Workflow V.S. Flow-Agent

GPTBots Workflow is designed to handle functional requests by executing a series of nodes sequentially to achieve specific functions. Flow-Agent, a specialized conversational workflow, operates on a dialogue-based model tailored for processing user interaction requests (e.g., scenarios requiring multi-turn conversations or dynamic conversational flows).

Compared to Workflow, Flow-Agent specializes in managing conversational logic within dialog-driven contexts. Each Flow is associated with a Conversation, dynamically accessing historical interaction data and maintaining complete session transcripts during live exchanges, which can be regarded as a workflow with memory.

Based on application scenario characteristics, you can build either a Flow-Agent or a workflow：

Flow-Agent Preferred Scenarios

For conversational interaction AI applications (e.g., smart customer service, AI assitants), we recommend Flow-Agent for its:
- Context Awareness: Integrates LLM for real-time conversation analysis and automatic dialogue memory maintenance.
- Interaction Adaptability: Native support for multi-turn dialogue design and natural language interfaces.
- Channel Compatibility: Seamless integration with social platforms like WhatsApp/Telegram.
Workflow Applicable Scenarios

For developing task-oriented processing AI applications (e.g., data batch processing, automation workflows), Workflow solution is advised.

Difference	Workflow	Flow-Agent
Essence	API	Muti-turn dialogue involving multiple Q&A interactions
Scenario	- Handle functional requests by executing a series of nodes sequentially to achieve specific functions. - Ideal for automated data processing scenarios, such as generating industry research reports, generating a poster, making picture books, etc.	- Specialized conversational workflow tailored for processing user interaction requests. - Interacts with users through conversation and completes complex operational logics. - Ideal for conversational applications that require complex logical processing in response to conversation requests, such as Chatbot, AI customer service, and virtual companions.
Learning cost	Relatively high: Workflow decouples variables from triggers, offering greater flexibity and power, but may require a moderately higher learning cost and cofiguration efforts.	Relatively low: Flow-Agent offers more intuitive operation optimized for multi-turn dialogue scenarios, though with slightly lower flexibility.
Operational logic	Schedules business through two core logics: variables and triggers. - Triggers are represented by the flow of connection lines. - Variables are defined and referenced by users within individual components.	Flow-Agent integrates variables and triggers into a unified mechanism, where the flow of connection lines not only initiates component execution but also facilitates data payload transmission.
Configuration	- Supports various components; - Canvas-based drag-and-drop operation	- Supports various components; - Canvas-based drag-and-drop operation; - Supports agents-related configurations (e.g., Memory, Welcoming Guide, Message Type, Tone, etc.)
Input	JSON format, supporting mainstream field types.	Natural language entered by users of the agent. Supports text, image, audio, and document inputs.
Output	JSON format, supporting mainstream field types.	Natural language generated by the agents. Provides text and audio contents.
Application	Designed for API-oriented single-turn request/response interactions: - Serves as an API for integration with any system. - Serves as a tool for integration with AI agents.	Optimized for human-facing multi-turn dialogue servce: - Serves as an API for integration with any system. - Supports integration into diverse UI contexts (Web chat, iframe, bubble components) and third-party platforms (WhatsApp, Telegram, WeChat Customer Service, etc.)