The Configure section is the main place where you control how Model HQ works and looks. You can change things like the design, choose your models, set up databases, manage RAG settings, and turn on safety and security options.

After setup, you get access to all these controls in one spot. Think of it like a control panel for the whole app. You can adjust how bots behave, how the screen looks, how data is stored, and how Model HQ connects to other tools.

These settings help users customize Model HQ to fit their own needs, workflows, and rules.

After completing the initial setup, users gain access to comprehensive configuration controls that shape how Model HQ behaves, appears, and integrates with external systems. The Configure interface serves as the central control panel for tailoring the application to specific use cases, organizational requirements, and security policies. Settings can be adjusted to control default bot behavior, user interface appearance, RAG performance parameters, database connections, safety controls, and enterprise integrations.

These configurations enable organizations to transform Model HQ from a general-purpose AI application into a customized solution that aligns with their unique workflows, branding guidelines, and compliance requirements.

Understanding these configuration options is essential for optimizing Model HQ's performance, security posture, and user experience across different deployment scenarios.

1. Launching the configuration interface

To begin, the Configure button (⚙️) located in the top right side of the main menu can be selected.

2. Configuration interface overview

After launching the configuration section, the interface displays the following key options:

The configuration interface is organized into the following main sections:

Each section provides granular control over specific aspects of Model HQ's functionality, enabling customization tailored to organizational needs and deployment environments.

3.1 App

The App section controls global application behavior, default experiences, visibility of features, and runtime modes within Model HQ. These settings determine how users interact with bots, agents, datasets, services, and enterprise integrations.

3.1.1 Default Bot

Defines which bot is launched by default when the application starts - the listed bots include Bots that are Model HQ template bots as well as any bots the user may have created with Model HQ.

Available Options:

• AC Repair Bot
• Bot with Agents
• Dataset Bot
• Demo Bot
• Fast Start Chatbot
• Model HQ Embedded Agent Bot
• Model HQ API Server Bot
• Model HQ Model Sampler
• Persona Bot
• SQL Bot

Recommendation:

• Use Fast Start Chatbot for quick onboarding and general usage.
• Use API Server Bots when running Model HQ primarily as a backend service.
• Use Bot with Agents for multi-agent workflows.

3.1.2 Agent Process Run Mode

Controls how agent execution details are displayed during runtime.

Options:

• Detailed Process Shows step-by-step execution details for each agent.
• Summary View Displays a concise execution summary.

When to use:

• Detailed Process for debugging and development.
• Summary View for end-user or production environments.

3.1.3 Dev Mode

Toggles developer-focused features and build options.

Options:

• ON Enables developer tools, configuration panels, and advanced options.
• OFF Hides developer options for end-user deployment.

Recommendation:

• Enable during development.
• Disable before deploying to non-technical users.

3.1.4 App Mode

Controls whether application-level apps are displayed.

Options:

• ON Displays available apps within the UI.
• OFF Hides app-level UI elements.

3.1.5 Connected Enterprise Servers

Enables connectivity to enterprise Model HQ servers.

Options:

• ON Allows connection to enterprise accounts and managed servers.
• OFF Disables all enterprise server connections.

Requirement:

• Must be enabled if using enterprise-hosted Model HQ infrastructure.

3.1.6 Air Gap Mode

Runs the application completely disconnected from external networks.

Options:

• ON Blocks all external connections, including data ingestion.
• OFF Allows internet and external system access.

Use cases:

• Secure or regulated environments.
• Offline or isolated deployments.

3.1.7 Main Menu Cards

Controls which primary feature cards appear in the main menu.

Available Cards:

• Agents
• Bots
• Sources
• Datasets
• Services
• Integrations
• Projects
• Models

Purpose:

• Customize UI visibility based on user roles or use cases.
• Simplify the interface for focused workflows.

3.1.8 Agent Processes

Select specific agent processes to expose in the main menu.

