Home » AI Chatbots » Upload Files
How to Upload Files to Train Your AI Chatbot
You can train your AI chatbot by uploading text files, PDFs, or pasting content directly into the admin panel. The system extracts the text, breaks it into searchable chunks, and creates vector embeddings so the chatbot can find and use the information when answering customer questions. Upload as many documents as you need at 3 credits per chunk.
Supported File Types
- Text files (.txt): Plain text is the most reliable format. No formatting to strip, no conversion issues. If you can export your content as plain text, this is the cleanest option.
- PDF files (.pdf): The system extracts text content from PDFs. Works well with text-based PDFs (documents created from word processors). Scanned image PDFs may not extract cleanly since the content is an image, not selectable text.
- CSV files (.csv): Useful for structured data like product catalogs, FAQ lists, or contact databases. Each row becomes a separate piece of training content.
- Direct paste: Copy and paste text directly into the admin panel. Good for small amounts of content or quick additions without creating a file first.
Step-by-Step Upload Process
Step 1: Open the chatbot embeddings section.
Log into your admin panel, go to the AI Chatbot app, and select the chatbot you want to train. Navigate to the embeddings or knowledge base area where training data is managed.
Log into your admin panel, go to the AI Chatbot app, and select the chatbot you want to train. Navigate to the embeddings or knowledge base area where training data is managed.
Step 2: Choose your upload method.
Select file upload to attach a document from your computer, or use the text paste option for content you want to enter directly. You can also point to a URL if your content is published online.
Select file upload to attach a document from your computer, or use the text paste option for content you want to enter directly. You can also point to a URL if your content is published online.
Step 3: Upload or paste your content.
For file uploads, select the file and submit. For text paste, enter or paste your content into the text area. There is no strict size limit, but very large documents (hundreds of pages) should be split into logical sections for better chunk quality.
For file uploads, select the file and submit. For text paste, enter or paste your content into the text area. There is no strict size limit, but very large documents (hundreds of pages) should be split into logical sections for better chunk quality.
Step 4: Tag the content.
Assign a tag or label to identify this batch of training data. Tags help you organize embeddings, for example "product-manual" or "faq" or "policies." Later, if you need to update or remove specific content, the tag makes it easy to find.
Assign a tag or label to identify this batch of training data. Tags help you organize embeddings, for example "product-manual" or "faq" or "policies." Later, if you need to update or remove specific content, the tag makes it easy to find.
Step 5: Process and embed.
The system breaks your content into chunks of roughly 200 to 500 words and creates a vector embedding for each chunk. This costs 3 credits per chunk. A 10-page document typically produces 15 to 30 chunks, costing 45 to 90 credits ($0.045 to $0.09). Processing usually completes in under a minute.
The system breaks your content into chunks of roughly 200 to 500 words and creates a vector embedding for each chunk. This costs 3 credits per chunk. A 10-page document typically produces 15 to 30 chunks, costing 45 to 90 credits ($0.045 to $0.09). Processing usually completes in under a minute.
Step 6: Verify with test questions.
After uploading, open the chatbot and ask questions that the uploaded content should answer. Verify the chatbot gives accurate responses. If it misses something, the relevant content may need to be restructured for better chunking. See How to Chunk Documents for Better AI Understanding.
After uploading, open the chatbot and ask questions that the uploaded content should answer. Verify the chatbot gives accurate responses. If it misses something, the relevant content may need to be restructured for better chunking. See How to Chunk Documents for Better AI Understanding.
Preparing Documents for Upload
The quality of your chatbot's answers depends heavily on the quality of the documents you upload. Here are practical guidelines:
- Write clearly and specifically. Content that says "Our premium plan includes X, Y, and Z for $49/month" is more useful than "We offer various plans at competitive prices."
- Use descriptive headings. Headings help the chunking process break content at logical boundaries. "Return Policy" is a better section header than "Section 4."
- Include the questions, not just the answers. A document that says "Q: What is your return policy? A: We accept returns within 30 days..." matches question embeddings better than one that just states the policy.
- Remove irrelevant boilerplate. Legal disclaimers, copyright notices, and navigation text from web copies add noise without useful information.
- Keep information current. Upload updated documents whenever your pricing, policies, or product details change. The chatbot uses whatever content is in its knowledge base.
Combining methods: Most businesses get the best results from a combination of approaches. Crawl your website for broad coverage of published content, then upload specific internal documents (employee handbooks, detailed product specs, sales scripts) that are not published online. Use direct paste for small, frequently updated information like current promotions or temporary policy changes.
Managing Uploaded Content
Over time, you will accumulate many sets of embeddings. Use tags to keep them organized. When information changes, you can delete the old embeddings for that tag and upload the updated document. There is no need to rebuild the entire knowledge base, just replace the specific content that changed. See How to Delete or Update Specific Training Data for the process.
Upload your documents and have a trained chatbot ready in minutes. No coding needed.
Get Started Free