April 11- VectorShift Product Update

Use case spotlight: Deploy a Search Interface on your AI workflows

With VectorShift, you can place a search interface on an AI workflow and deploy it directly into your product, website, or app! The search interface will show citations and the specific chunk of information leveraged to generate the answer!

Product Updates

The VectorShift team has been hard at work over the last few weeks - this is probably our largest product update yet!

1. Version Control: Revert back to a previous deployed version of a pipeline. This is helpful if you made changes to a pipeline that you do not want. Find this under the “Edit Information” pane within the no-code builder.

2. Auto-Saving: pipelines will auto-save - never lose your progress again!

3. Multiplayer Mode: work on a pipeline together with teammates and see changes updated live (e.g., new nodes, new connections). Automatically enabled for all accounts.

4. Offline Mode: Continue to edit pipelines if you lose internet connection. When your computer gets access to internet again, the pipeline will save.

5. Search Interface: place a search interface on a pipeline or knowledge base. Do so within the pipeline builder (Deploy >> Search) or on the new “Search” tab.

6. Enhanced CSV reader: we made significant accuracy improvements in the ability to query over tabular data. Utilize the CSV query node under the “Data Loader “ tab.New

7. Data Sources / Integrations

   1. Dropbox: live-sync with your dropbox and leverage the data in your AI workflows.

   2. One Drive: live-sync with your OneDrive and leverage the data in your AI workflows.

   3. S3: live-sync with your S3 storage and leverage the data in your AI workflows.

   4. Confluence: live-sync with your Confluence knowledge base and leverage the data in your AI workflows.

   5. Slack (private channels): in addition to public channels, live-sync with data from Slack private channels and leverage the data in your AI workflows.

   6. Zapier / Make nodes: have a pipeline run a Zapier / Make.com workflow.

   7. Zendesk: search tickets, create new tickets, and trigger a pipeline with a new Zendesk ticket.

   8. Notion (create new block): add text to an existing Notion page.

   9. Postgres: connect and query your Postgres database.

   10. Pinecone: connect and query your Pinecone database.

8. Re-scrape URLs: when loading URL data into a knowledge base, define how often you want the data to be re-scraped and re-embedded. This will ensure the data is fresh.

9. Citations: Enable LLM to site sources in the output (URLs and files). Access by clicking on the gear on an LLM.

10. Significant improvements to chatbots

    1. Overall UI improvements

    2. Edit display details

    3. Add assistant / launcher / avatar image

    4. Enable AI generated follow up questions

    5. Copy and download responses

11. Anthropic Claude 3: New heavy-weight LLM model! Find it in the Anthropic node under the LLMs tab.We are cooking up a lot more improvements over the next few weeks - stay tuned!Video Feature: An overview of building Generative AI apps in No-code

LLM Guide Highlight:

Maximizing LLM performance with Effective Chunking Strategies for Vector Embeddings: this article outlines the goals of chunking and available chunking strategies for LLM vector databases.