External Pipeline Requirements

External pipeline detailed requirements

Overview

External services must implement specific requirements to be compatible with GLChat's external pipeline registration feature. This guide covers the core API endpoints, attachment handling, and conversation history support.

Key Requirements

1. Implement /verify endpoint - For registration and metadata retrieval

2. Implement /responses endpoint - OpenAI-compatible Responses API interface

3. Support header authentication - For API key validation

4. Return valid JSON responses - With required fields

5. Support attachment handling - For file-based queries (Optional)

6. Support conversation history --- For multi-turn conversations (Optional)

Required Endpoints

1. Service Discovery Endpoint

Endpoint: GET /verify

Purpose: Called by GLChat during pipeline registration to verify connectivity and retrieve service metadata.

Request Headers:

Header	Value	Required	Description
X-API-Key	{api_key}	Yes	API key for authentication

Response Schema:

{
  "name": "my-ai-service",
  "version": "1.0.0",
  "description": "A custom AI service",
  "model_names": ["gpt-5", "opus-4.6"],

  "supports_attachments": true,
  "supported_file_types": {
    "document": [".pdf", ".txt", ".md", ".docx"],
    "image": [".jpg", ".png", ".jpeg"]
  },
  "max_file_size": 10485760,
  "max_file_count": 5
}

Field Specifications:

Field	Type	Required	Description
name	string	Yes	Non-empty after stripping whitespace
version	string	Yes	Non-empty after stripping whitespace
description	string	No	Optional, string if provided
model_names	string[]	No	Optional, array of non-empty strings if provided
supports_attachments	boolean	No	Whether the pipeline accepts file attachments
supported_file_types	dict	No	File type categories with extensions
max_file_size	integer	No	Maximum file size in bytes
max_file_count	integer	No	Maximum number of files per request

⚠️ Notes: The model names for external pipeline are currently not used in GLChat. This will be addressed as a future enhancement.

2. Responses Endpoint

Endpoint: POST /responses

Purpose: Called by GLChat during chatbot conversations to generate AI responses. Supports simple queries and multi-turn conversations with attachments.

Note:

1. GLChat uses OpenAI's newer Responses API (not the legacy Chat Completions API).

2. The service should support streaming modes

3. The service /responses should mimic OpenAI Responses. Here's guides from OpenAI https://developers.openai.com/api/reference/resources/responses/methods/create

Example

For example, follow instructions on External Pipeline Main Section

⚠️ Disclaimer: The attached code is just a minimum dummy example