Table of Contents

• AI Co-Pilot Guide - Enterprise Edition
• Configuring SWIRL AI Co-Pilot, Enterprise Edition
  • Roles for Generative AI/Large Language Models
  • Adding Chat to an AI Provider
  • GAI/LLM Requirements

AI Co-Pilot Guide - Enterprise Edition

This document applies only to SWIRL AI Co-Pilot, Enterprise Edition.

Configuring SWIRL AI Co-Pilot, Enterprise Edition

Roles for Generative AI/Large Language Models

SWIRL AI Connect, included with Co-Pilot, defines four roles for GAI/LLMs, as noted below. SWIRL AI Co-Pilot adds a fifth role, "Chat". This can be assigned to any sufficiently capable LLM.

Role	Description	Default
reader	Providing embeddings for SWIRL's Reader LLM to use when re-ranking search results	spaCy
query	Provide completions for query transformations	OpenAI GPT-3.5 Turbo
connector	Provide completions for direct questioning (not RAG)	OpenAI GPT-3.5 Turbo
rag	Provide completions for Retrieval Augmented Generation (RAG) using data retrieved by SWIRL	OpenAI GPT-4
chat	Provide messaging support for SWIRL AI Co-Pilot	OpenAI GPT-4

Adding Chat to an AI Provider

• Go to the swirl/aiproviders endpoint. For example, if using the default local install: http://localhost:8000/swirl/aiproviders

• Edit any AI Provider by adding the id value to the end of the swirl/aiproviders URL. For example: http://localhost:8000/swirl/aiproviders/4/

• Review the provider definition to ensure that the active property is true.

• Review the tags and default lists in the definition. For an AI Provider to provide Co-Pilot services:

1. Make sure active is true
2. Verify that api_key has a valid API key
3. Ensure model and any items in config are filled in correctly
4. The provider has chat in the tags list
5. The provider has chat in the defaults list

For example, here is the preloaded OpenAI GPT-4 provider, set for chat as well as RAG:

    {
        "id": 16,
        "name": "OpenAI GPT-4",
        "owner": "admin",
        "shared": true,
        "date_created": "2024-03-04T15:15:16.940393-05:00",
        "date_updated": "2024-03-04T15:15:16.940410-05:00",
        "active": true,
        "api_key": "<your-openai-api-key>",
        "model": "gpt-4",
        "config": {},
        "tags": [
            "query",
            "connector",
            "rag",
            "chat"
        ],
        "defaults": [
            "rag",
            "chat"
        ]
    }

Assuming the AI provider is correctly configured, you should be able to load the Co-Pilot using a browser. If you are using the default installation, this will be:

http://localhost:8000/galaxy/chat

Please refer to the AI Connect, Enterprise Edition, section on Connecting to Enteprise AI for more information on configuring AI Providers.

GAI/LLM Requirements

SWIRL Co-Pilot expects the following from GAI/LLMs that provide chat functionality:

• Support for receiving the chat history in each message, in reverse chronological order, i.e. the format used by the OpenAI Chat Completions API

• Prompt support of at least 3k tokens per message, with 6k+ preferred

• Currently, we recommend OpenAI GPT-4 (all variants including 'o'), Anthropic Claude 3 or Google Gemini Pro/Ultra. But there are many incredible models, and any of them that support the chat history (noted above) may work. Please let us know what works, and doesn't work!