Knowledge Base (RAG)

The Knowledge Base module in DeepExtension allows users to convert documents into vector representations for use in retrieval-augmented generation (RAG) and related knowledge-aware AI workflows. This is especially useful for enterprise applications that require grounding LLM outputs in internal documentation.

---

Create a Knowledge Base

To set up a new Knowledge Base:

1. Click "Create a Knowledge Base" on the Embedding page.
2. Fill in the Knowledge Base name and a short description.
3. Select an embedding model. If no models appear, register a usable embedding model via Third-party Models.
4. Click "Save" — your Knowledge Base will be created immediately.

---

Add Embedded Documents

To add documents to an existing Knowledge Base:

1. Click "Edit" on the Knowledge Base you want to modify.
2. You’ll see a list of already embedded documents (initially empty).
3. Click "Upload Document", then select a file in .pdf, .docx, .txt, .xlsx, or .csv format from your local directory.
4. Choose whether to enable “Parsing During Creation” (this will trigger immediate embedding using the selected model) or skip for later parsing.
5. Click "Execute" to begin uploading. The process runs in background mode.
6. If an error occurs, click "Log" to inspect detailed error messages.

---

Document Embedding Behavior

While document upload and embedding are supported, full Knowledge Base search and RAG-style retrieval are still under development and listed on our roadmap for future releases.

---

DeepExtension — Make your documents searchable, contextual, and ready for generation