Interacting with your private PDF's, Docx, PPTX, and more
How to chat with your private documents on a self-hosted ChatGPT alternative?
Introducing MirageGPT, a cutting-edge private AI built on the advanced Llama 2 model by Meta, which offers unprecedented capabilities while safeguarding your privacy. With its self-hosting feature and integrated knowledge bases, you can interact seamlessly with your private documents. From annual reports to data processing agreements. Let's delve into how you can use MirageGPT to converse with your own PDFs, illustrating this with Apple's 2023 annual report.
How MirageGPT works with your private documents
Our MirageGPT knowledge base integration lets you chat with your complex documents, enabling you to ask direct questions and understand intricate details in your own words. Here's how it works:
- Step 1: Creating a new knowledge base. The first step is to create a new knowledge base by providing a unique name and description. In this example we are going to create a knowledge base named "Annual Reports" to answer questions about Apple's Annual Report.
-
Step 2: Uploading documents. Once the knowledge base is set up, you can upload various types of files, including PDFs, Docx, PPTX, HTML, TXT, CSV, EML, and more.
We are going to upload a PDF which is named "Apple FY23 Q2 Consolidated Financial Statements.pdf". Let's click on the upload pane or drop or file and wait for the upload to be finished. - Step 3: Processing and Indexing. After the documents are uploaded, MirageGPT processes them and indexes the information. This step may take a few minutes depending on the size and complexity of the documents. Our PDF analyzer also incorporates table data and employs OCR technology to identify text within images. Therefore, even scanned PDFs or PDFs with complex tables can be uploaded and utilized for interaction with MirageGPT.
-
Step 4: Interaction with the Knowledge Base. We are on the brink of interacting with our new private knowledge base with our annual report PDF in it. Our AI tool, MirageGPT, allows us to ask questions, seek clarification, and gain insights, all in a simple and natural language.
However, before we dive in, we must first select our new knowledge base for MirageGPT to draw answers from. Begin by clicking on the knowledge base icon. - A dropdown menu is now available for selecting the knowledge base that MirageGPT will utilize to retrieve information for answering your questions or executing your tasks.
- We choose the recently created "Annual Reports" knowledge base, which contains the Apple Annual Report PDF that we uploaded.
-
After selecting the appropriate knowledge base, we are now able to pose questions related to the documents uploaded within it. However, its functionality is not limited to merely asking questions. For instance, you could request it to compose an email based on information from your document.
But for now we are going to ask it "What where the net sales in Europe in 2023 looking at three months ended". - We can look at the original PDF that we uploaded to check if it gave the correct answer.
- As you can see in the PDF, the correct answer for the question: "What where the net sales in Europe in 2023 looking at three months ended" is indeed $23,945 million.
Other Interaction with Apple's Annual Report
Here's how other interaction look:
Question: What where the total cost of sales in 2022, three months ended?
MirageGPT: The total cost of sales for Apple Inc. for the three months ended March 26, 2022 was $54,719.
Question: What where they April 1, 2023?
MirageGPT: Based on the information provided in the financial statements for the three months ended April 1, 2023, the cost of sales were $52,860.
Expanding Possibilities with MirageGPT
Having the ability to pose intricate queries and the capacity to further clarify them is incredibly powerful, and opens up numerous possibilities. You can now effortlessly inquire about legal or financial documents without needing to understand professional jargon. You can rapidly search through training materials or manuals, quickly locating information on invoices.
The Concept of a Knowledge Base
This leads us to the concept of a knowledge base. You might be wondering why you can upload multiple documents into a single knowledge base. The reason is that this allows the ability to search across several documents. Suppose you've uploaded all invoices from the past year. You can then ask, "Which invoice mentions web design costs?" MirageGPT can search your knowledge base for the web design costs and report that it found this information on invoice #38392. It can also indicate that the cost was 10,000 euros, which covered 80 hours of work at a rate of 125 euros per hour. You can then ask for the total cost of invoice #38392. This opens up new possibilities to search and use your data in questions and tasks you delegate to MirageGPT.
Executing Tasks with MirageGPT
Yes, it's also possible to execute tasks based on your data. If you instruct MirageGPT to "Compose a formal email to the sender of quote #1002 expressing our concerns about their high costs", it can retrieve the information from that quote, analyze it, and execute your command using this data.
Conclusion
As demonstrated, MirageGPT provides a new way to interact with your documents, making it easier to access the exact information you need. With the integration of private knowledge bases, you can delve into your documents' details with unprecedented depth and clarity. Start a conversation with your documents today with MirageGPT!