Last updated: January 20, 2025
<aside> 💬
Dokko is a state-of-the-art chatbot platform that harnesses the power of advanced AI and sophisticated natural language understanding to revolutionize how you manage, access, and share information. With intuitive chatbot interfaces accessible from any device, Dokko makes your team more connected and informed than ever before, enabling effortless communication and streamlined knowledge exchange.
</aside>
Many organizations grapple with the issue of siloed information, which impedes decision-making and reduces overall efficiency. Dokko addresses this head-on by developing a centralized, intuitive system that integrates data from various sources into a unified access point, using the simplest interface of all: natural, free-flowing text. With Dokko, there’s no need for users to learn new tools or adopt complex techniques. Instead, they can continue to use their existing tools and interact with the system using the familiar and comfortable format of everyday language, greatly enhancing productivity and facilitating smoother, more effective knowledge sharing.
If you want to try Dokko as quickly as possible, there are just a couple of actions you need to perform. You will be assigned a default AWS S3 document source when your account is created, but you will need to assign an LLM API key to it so your chatbots can communicate with LLM providers of your choice. To do this:
Go to the LLM Configuration / LLM API Keys section, choose Add new key, choose a friendly name, enter the key, and choose the provider name. Dokko supports multiple providers, including OpenAI, Groq, Anthropic and Google. Support for other providers and services will be added in the future. LLMs running on your own infrastructure are also supported in the Enterprise subscription model.
Instead of using API keys directly, which would be insecure, you will use their names - as assigned by you - ensuring that your keys and settings stay organized and protected no matter how many keys you work with.
Dokko uses several prompts for document preprocessing, chat interactions, and possibly different purposes. You can view and fine-tune these prompts in Settings / Prompt templates.
Prompts set up the default general tone and behavior of the chatbot. Each repository can have its own prompt template for each of the different tasks in Dokko workflow: preprocessing, chat, response analysis.
Settings / LLM Configuration sets screen defines which Large Language Models are in charge for pre-processing, chat and response analysis; it also allows you to set their parameters (temperature and max tokens), as well to associate prompt templates and API keys to them. Our default configuration uses OpenAI gpt-4o-mini for pre-processing and chat, while response analysis is optional and usually uses llama-8b-instant at Groq.
Go to Settings / Sources and click on the Edit icon next to your default document source. By default, new users get access to their AWS S3 storage bucket. Enter the LLM configuration set in the corresponding text field and click on Save.
Go to the Files section, right-click, and choose Create new folder. You can rename the folder by right-clicking on it, and you need to choose Mark as repository after that. Choose appropriate permissions for your new repository.
Upload several documents of your choice to it using the Upload files option from the toolbar.
When you are done, click on the Update knowledge base button in the upper right corner.
After several minutes, you will receive a notification that the knowledge base is updated, as indicated at the notification icon in the bottom left corner. You can see a list of all recent notifications by clicking on that button.
Go to the Playground section. Choose the default Document source type in the Widget script attributes section and choose permissions required to access your documents. You don’t have to configure anything else - your chatbot is ready to be tested in the right section of this screen!
When you are done with testing, you can view and download the JavaScript code to embed your chatbot on the web page of your choice.
Let’s move on to more advanced concepts and describe how to set up and use Dokko in different scenarios.
Dokko organizes your information in repositories that are stored in one or multiple document stores.
Let’s describe these core concepts in depth.