+91 253 6630710
781.258.1274
+44 (0) 23 9204 6323GenAI Accelerators
Talk to us now!
Solve challenging business problems using our accelerators & start your analytics journey with a strong and confident foundation.
VartikGPT Secure Extensible Enterprise Chat
One of the clever technologies of our chat-based accelerator is the Data Ingestion pipeline that easily works with unstructured and semi-structured data. On this front, the system uses Multi-chunking mechanisms like Recursive and Semantic chunking to optimize the storage of information in Vector database. Converting text to embeddings using an open-source embedding framework, it offers support for multiple LLMs including open-source ones. It also has caching for the messages; chat history management; and routing for cost optimization. Also, It can be extended to domains like HR, Finance, IT through configurable Prompts and separate Vector database indexes.
Generic Data Ingestion Framework
Our enterprise-ready, secure, and scalable data ingestion framework efficiently handles data from private knowledge bases. It supports various chunking strategies and integrates with Vector databases such as Pinecone, Qdrant, and Azure AI Search. The framework accommodates multiple LLMs, including open-source options, for text embedding. It can ingest a wide range of data formats, including PDFs, DOCX, PPT, emails, CSV, MD, and more.
Dynamic Document Generation
Metadata and rule-based document construction block generator that employs multi-level Dynamic Generation Map to streamline document chunk dependencies. It has configurable Prompt definitions in YAML files in datastore to define the semantic of the user’s input. It uses Aress Generic Data Ingestion Framework to download historical data for context relevance in the document generation. This accelerator can be reused in new cases with a little modification Therefore, this accelerator can be used as reusable framework for new use cases.