Glow Server Package

Server-side logic for Glow AI chat

Overview

@zooly/glow-server provides the server-side logic for the Glow chat system: streaming LLM responses, tool execution, cache lookup, context switching, and chat lifecycle management.

Package Details

Package Name: @zooly/glow-server
Location: packages/glow-server
Type: Server-side module (Node.js)

Key Exports

Request Handlers

Export	Purpose
`handleGlowChatStream`	Entry point for streaming; validates session, delegates to cache/core, returns stream response
`handleGlowChatCore`	Main LLM flow: prompts, tools, streamText
`handleGlowChatWithCache`	Wraps core; checks question cache when enabled
`handleGlowChatGenerate`	Generate flow: getOrCreateGlowChat + handleGlowChatWithCache + stream

Chat Lifecycle

Export	Purpose
`getOrCreateGlowChat`	Create new chat or return existing by chatId
`getGlowChatData`	Load glowChat + glowSettings by glowChatId
`createGlowChat`	Create chat with options
`createNewGlowChat`	Create new chat for a flow slug

Context & Tools

Export	Purpose
`changeChatContext`	Switch chat to new flow by slug
`returnToContext`	Return to previous flow
`buildContextStack`	Build context from tool call logs
`getCurrentNodeTools`	Resolve tools for current node
`getToolsList`	All tools for a flow
`executeApiTool`	Call external API for API tools
`toolsBase`	Base tools (moveToNodeId, changeChatContext, returnToContext)

Prompts & Helpers

Export	Purpose
`getGlowChatSystemPrompt`	Flow-level system prompt
`getGlowChatInstructionsSystemPrompt`	Node-level instructions
`getFirstNodeMetadata`	First node metadata from flow
`getCurrentNodeMetadata`	Current node metadata
`moveToNodeIdInstructions`	Instructions for moveToNodeId tool

Utilities

Export	Purpose
`extractQuestionFromMessage`	Extract question text for cache lookup
`transformNodesMetadata`, `transformField`, `transformFields`	Field transformers
`generateZodSchema`	Generate Zod schema from tool fields
`defaultOnFinish`	Default onFinish handler
`prepareStepHandler`	Prepare step handler
`onStepFinishHandler`	Step finish handler
`saveMessageData`	Save message data

Model

Export	Purpose
`defaultModel`	Default model config (openai/gpt-4.1)

Dependencies

ai - AI SDK (streamText, convertToModelMessages, etc.)
@zooly/db - Database access, cache, embeddings
@zooly/types - GlowChat, GlowSettings, GlowTool, etc.
@zooly/util - extractQuestionFromMessage
@zooly/llm-mock - Mock LLM for testing
next - Request types
next-auth - Session types
zod - Schema validation
uuid - ID generation

Usage

Streaming Handler

import { handleGlowChatStream } from "@zooly/glow-server";

export async function POST(request: Request) {
  const session = await getServerSession(authOptions);
  return handleGlowChatStream(request, session ?? undefined);
}

Get or Create Chat

import { getOrCreateGlowChat } from "@zooly/glow-server";

const { glowChat, glowSettings } = await getOrCreateGlowChat({
  chatId: "my-chat-id",
  glowSlug: "backstage-mini-app",
  restart: false,
});

See App Integration for full integration steps.

Glow Client Package Auth Overview

On This Page

Overview Package Details Key Exports Request Handlers Chat Lifecycle Context & Tools Prompts & Helpers Utilities Model Dependencies Usage Streaming Handler Get or Create Chat