Examples:

• AC Field Tech Support
• Cloud API Agent
• Conditional Agent
• Contract Analyzer
• Customer Support
• Dataset Analysis
• Financial Data Extractor
• Image Tagger
• Image Generation Agent
• Research Process
• Summarize Website

Behavior:

• Only selected agents appear as quick-access options.
• Useful for role-based or task-specific setups.

3.1.9 Custom Bot Apps

Choose which custom bots appear in the main menu.

Available Bots:

• Bot with Agents
• Dataset Bot
• Demo Bot
• Fast Start Chatbot
• Model HQ Embedded Agent Bot
• Model HQ API Server Bot
• Model HQ Model Sampler
• Model HQ API Server Biz Bot
• Model HQ Biz Bot
• Persona Bot
• SQL Bot

Purpose:

• Control bot availability per deployment.
• Reduce UI clutter.

3.1.10 Source Cards

Select source configurations to display in the main menu.

Examples:

• test_source.jsonl

Used to provide quick access to predefined data sources.

3.1.11 Dataset Cards

Controls which datasets are accessible from the main menu.

Examples:

• test_dataset.jsonl

Ideal for fast dataset-driven workflows.

3.1.12 Custom Service Cards

Displays custom service integrations in the main menu.

Examples:

• AC_Field_Tech_Support

Used to expose REST, MCP, or backend services directly in the UI.

3.1.13 Auto-Restart UI

Controls automatic UI restart behavior if a session or connection drops.

Options:

• ON Automatically restarts the UI on disconnect.
• OFF Requires manual restart.

Note:

• If enabled, use the on-screen shutdown button to stop the application.

3.1.14 Reset Defaults

Resets all App settings to their default values.

Use case:

• Recover from misconfiguration.
• Quickly revert experimental changes.

These App settings allow fine-grained control over user experience, security posture, feature exposure, and runtime behavior across development, enterprise, and air-gapped deployments.

3.2 Services

This is a master panel of services that are available to use in creating agents. Making the selection here will ensure that each of these services are displayed as an option in the Nodes in agents. (Note: Services outside of this master list can be selected at time of use in the agent canvas if not pre-selected here.)

3.2.1 Core Services

3.2.2 Classifiers

3.2.3 Datasets

3.2.4 Specialized Services

3.2.5 Integrations

3.2.6 Custom Services

Custom services were described as those services that had been created within Model HQ to support domain-specific workflows, specialized automation, or organization-specific requirements. These services extend the standard catalog by incorporating tailored logic, configurations, and business rules.

3.3 UI

The UI section enables fast and comprehensive customization of Model HQ's visual appearance, including bot names, icons, colors, and other interface elements.

This configuration panel allows the entire Model HQ appearance to be customized according to organizational branding requirements. The following elements can be modified:

• Theme: Light or dark mode selection
• Title: Application window title
• App Title: Display name shown in the interface
• Company Name: Organization name displayed throughout the application
• Company Website: URL link associated with the organization
• Icon: Custom application icon/logo
• Header Color: Color scheme for the top navigation bar
• Footer Color: Color scheme for the bottom interface elements
• Background Color: Main interface background color

These customization options are designed to enable quick and seamless enterprise branding, allowing organizations to align Model HQ with their unique visual identity and corporate standards.

3.4 RAG

The RAG (Retrieval Augmented Generation) section controls how Model HQ retrieves, ranks, and injects external knowledge into model prompts. These settings directly affect answer quality, relevance, performance, and memory usage when working with documents and sources.

3.4.1 Text Chunk Size

Defines the target size of text segments created during document parsing.

Purpose:

• Controls how source documents are split before embedding.
• Smaller chunks improve precision.
• Larger chunks preserve more context.

Example:

600

Guidance:

• Use smaller values for highly structured or technical documents.
• Use larger values for narrative or long-form content.

3.4.2 Context Top N

Specifies the number of top-ranked text chunks selected to build the context for the model.

