Skip to Content
DocumentationFeaturesChat Flows

Chat Flows

Build multi-turn conversational workflows with Claude using message documents and the messagesSearchTag pattern.

Example

import { z } from 'zod';
import { LlmGenerateTextTool, LlmMessageDocument } from '@loopstack/llm-provider-module';
import { BaseWorkflow, Initial, InjectTool, Transition, Workflow } from '@loopstack/common';
 
@Workflow({ uiConfig: __dirname + '/chat.ui.yaml' })
export class ChatWorkflow extends BaseWorkflow {
  @InjectTool({ provider: 'claude', model: 'claude-sonnet-4-6' }) llmGenerateText: LlmGenerateTextTool;
 
  @Initial({ to: 'waiting_for_user' })
  async setup() {
    await this.repository.save(
      LlmMessageDocument,
      { role: 'user', content: this.render(__dirname + '/templates/systemMessage.md') },
      { meta: { hidden: true } },
    );
  }
 
  @Transition({ from: 'waiting_for_user', to: 'ready', wait: true, schema: z.string() })
  async userMessage(payload: string) {
    await this.repository.save(LlmMessageDocument, { role: 'user', content: payload });
  }
 
  @Transition({ from: 'ready', to: 'waiting_for_user' })
  async llmTurn() {
    const result = await this.llmGenerateText.call({
      model: 'claude-sonnet-4-6' },
      messagesSearchTag: 'message',
    });
    await this.repository.save(LlmMessageDocument, result.data!, { id: result.data!.id });
  }
}

YAML Config

title: 'Chat Assistant'
 
ui:
  widgets:
    - widget: prompt-input
      enabledWhen:
        - waiting_for_user
      options:
        transition: userMessage

How Message Accumulation Works

  1. All messages are saved as LlmMessageDocument — automatically tagged with message
  2. messagesSearchTag: 'message' tells llmGenerateText to collect all documents with that tag as conversation history
  3. Each new message adds to the conversation — the LLM sees the full history on every turn

Chat Loop Flow

setup → waiting_for_user → [user sends message] → ready → llmTurn → waiting_for_user (loop)
  1. @Initial — Create system message (hidden from UI)
  2. Workflow enters waiting_for_user — UI shows the prompt-input widget
  3. User sends message → userMessage fires, saves user message as document
  4. llmTurn fires — calls Claude with full message history, saves response
  5. Workflow returns to waiting_for_user — loop continues

Combining with Tool Calling

Add tool calling to a chat flow by combining the patterns from AI Tool Calling:

// Configure tools on @InjectTool
@InjectTool({ provider: 'claude', model: 'claude-sonnet-4-6', tools: ['getWeather', 'searchDatabase'] })
llmGenerateText: LlmGenerateTextTool;
 
@Transition({ from: 'ready', to: 'prompt_executed' })
async llmTurn() {
  const result = await this.llmGenerateText.call({
    messagesSearchTag: 'message',
  });
  this.llmResult = result.data;
}
 
@Transition({ from: 'prompt_executed', to: 'awaiting_tools', priority: 10 })
@Guard('hasToolCalls')
async executeToolCalls() { ... }
 
@Transition({ from: 'prompt_executed', to: 'waiting_for_user' })
async respond() {
  await this.repository.save(LlmMessageDocument, this.llmResult!.message, {
    meta: { response: this.llmResult!.response, provider: 'claude' },
  });
}

Registry References

Last updated on