Chatflow

A Chatflow is LanSphere’s multi-turn conversational app: users interact with the app continuously in dialogue, and the app carries context memory, answering turn by turn along the flow you orchestrated.

This article covers Chatflow use cases, how it differs from Workflow, the five conversational features, the “Reply” node, debugging, and publishing channels. It is for developers building conversational apps.

Use Cases

Chatflows fit open-ended Q&A that needs multi-turn interaction and follow-ups. Typical scenarios:

• Smart customer service: Answer common questions from employees or the public, with multi-turn follow-ups and added detail.
• Knowledge Q&A: Combined with a Knowledge Base, answer questions about policies, products, and technical docs with sources.
• Business inquiry assistant: Provide process consultation and handling guidance around a business area (e.g., administration, HR, IT services).

How It Differs from Workflow

Tip: When choosing an app type, judge the interaction form first — “fill in a form once, get one result” → Workflow; “back-and-forth over multiple turns” → Chatflow.

Multi-Turn Conversation Memory

A Chatflow carries context across turns: users do not have to repeat background info every turn — the app understands the current question in light of previous questions and answers. For example, if a user first asks “How many days of annual leave do we get?” and then asks “What about sick leave?”, the app understands the second question is still about leave policies.

This makes Chatflow a natural fit for open-ended Q&A: what you orchestrate is “how one turn is handled”; continuity across turns comes from conversation memory.

Creating a Chatflow

1. Click “New Project” at the top of the left menu.
2. Choose the app type “Chatflow” in the pop-up.
3. Fill in the app name (required, 1–30 characters; Chinese characters, upper/lowercase letters, digits, and hyphens only). Description is optional (up to 300 characters); custom avatar upload supported.
4. Confirm to enter the “App Development” page and orchestrate the flow.

Canvas operations (click-to-add nodes, list search, zoom, undo/redo, change history), connections and variable references, Environment Variables, Variable Inspector, and the Checklist all work the same as in Workflow — see “Workflow.”

Conversational Features

Chatflow offers five conversational features to refine the end-user experience, enabled and configured as needed on the “App Development” page:

Opening Message

A welcome message with preset questions shown automatically when a user first enters the conversation. The Opening Message explains what the app can do; preset questions guide users into their first turn.

Suggested Questions

Automatically recommends likely next questions based on the current answer, helping users keep digging instead of wondering what to ask next.

Citation & Attribution

Marks the Knowledge Base sources cited in an answer so users can verify them. Works together with the Knowledge Base node — ideal for knowledge Q&A where answer credibility matters.

Sensitive Word Review

Reviews user input and app output for sensitive content — for government and enterprise scenarios with content-compliance requirements.

File Upload

Lets users upload files into the conversation as part of the Q&A — e.g., upload a document and ask questions about it.

The “Reply” Node

The “Reply” node is how a Chatflow outputs messages to the user, wherever it sits in the flow:

• It can output multiple times mid-flow. For example, send a “Looking it up…” message first, then output the final result when the query completes.
• Each time a “Reply” node executes, the user receives one message.

Note the difference from the Workflow “Output” node: a Workflow has no concept of conversation, and its “Output” node returns one result when the flow ends; a Chatflow serves multi-turn dialogue and outputs whenever needed via “Reply” nodes.

Available Nodes at a Glance

Chatflow and Workflow share the same node system: the trigger node (Input), AI nodes (LLM, Knowledge Base, Agent, Intent Recognition), logic nodes (Conditional Branch, Loop, Iteration), data-processing nodes (Code Execution, Variable Assignment, Variable Aggregator, Variable Formatter, Parameter Extraction, Document Parser, HTTP Request, Array Operations), tool-type nodes (System Plugins, Workflow, MCP, Custom Tools) — plus the Chatflow-only “Reply” node.

See “Node Reference” for each node’s purpose and configuration.

Debugging

Chatflow debugging happens as a multi-turn conversation in the debug panel: simulate a real user’s consecutive questions to verify context memory, branch logic, and answer quality.

Each turn can be expanded to inspect that turn’s per-node execution, including each node’s input, output, and error info — making it easy to tell whether a wrong answer came from knowledge retrieval, model generation, or a conditional branch. Past runs are in “Debug History.”

Tip: Debug with at least an “opening question + follow-up” combination to verify context memory behaves as expected.

Publishing

Publish to App Square

Click “Publish to App Square” in the “Save” button’s hover drop-down to save the latest content and list the app. Once listed, everyone on the platform (including Viewers) can use it in the App Square. A conversational app page provides a conversation history panel (new conversation, switch between past conversations) and an app info drawer.

Publish as a Lansenger AI Assistant

Chatflow supports publishing as a Lansenger AI Assistant. Once published, the app is listed in the Lansenger client’s “AI Assistant Marketplace,” and end users can chat with it right inside Lansenger. The entry is the app details page → “Publishing Channels” tab → “Lansenger AI Assistant.”

Before publishing:

• The publishing account must be linked to Lansenger; if not, follow the pop-up to the Lansenger authorization center — publishing runs automatically once QR-code authentication passes.
• Publishing as a Lansenger AI Assistant requires the Admin role or above.
• The assistant title defaults to the app name (within 10 characters) and the description defaults to the app description (within 20 characters).
• After a successful publish, the button changes to “Published” and cannot be repeated.

For the remaining channels — web link, API access — and Version Management, see “Publishing Channels.”

Related Reading

• “Workflow”: The single-run automation app — shared canvas and debugging capabilities.
• “Node Reference”: Purpose, key settings, and outputs of each node.
• “Publishing Channels”: Full flows for publishing to the App Square, as a Lansenger AI Assistant, and more.
• “Logs & Monitoring”: Viewing and annotating Conversation Logs, and the Data Dashboard.