Yes — 0xDoktor works exclusively with your content
0xDoktor does not use any pre-trained knowledge about your product. It answers questions based solely on the documents you upload. This is the core design principle: your knowledge base is the only source of truth.
This means that before the widget can answer questions, you need to ingest your documentation. The more complete and up-to-date your knowledge base, the better the answers.
What formats can I ingest?
0xDoktor currently supports:
- Markdown (
.md) — ideal for documentation sites, wikis, READMEs - PDF — product manuals, whitepapers, policy documents, help guides
- Web pages (via URL) — any publicly accessible HTML page; 0xDoktor extracts the main content and ignores navigation, ads, and boilerplate
Support for additional formats (DOCX, HTML files, Confluence exports) is on the roadmap.
How does ingestion work?
When you upload a document, 0xDoktor:
- Parses the raw content and extracts clean text
- Chunks the text into overlapping segments of approximately 512 tokens each
- Embeds each chunk into a vector using a semantic embedding model
- Stores the chunks and their vectors in your private vector database
After ingestion, the content is immediately available for the widget to search. The whole process for a typical documentation page takes a few seconds.
How many documents can I ingest?
The number of documents depends on your plan. Starter plans support hundreds of pages; enterprise plans support tens of thousands. Contact the team if you have a large documentation corpus.
What happens when I update a document?
When you re-ingest an existing document (same URL or filename), 0xDoktor replaces the old chunks with the new ones. You do not need to manually delete the previous version. The updated content is available immediately after ingestion completes.
This makes it straightforward to keep your knowledge base in sync with your product documentation: update the source, re-ingest, and the widget immediately reflects the changes.
Can I organize content into multiple knowledge bases?
Yes. Within a single tenant, you can create multiple knowledge bases. This is useful for:
- Language separation: one KB per language for multilingual products
- Audience segmentation: a KB for end-users and a separate KB for developers
- Product areas: separate KBs for different parts of your product
You specify which KB to use on each page via the data-kb attribute in the embed script.
Can I preview what the widget will answer before publishing?
Yes. The admin dashboard includes a test interface where you can type questions and see exactly what the widget would answer, including which chunks it retrieved and the confidence score. This lets you validate your knowledge base coverage before deploying to users.
What if I want to prevent the widget from answering certain questions?
You can configure topics or keywords that the widget should decline to answer or escalate to a human agent. This is done through the tenant configuration in the admin dashboard. Common use cases include sensitive legal or medical topics where a human must be involved.