Knowledge Base

Upload documents to give your agents access to specific information through RAG (Retrieval-Augmented Generation). Instead of relying solely on the system prompt, your agents can search through uploaded documents in real time and ground their responses in your actual content: product guides, FAQs, policies, pricing sheets, and more.

How It Works

When you upload a document, Cloudax Connect processes it through a multi-stage pipeline:

Chunking: The document is split into smaller, semantically meaningful passages. This ensures that each piece of text is focused enough for accurate retrieval.
Embedding: Each chunk is converted into a vector embedding (a high-dimensional numerical representation that captures the meaning of the text).
Storage: The embeddings are stored in a vector database, indexed and ready for fast similarity searches.

During a live conversation, when the caller asks a question, the agent converts the question into an embedding, searches the vector database for the most relevant chunks, and incorporates that information into its response. This means the agent's answers are grounded in your actual documentation rather than general training data.

Accuracy through grounding

RAG significantly reduces hallucination. By retrieving real content from your documents, the agent can cite accurate figures, follow your specific processes, and stay on-brand.

Creating a Knowledge Base

Navigate to the Knowledge page

From the main navigation, click Knowledge. This page lists all your existing knowledge bases and their document counts.

Create a new knowledge base

Click Create Knowledge Base. Enter a descriptive name (e.g. "Product FAQs" or "Return Policy") and an optional description that explains what this collection contains. The description helps you and your team stay organised as you add more knowledge bases.

Upload documents

Once the knowledge base is created, click Upload or drag and drop files into the upload area. You can upload multiple files at once. Each file will begin processing automatically; see Document Processing below for details on the pipeline.

Supported File Types

Cloudax Connect accepts the following document formats:

Format	Extension	Notes
PDF	`.pdf`	Text-based PDFs work best. Scanned image PDFs may have reduced accuracy.
Plain Text	`.txt`	Simple and reliable. Great for FAQs and structured content.
Word (modern)	`.docx`	Full support for paragraphs, lists, and tables. Images are ignored.
Word (legacy)	`.doc`	Supported but `.docx` is recommended for better parsing reliability.
Markdown	`.md`	Ideal for developer documentation and structured guides.

Optimise your documents

For the best results, use clean, well-structured documents. Add clear headings, short paragraphs, and avoid embedding critical information inside images or complex tables. The clearer the text, the more accurately the agent can retrieve and reference it.

Document Processing

After uploading, each document moves through a processing pipeline. The current status is displayed next to every document in the UI:

Uploading: The file is being transferred to Cloudax Connect.
Processing: The document is being parsed and split into chunks.
Embedding: Chunks are being converted into vector embeddings.
Completed: The document is fully indexed and ready for retrieval.
Failed: Something went wrong during processing. Check the file format and try re-uploading.

Processing happens entirely in the background; you do not need to wait on the page. Most documents are processed within a few seconds; larger files (several hundred pages) may take a minute or two.

Failed documents

If a document fails to process, it will not be available for retrieval. Common causes include corrupted files, password-protected PDFs, and scanned-only PDFs with no extractable text. Try converting the file to a different format and re-uploading.

Connecting to Agents

A knowledge base on its own does not affect any agent; you need to explicitly link it. To connect a knowledge base to an agent:

Open the agent you want to configure in the agent editor.
Select the Knowledge tab in the left sidebar.
Choose one or more knowledge bases from the list. The agent will search all linked knowledge bases when answering questions.

You can link the same knowledge base to multiple agents, and a single agent can be connected to multiple knowledge bases. This makes it easy to share common content (like company policies) across your entire fleet of agents whilst also giving specific agents access to specialised information.

Keep knowledge bases focused

Rather than uploading everything into one large knowledge base, create smaller, topic-focused collections. This improves retrieval accuracy because the search space is more targeted. For example, separate "Product Catalogue" from "Return Policy" from "Shipping FAQs".

Organising Documents

As your knowledge base grows, keeping documents well-organised becomes important. Cloudax Connect provides several features to help:

Drag and drop: Move documents between knowledge bases by dragging them from one collection to another. The document will be re-indexed under its new knowledge base automatically.
Filtering: Filter documents by processing status, file size, file type, or upload date to quickly find what you need.
Bulk actions: Select multiple documents to delete or move them in a single operation.

Review and refresh regularly

Documents can become outdated. Set a reminder to review your knowledge bases periodically; remove obsolete content and upload updated versions. The agent will always use the latest indexed content, so keeping your documents current ensures accurate responses.