Purpose:

• Limits how many chunks are initially considered relevant.
• Improves speed by narrowing the search space.

Example:

5

3.4.3 Context Target Size

Defines the target token size for the final context passed to the model.

Behavior:

• If the target size is not reached using the selected top N chunks, additional chunks are added until the target is met.

Example:

500

Impact:

• Larger values provide more context but increase token usage.
• Smaller values reduce latency and cost.

3.4.4 Reranker Max Samples

Sets the maximum number of text chunks evaluated in memory by the reranker.

Purpose:

• Controls how many candidates are semantically re-scored.
• Higher values improve ranking accuracy but increase memory and compute usage.

Example:

1000

3.4.5 Context Comparison Prompt

Adds a system instruction to the prompt when multiple source documents are used.

Default Example:

Here are several sources - please use as the basis for answering questions, and cite the specific source, if used, in generating your answer.

Use case:

• Ensures answers reference and differentiate between multiple documents.
• Encourages source-aware responses.

3.4.6 RAG Prompt Instruction

Custom instruction passed directly into the RAG pipeline.

Purpose:

• Guides how retrieved context should be interpreted.
• Can enforce tone, citation format, or reasoning style.

Example Use Cases:

• Enforce strict citation requirements.
• Limit answers to retrieved content only.
• Control summarization vs extraction behavior.

3.4.7 RAG Model

Selects the language model used to generate responses using retrieved context.

Example:

llama-3.2-3b-instruct-ov

Recommendation:

• Use instruction-tuned models for best RAG performance.
• Smaller models improve speed, larger models improve reasoning.

3.4.8 Reranker Model

Defines the semantic ranking model used to reorder retrieved text chunks.

Example:

jina-reranker-v1-tiny-en-ov

Role:

• Improves relevance by re-ranking chunks beyond vector similarity.
• Critical for multi-document or noisy datasets.

3.4.9 Embedding Model

Selects the model used to convert text into vector embeddings.

Example:

all-mini-lm-l6-v2-ov

Impact:

• Affects retrieval accuracy and embedding performance.
• Smaller models are faster, larger models capture deeper semantics.

3.4.10 Use Wikipedia as Source

Controls whether Wikipedia is included as an external retrieval source.

Options:

• ON Allows Wikipedia content to be used during retrieval.
• OFF Restricts retrieval to configured sources only.

Recommendation:

• Enable for general knowledge queries.
• Disable for enterprise or private datasets.

3.4.11 Update Behavior

RAG settings can be updated at any time.

Notes:

• Changes take effect immediately for new queries.
• No restart is required.

These RAG configurations allow fine-tuned control over document retrieval, ranking, and prompt construction, enabling accurate, scalable, and context-aware AI responses across diverse data sources.

3.5 DB

The DB section provides tools for managing resources on the local Model HQ database.

The local Model HQ database can be configured to build, view, delete, and manage resources. This database is utilized for querying SQL tables in Chat and Agents, enabling structured data interactions and query-based workflows. Database management capabilities include schema creation, table configuration, and resource cleanup operations.

3.6 Prompts

The Prompts section allows system-level prompts to be added and pre-configured for reuse across sessions.

Custom prompt templates can be created and stored for consistent behavior across different workflows. These pre-configured prompts help standardize model interactions, enforce specific response formats, and maintain consistent tone or style requirements throughout the application.

3.7 Server

The Server section configures the Backend API Server when running Model HQ in Headless mode. These settings define how external clients, agents, or applications connect to Model HQ without using the built-in UI.

3.7.1 Localhost or External IP

Determines whether the backend server is accessible only on the local machine or exposed over the network.

Options:

• Localhost Restricts access to the local machine only.
• External IP Allows other machines and services to connect.

Guidance:

• Use Localhost for development and testing.
• Use External IP for production or shared environments.

3.7.2 IP Address

Specifies the IP address the backend server binds to.

