Comment on page
Welcome to VectorShift
VectorShift is a no-code, drag-and-drop interface to build, design, prototype, and deploy generative AI workflows. Example use cases:
- Chatbots: Build a fully functional chatbot and embed it into your website in minutes. Chatbot is able to respond to user queries and answer questions based on a knowledge base (e.g., product documentation, support articles)
- Document Search: Summarize and answer questions about documents, videos, audio files, and websites instantly.
- Content Creation: Create marketing copy, personalized outbound emails, call summaries, proposals, and graphics at scale and in a predetermined format and style.
- Analyst: Replicate the logical thinking of an analyst to generate synthesized output.
Our site is structured into the following tabs:
- 1.Overview: the landing page. Jump either to our pipeline builder or marketplace.
- 2.Marketplace: a library of pre-built pipelines. Contribute to the marketplace by sharing your pipelines with other users! Within the marketplace, you can search for pipelines and “import” them into your workspace.
- 3.Pipelines: where you can find pipelines you have built. The “build” button opens up the pipeline builder, our drag-and-drop tool to build generative AI workflows. Additionally, you can utilize the pull-down menu on the top right to find your “imported pipelines” from the marketplace. Once selected, you can select a pipeline and click “add to my pipeline” to add to your pipeline workspace.
- 4.Files: you can find your stored files. You can upload files here and utilize them later in the pipeline builder.
- 5.Chatbots: turn pipelines you have built into Chatbots. Test the Chatbots in our interface before deploying to your end users via API.
At its core, a pipeline will have the following components:
- 1.Input / Output Node: Information (e.g., written text, video, audio, files) provided in the input node is used as an “input” to connecting nodes. The output node represents the result of the pipeline.
- 2.Large Language Model (LLM): an LLM is a trained deep-learning model that can generate output in a human-like fashion. You can connect other nodes (e.g., the input node) to the LLM to use as inputs into the model to instruct the model on what to do.
- 3.Data Loaders: nodes that allow you to import data into a pipeline. We currently offer the following data loaders: files, URL (data scraping), PDF, Wikipedia, YouTube video, and Arxiv.
- 4.VectorDB: vector databases are a type of database that stores unstructured data in vector embeddings (e.g., vector representation of the data) to facilitate retrieval (e.g., allow users to query to identify relevant data).