Uploading source data to knowledge models

After you upload your files to a knowledge model you create, the platform will process them and add the data to the model. The knowledge model will then use this data to respond to users' questions and prompts, and reference the original sources.

Constructor Model supports multimodality: it processes tables and images in uploaded documents (for example, PDFs with pictures) and uses them in the model's responses.

The supported file formats are:

• .csv
• .html
• .ipynb
• .json
• .md
• .pdf
• .ppt
• .pptx
• .py
• .txt

To upload source files with the data for your knowledge model:

1. In Constructor Platform, go to Knowledge models > My models.
2. In the list of your knowledge models, click the required model to open it.
3. Click the Data tab.
   
4. Click + Upload, select whether to upload a file or folder and drag-and-drop the necessary items. Model will process the uploaded data to create the corresponding knowledge model.
   While the platform processes the data, the files pass the statuses Pending, Uploading, Indexing, and In use. All the possible statuses are as follows:
   • Uploading — the file is being uploaded.
   • Pending — the file is pending for indexing.
   • Indexing — the file data is being split into chunks.
   • In use — the file is used by the knowledge model.
   • Not used — the file is disabled from reaching it by the knowledge model.
   • Error — an error occured when indexing the file. Try checking that the file format is correct.
   • Obsolete — the file is no longer relevant or cannot be used.
5. Wait until the file (or files) status changes to In use. This indicates that the model can now access and use the data from the uploaded files when responding in the chat.