knowledge Base Docking::
- Select the preconfigured platform (OpenAI Storage/Dify Datasets) in the Integrate screen.
- Enter the API Key of the target platform (prior application required)
- Setting up field mapping relationships (auto-matching or manual adjustment)
- API call::
- Get the unique Endpoint provided by the platform
- Get latest data via POST request (supports filtering by time/keyword)
- Return to Example:
{"status": 200, "data": [{"title": "...", "embedding": [...]}]}
- Webhook Push: Configure the notification address to automatically trigger data synchronization when a timed task is completed
- File export and re-import: Download JSON-LD format files (containing semantic tags), loaded through frameworks such as LangChain
For customization needs, it is recommended:
- Use the "customKeys" parameter to define business-specific fields (e.g. customer industry)
- Dynamic structuring of data in conjunction with the "natural language extraction" function ("Next time, please include the author's title")
- Directly reference the dataset as a context source when creating a Web App in platforms like Dify
All integration methods preserve data lineage, traceability of original documents and modification records, and meet corporate audit requirements.
This answer comes from the articleSupametas.AI: Extracting Unstructured Data into LLM Highly Available DataThe
May not be reproduced without permission:AI productivity tools " How can the data processed by Supametas.AI be integrated into existing AI systems?
English 
简体中文 
日本語 
Deutsch 
Português do Brasil