Corporate data and relationships analysis with RAG
This video shows how you can integrate searches and AI in an application and use it to analyze corporate relationships, events, facts and related information.
The application is made of several components and fetches data from several sources. It is meant for a serious analysis work and its current objective is to help an analyst work faster and robustly.
The source of information for the RAG can be files/text that you upload, as well as several vetted and official sources of corporate information, such as national corporate registries, international institutions APIs (the EU Commission and the US Senate), the SEC and other national regulators in other countries.
The system uses embeddings created with Voyage AI, stored in Pinecone and processed with Anthropic's Claude (Haiku).
The information is uploaded/pasted or fetched and is added to the vector store. Further information can be upserted to the vector store at any moment, depending on the analyst's needs. the information is stored in a namespace, which is tied to each user's unique id and mapped internally to the user email, so that
1) each user's data is isolated
2) even an inspection of the vector database would not allow to identify a user
I kept this video short to allow viewers to focus on some of the powerful features. The application is densely packed with tools and information sources - try it out!