AI Provider Installation Guide for Administrators

Purpose of This Guide

This guide provides administrators with step-by-step instructions for configuring AI providers within the QuikData platform.

QuikData uses a centralized AI Provider table (DRControl.dbo.AIProvider) to register and manage cloud-based and local AI models that power features such as summarization, document classification, privilege analysis, and other AI-assisted review capabilities.

Because AI provider configuration is global across the entire QuikData installation (not per Environment or Room), correct setup is essential for ensuring consistent model availability, accurate billing, and reliable system performance.

This guide explains how to:

• Register supported AI providers (OpenAI, Anthropic, Jan, TabbyAPI, Azure OpenAI)

• Configure model metadata, capabilities, and API interaction details

• Encrypt and store provider API keys

• Maintain accurate token cost and pricing information

• Insert one row per model into the AIProvider database table

What You Will Need

Before configuring AI providers within the QuikData platform, ensure you have the following:

1. QuikData Version 5.8 or greater

2. Access to the QuikData DRControl Database
   Administrator or DB access is required to insert rows into the dbo.AIProvider table.

3. A Registered API Account with a Supported AI Provider
   QuikData currently supports:

   • OpenAI 

   • Anthropic

   • Azure OpenAI (uses ProviderName = OpenAI)

   • Jan (local LLM hosting)

   • TabbyAPI (local LLM hosting)

   • Other OpenAI-compatible API endpoints

   Providers must expose either:

   • An OpenAI Chat Completions–compatible API, or

   • An Anthropic Messages–compatible API

4. An Encrypted API Key
   API keys must be encrypted using the QuikData Key Encrypter Utility before being stored in the database.
   If you do not have this utility, a QuikData Support Representative can encrypt the value for you.

Supported AI Providers

The ProviderName field determines which AI engine integration QuikData uses.
Only the following values are supported:

Notes:

• Each model variant must be registered as a separate row.

• OpenAI is used for both standard OpenAI and Azure OpenAI APIs.

• Jan and TabbyAPI enable fully local inference (no cloud dependency).

How to Register an AI Model in QuikData

To make an AI model available inside the QuikData platform, administrators must insert a row into the DRControl.dbo.AIProvider table.
Each row corresponds to one model — for example:

• GPT-4o

• GPT-4o-mini

• Claude 3.5 Sonnet

• A TabbyAPI-hosted local LLaMA model

This guide will walk you through:

1. The fields that must be populated

2. What each field represents

3. How QuikData uses these values internally

4. An example SQL INSERT statement you can use as a template to register your AI model

The next section describes each column in the AIProvider table so you can correctly supply the required values when adding new models.

AIProvider Table – Field Definitions

The following table describes every column required to register an AI model in QuikData.

Example: Registering AI Models

Below is an example showing how to register two OpenAI GPT-4o models.

INSERT INTO [AIProvider] (
    [ProviderName],
    [EngineName],
    [EngineDescriptionLocalizedJSON],
    [ApiEndpointURL],
    [MaxContextWindowTokens],
    [MaxOutputTokens],
    [SupportsStructuredOutput],
    [SupportsImageAnalysis],
    [DefaultAPIKeyEncrypted],
    [InputTokensCostPerMillion],
    [OutputTokensCostPerMillion],
    [InputTokensPricePerMillion],
    [OutputTokensPricePerMillion]
) VALUES (
    N'OpenAI',
    N'gpt-4o-2024-08-06',
    N'{ "EN": "Ideal for complex, nuanced tasks requiring the highest quality outputs, with a substantial context window." }',
    N'https://api.openai.com/v1/chat/completions',
    128000,
    16384,
    1,
    0,
    N'<encrypted API key>',
    2.500,
    10.000,
    7.500,
    30.000
);

INSERT INTO [AIProvider] (
    [ProviderName],
    [EngineName],
    [EngineDescriptionLocalizedJSON],
    [ApiEndpointURL],
    [MaxContextWindowTokens],
    [MaxOutputTokens],
    [SupportsStructuredOutput],
    [SupportsImageAnalysis],
    [DefaultAPIKeyEncrypted],
    [InputTokensCostPerMillion],
    [OutputTokensCostPerMillion],
    [InputTokensPricePerMillion],
    [OutputTokensPricePerMillion]
) VALUES (
    N'OpenAI',
    N'gpt-4o-mini-2024-07-18',
    N'{ "EN": "Cost-efficient model offering strong performance for lightweight tasks." }',
    N'https://api.openai.com/v1/chat/completions',
    128000,
    16384,
    1,
    0,
    N'<encrypted API key>',
    0.150,
    0.600,
    0.450,
    1.800
);

Notes for Administrators

• Execute this script against the QuikData control database (the database that contains the AIProvider table).

• You may prefix the table with schema or database name if required by your environment:

  • [dbo].[AIProvider]

  • [ClientName_Control].[dbo].[AIProvider]

• Only one row should be inserted per AI model.

Post-Update Procedures

After updating the AIProvider table, the QuikData platform must refresh its internal AI configuration cache.
AI Providers are loaded once during web application startup and remain cached until explicitly refreshed.

To ensure newly added or updated AI models become available to users:

1. Log into your QuikData instance as an Administrator.

2. Navigate to:
   Site Admin → Site Settings → Refresh Cache

3. Click Refresh Cache to reload all cached configuration data, including AI Providers.

This operation instructs the QuikData application to:

• Reload the AI Provider list

• Update available models displayed in dropdowns

• Apply any changes made to model descriptions, costs, or API configurations

• Ensure all AI features use the most current provider information

Important:
If the cache is not refreshed, newly added models will not appear in the UI until the next full application restart.