Gemini 2.5 Converse Agents: This FULLY FREE Way to create CONVERSATIONAL AI Agents is AMAZING!

Conversational AI Workflows in Vector Shift: A Game Changer

Key Concepts:

• Conversational AI Workflows
• Talk Nodes (Message, Image, Cards, Carousels)
• Listen Nodes (Capture Response, Buttons)
• Lead Collection Chatbot
• Google Sheets Integration
• Retrieval Augmented Generation (RAG)
• Knowledge Base Node
• LLM Node (OpenAI)
• Looping
• Cyclic Input
• Chatbot Deployment (Embed, iFrame, Chat GPT Style Interface)
• Message Delay

1. Introduction to Conversational AI Workflows

Vector Shift has introduced Conversational AI Workflows, a new AI workflow mode designed for building custom and interactive chatbots. This feature offers a significant upgrade in control compared to standard pipelines.

2. Distinguishing Conversational AI from Standard Pipelines

Standard pipelines typically follow a direct input-to-output process. Conversational AI workflows, however, allow users to define specific actions at each stage of the conversation, enabling step-by-step user guidance.

3. No-Code Builder and Conversational Nodes

The no-code builder includes a new "conversational" tab with "talk" and "listen" nodes.

• Talk Nodes: Used for the bot to send information (e.g., text messages, images, cards, carousels). The "message node" is used for text.
• Listen Nodes: Used for the bot to receive input (e.g., capture response node for user input, buttons).

4. Building a Lead Collection Chatbot: A Step-by-Step Example

A simple lead collection chatbot is built to demonstrate the functionality:

1. Start Flag: Initiates the chat.
2. Message Node 1: Asks "What is your name?"
3. Capture Response Node 1: Captures the user's name.
4. Message Node 2: Asks "What is your email?"
5. Capture Response Node 2: Captures the user's email.

The chatbot waits for user input at each capture step, providing fine-grained control.

5. Integrating with Google Sheets

The collected lead information (name and email) is saved to a Google Sheet:

1. Google Sheets Node: Configured to add a new row.
2. Column Mapping: The output of the first capture node (e.g., "capture zero") is mapped to the "name" column in Google Sheets, and the second capture node (e.g., "capture one") is mapped to the "email" column.

After both name and email are captured, a new row is automatically added to the Google Sheet.

6. Transitioning to an AI-Powered Q&A

After lead collection, the bot transitions into an AI assistant:

1. Message Node 3: Asks "Thanks, how can I help you today?"
2. Capture Response Node 3: Captures the user's question.
3. Knowledge Base Node: Points to a website (e.g., the Vector Shift URL) to scrape for context.
4. LLM Node (OpenAI):
   • Prompt: Instructs the LLM to answer the question based on the provided context.
   • Input: The user's question from the capture node and the context from the knowledge base reader output.
5. Message Node 4: Sends the LLM's response back to the user.

This setup allows for a seamless transition from deterministic steps (lead collection) to AI-powered Q&A.

7. Looping and Cyclic Input

The conversation can be looped back to the "How can I help you today?" message node, allowing the user to ask multiple questions. Vector Shift highlights the loop visually.

Cyclic Input: A feature that allows the bot to vary its prompts on subsequent loops (e.g., changing "How can I help?" to "Do you have any additional questions?"), making the conversation flow more naturally.

8. Chatbot Deployment

The deployment process is similar to other chatbots in Vector Shift:

1. Deploy Changes: Save the workflow.
2. Export: Select "chatbot."
3. Name: Give the chatbot a name.
4. Message Delay: Set the time delay between each message from the bot (e.g., 0.5 seconds).
5. Export: The chatbot can be embedded into a website using a script or iFrame, or tested using the provided Chat GPT-style interface.

9. Conclusion

Conversational AI Workflows in Vector Shift offer a significant upgrade for building guided, interactive, and intelligent chatbots. The ability to define each stage of the conversation, collect information, integrate with tools like Google Sheets, and implement looping capabilities makes this a powerful tool for lead generation, customer support, and other applications requiring precise control over the conversation flow.