Indexing APIs
Supported File Types
Raw document types
The upload endpoint supports several raw document types. Vectara extracts text
from these documents and sections them as best it can. This provides a
convenient way to index text, yet the caller has less control compared to when
providing the Document proto message themselves. The following raw document
types are supported:
- Commonmark / Markdown (
mdextension). - PDF/A (
pdf). - Open Office (
odt). - Microsoft Word (
doc,docx). - Microsoft Powerpoint (
ppt,pptx). - Text files (
txt). - HTML files (
.html). - LXML files (
.lxml). - RTF files (
.rtf). - ePUB files (
.epub). - Email files conforming to RFC 822.
Semi-structured documents
In addition, the upload endpoint supports
sending semi-structured documents through this endpoint that reflect a
Document proto message. Those can be sent in the following formats:
-
pb: Contains binary serializedDocumentproto message. -
pbtxt: ContainsDocumentproto message in proto text format. -
json: ContainsDocumentproto message in json text format.
For more details of how to format these types of files, read the formatting document