WeKnora

What it is

WeKnora is a tool that processes raw documents and provides three main outputs: a question-answering RAG, a reasoning agent, and a self-updating wiki. It handles parsing, chunking, embedding, and retrieval of documents. WeKnora is built using Go and comes with a single-binary deployment option.

The tool is designed to handle the complexities of document processing, including incremental reindexing and document versioning. This opinionated approach aims to simplify the deployment and management of the tool.

Where it fits in a deployment stack

RAG tools like WeKnora are typically used to improve the accuracy and efficiency of question-answering systems. They sit between the data storage and the application layers, providing a critical link between the raw data and the end-user interface.

In a production system, a RAG tool would normally be surrounded by other components such as data ingestion pipelines, data storage solutions, and application servers. The RAG tool would receive input from the data ingestion pipeline, process it, and then provide output to the application server.

The goal of a RAG tool is to reduce the complexity and latency associated with querying large datasets, making it an essential component in many modern applications.

How to evaluate it

To evaluate WeKnora, you need to consider its performance, scalability, and ease of use. This involves testing the tool with your specific dataset and use case, and verifying its output against expected results.

• Feed a sample dataset of 10,000 documents into WeKnora and measure the time it takes to process them.
• Inject failures into the system, such as network errors or document parsing errors, and verify that WeKnora recovers correctly.
• Inspect the output of the question-answering RAG and verify that it provides accurate answers to a set of test questions.
• Measure the p95 latency of the retrieval step under 50 concurrent sessions.
• Verify that the self-updating wiki correctly updates its content when the underlying document corpus changes.
• Test the tool's ability to handle incremental reindexing and document versioning.

What to verify before you commit

Before committing to WeKnora, you need to verify several key aspects of the tool, including its licensing, data handling, and operational costs. You should also consider the potential risks of vendor lock-in and the tool's behavior in case of failures.

• Check the licence terms to ensure they align with your organization's requirements.
• Verify that WeKnora handles sensitive data in accordance with your organization's data handling policies.
• Estimate the operational costs of running WeKnora, including hardware, maintenance, and support costs.
• Check for any potential lock-in risks, such as proprietary data formats or limited export options.
• Verify that WeKnora provides adequate support and documentation for troubleshooting and maintenance.
• Check the tool's behavior in case of failures, such as data corruption or system crashes.