Skip to content

Maira Overview

Maira allows you to make your own GPT based solution, trained with your own data, customized according to your needs. Here are some basics to get you started. Find Maira related endpoints in our sandbox.

Lets get familiarized with a Few core concepts and abilities of Maira below

Documents and Datasets

Documents are single piece of information, such as Address of Gigalogy, What is Gigalogy personalization, etc. A Document must belong to a Dataset.

Dataset are collections of documents in a single file. For instance, single dataset may contain documents with information such as Address of Gigalogy, What is Gigalogy personalization, What is Maira and so on.

These documents and datasets are your information that is used to train Maira for your application.

We have discussed about Dataset and Document management in detail in the "Maira dataset management" Tutorial.

Profiles

Profile is a unique and powerful feature of Maira that lets you create multiple settings and modes for you application and switch between them rapidly as per your need.

With each request sent to Maira via the endpoints POST /v1/gpt/ask or POST /v1/gpt/ask/vision, we include a parameter called profile_id. This parameter points to a GPT profile that contains information guiding GPT on how to process the query and respond.

We have covered about Profiles in details in the "Maira Profiles" tutorial.

Let's look at some additional concepts below that will give you a head start in Maira.

Queries, Prompts, and Conversations

  • Queries: Queries are the "messages" or inputs that come from the end user of the application.
  • Prompts: Prompts are the complete set of instructions or text, including the "Query," that is sent to Maira to generate a response.
  • Conversations: In Maira, a Conversation consists of one Prompt and the corresponding response.

Conversation types, Sessions and Context

  • Conversation types: There are two types of conversation in Maira. Chat and Questions. Chat type conversation takes previous conversations into context to reply. Whereas Question type only considers the present query and not aware of the previous interaction.
  • Sessions: When Chat type conversations takes context into consideration, it checks for the conversations under same "Session".
  • Context: Context refers to the information that Maira uses to provide relevant and coherent responses in a conversations. It helps Maira maintain continuity and have a deep understand of the query. Not only Chat type conversations has context, each prompt can have context as we provide information in the Profile.

Conversation Summaries

Maira can summarize conversations as per your need. Maira will check the conversations of a defined time-period, and generate a full summary of all the conversations and present it in a way you prefer. In addition, you can edit this summaries and further train Maira with this data.

Reference documents and Evaluation

Maira generates answers mainly from two sources:

  1. From your data
  2. From LLMs' general knowledge

You can control how strictly Maira should balance these two sources of knowledge. With this in mind, whenever Maira generates an answer, like all LLM based applications, we want to know how certain we can be about the response and how much of it is based on the provided data.

To address this, whenever Maira generates a response, it shows "Reference documents" from which the answer was extracted. Additionally, Maira provides an "Evaluation" score, which indicates how similar the entire conversation is compared to the information in your document. This evaluation score, along with the reference documents, helps you build trust and confidence in the responses generated by Maira.

Next, in the following tutorials, we will dive into the details of how to integrate Maira into your system and build powerful applications.