Building Chatflows

Chatflows are the core of InnoSynth-Forjinn, allowing you to visually design and deploy AI-driven workflows. This guide will walk you through the process of building your own chatflows.

Accessing the Chatflow Builder

1. Navigate to Chatflows: From the left-hand sidebar, click on "Chatflows".
2. Create New: Click the "Add New" button to open a blank canvas. If you want to edit an existing chatflow, click on its name from the list.

The Chatflow Canvas

The chatflow builder consists of a central canvas where you drag, drop, and connect components to define your workflow.

• Left Panel (components library): This panel contains all available components, categorized for easy access (e.g., LLMs, Tools, Memory, Prompts).
• Central Canvas: This is where you assemble your workflow by dragging components and connecting them.
• Right Panel (Component Configuration): When you select a component on the canvas, its specific configuration options appear here.
• Bottom Right (Chat Interface): A mini-chat window for testing your chatflow in real-time.

Steps to Build a Chatflow

1. Drag and Drop Components:
   • Browse the components library in the left panel.
   • Drag the desired components onto the canvas. Common starting components include:
     • LLMs (Large Language Models): For AI reasoning and text generation.
     • Prompt Templates: To structure inputs for LLMs.
     • Tools: To enable your AI to interact with external services (e.g., search, databases).
     • Memory: To give your AI conversational context.
     • Document Loaders/Vector Stores: For RAG (Retrieval Augmented Generation) applications.

2. Connect Components:
   • Components have input and output ports (small circles) on their edges.
   • Drag a line from an output port of one component to an input port of another to create a connection.
   • Connections define the flow of data and execution in your workflow. For example, the output of a "Prompt Template" might connect to the input of an "LLM."

3. Configure Each Component:

   • Click on a component on the canvas to select it.
   • The right panel will display its configuration options.
   • LLM Components: Select the specific model (e.g., GPT-4, Claude), provide API keys (if not already set in Credentials), and adjust parameters like temperature or max tokens.
   • Prompt Templates: Write your prompt, using variables (e.g., {{input}}) that will be filled by other components or user input.
   • Tool Components: Configure the tool's specific parameters, such as search queries or database connection details.
   • Memory Components: Choose the type of memory (e.g., conversational buffer, chat history) and its settings.

4. Add Chat Output:

   • Always include a "Chat Output" component at the end of your main workflow chain. This component ensures that the final AI response is displayed in the chat interface.

Example: Simple Q&A Chatflow

1. Drag Prompt Template to canvas.
2. Drag OpenAI Chat Model to canvas.
3. Drag Chat Output to canvas.
4. Connect Prompt Template output to OpenAI Chat Model input.
5. Connect OpenAI Chat Model output to Chat Output input.
6. Configure Prompt Template: Answer the following question: {{question}}
7. Configure OpenAI Chat Model: Select your OpenAI credential and model.
8. Save your chatflow.

Now you can test this chatflow by typing a question into the chat interface.

For more details on specific components, refer to the components library.