Chatflows: Build Conversational AI Pipelines
Chatflows are visual pipelines that connect LLMs, memory, tools, and data sources into a deployable conversational AI — no code required.
A chatflow is Forjinn's fundamental building block for conversational AI. On the canvas, you drag, drop, and connect nodes — each representing a component like a language model, a memory store, a retrieval tool, or a data source — and wire them together into a working pipeline. When you deploy a chatflow, it becomes a live endpoint your applications or users can chat with.
Flow types
Forjinn supports several flow types under the chatflows umbrella. You select the type when you create a new flow, and it determines which nodes are available on the canvas.
Chatflow
A single-chain conversational pipeline. Best for chatbots, Q&A assistants, and RAG applications that follow a linear question-to-answer pattern.
Multi-agent
A pipeline that orchestrates multiple AI agents working together. Agents can delegate tasks to each other and synthesize results before responding.
Assistant
A managed assistant backed by OpenAI Assistants, Azure OpenAI, or a fully custom implementation. Ideal when you want built-in thread management and file handling.
Key properties
Every chatflow has a set of properties you can configure from the flow settings panel:
| Property | What it does |
|---|---|
| Name | A human-readable label shown in the dashboard and API responses. |
| Description | Optional context displayed to teammates and in the marketplace. |
| Category | A tag for organising flows in your workspace. |
| Deployed | Toggles whether the flow is live and accepting requests. |
| Public | Controls whether the flow can be accessed without authentication. |
Creating a chatflow
In the left sidebar, click Chatflows. Then click Add New in the top-right corner.
Enter a name, select the flow type (Chatflow, Multi-agent, or Assistant), and optionally add a description and category. Click Add.
The visual canvas opens. Use the node panel on the right to search for components — language models, memory, retrievers, tools — and drag them onto the canvas. Connect nodes by drawing edges between their input and output handles.
Click any node to open its settings panel. Fill in required fields such as model name, temperature, or credential references. Nodes with missing required fields are highlighted in red.
Click the chat bubble icon in the top-right toolbar to open the built-in chat panel. Send messages to test your flow without leaving the canvas.
Click Save to persist your changes. Toggle Deploy in the flow settings to make the flow live. A deployed flow is reachable via the Forjinn API.
Public vs. private flows
By default, a deployed chatflow requires an API key to access. You can make a flow public so that anyone with the URL can interact with it — no authentication needed.
Making a flow public exposes it to unauthenticated requests. Only use this setting for flows you intend to embed in public-facing products or share openly.
To change visibility, open the flow settings panel and toggle Make Public. Public flows also get an embeddable chat widget URL you can drop into any website.
Executions
Every time a user sends a message to a deployed flow, Forjinn creates an execution record. You can review execution history in the Executions page in the sidebar.
Each execution passes through the following states:
| State | Meaning |
|---|---|
INPROGRESS | The flow is actively processing the request. |
FINISHED | The flow completed successfully and returned a response. |
ERROR | The flow encountered an unhandled error during processing. |
TERMINATED | The execution was cancelled before it could finish. |
TIMEOUT | The flow did not complete within the allowed time limit. |
STOPPED | A user or administrator manually stopped the execution. |
Use the Executions page to debug unexpected responses. Each record shows the input, output, and which nodes were visited during that run.
Chatbot configuration
Once a flow is deployed, you can customise the appearance and behaviour of its chat widget from the Configuration tab in the flow settings. Options include the welcome message, suggested starter prompts, file upload permissions, and follow-up prompt suggestions.