Behavior:

• When Localhost is selected, this is automatically set to 127.0.0.1.
• When External IP is selected, provide a valid local or public IP.

Example:

192.168.29.93

3.7.3 Port

Defines the port on which the Backend API Server listens.

Default:

8088

Notes:

• Change only if the default port is already in use.
• Ensure the port is open in firewall and security group rules if exposed externally.

3.7.4 Workers

Controls the number of lightweight worker processes handling incoming requests.

Purpose:

• Enables basic concurrency for API calls.
• Designed to remain lightweight.

Example:

4

Recommendation:

• Keep this value low.
• Increase only if you experience measurable performance bottlenecks.

3.7.5 Trusted Key

Optional shared secret used to authenticate API requests.

Usage:

• Acts as a simple access control mechanism.
• Must be provided by clients when authentication is enabled.

Example:

my-secure-trusted-key

3.7.6 Require Trusted Key

Controls whether the Trusted Key is mandatory for API access.

Options:

• Require Key All API requests must include the trusted key.
• No Key API is accessible without authentication.

Security Guidance:

• Enable Require Key for any network-exposed or production deployment.
• Use No Key only in isolated or local environments.

3.7.7 Access Behavior

When enabled and configured correctly:

• Agents and services can communicate with Model HQ via REST APIs.
• Headless mode allows full automation without the UI.

3.7.8 Applying Changes

Server configuration updates take effect immediately.

Notes:

• Restart is not required unless explicitly prompted.
• Verify connectivity after changes using a test API call.

This configuration enables secure and flexible deployment of Model HQ as a backend service, supporting both local development and production-grade headless integrations.

3.8 Controls

The Controls section defines global governance, security, validation, and safety behaviors for model execution and inference. These settings influence how models are loaded, validated, executed, and how sensitive data is handled across the platform.

3.8.1 Inference Persistence

Controls whether inference results are stored locally.

Options:

• Do Not Save Inference data is not persisted.
• Save All inference outputs are saved to the local database.

Guidance:

• Disable saving for privacy sensitive or transient workloads.
• Enable saving for debugging, audits, or analytics.

3.8.2 Model Repository Source

Specifies the repository used to download models.

Options:

• Hugging Face Pulls models from the Hugging Face ecosystem.
• Azure Pulls models from Azure based model storage.

Use Case:

• Select Azure for enterprise managed environments.
• Select Hugging Face for broader open model access.

3.8.3 Model Validation on Download

Controls hash verification when a model is first downloaded.

Options:

• Validate on Download Ensures model integrity at download time.
• Skip Validation Downloads without integrity checks.

Recommendation:

• Keep validation enabled in production environments.

3.8.4 Model Validation on Load

Controls whether integrity checks occur each time a model is loaded from disk.

Options:

• Validate on Every Load Verifies model consistency on every read.
• Skip Validation Skips repeated validation for faster startup.

Tradeoff:

• Validation improves safety.
• Skipping improves performance.

3.8.5 Cloud API Access

Controls whether public cloud APIs can be used.

Options:

• Enable Cloud API Allows use of external cloud based model APIs.
• Disable Cloud API Restricts execution to local models only.

Security Note:

• Disable cloud APIs for air gapped or compliance restricted environments.

3.8.6 Prompt Preview

Controls visibility of the final prompt before execution.

Options:

• On Displays the constructed prompt.
• Off Executes without preview.

Use Case:

• Enable during development and debugging.
• Disable for streamlined end user experiences.

3.8.7 Enforcement Action

Defines how the system responds when sensitive patterns are detected.

Options:

• Redact Automatically masks detected content.
• Warn Flags content but allows execution.

3.8.8 Pattern Redaction

Specifies which sensitive data patterns are detected and handled.

Available Patterns:

• us_ssn
• aba_routing_numbers
• email
• credit_card
• us_dl
• us_passport
• date
• iban
• in_pan
• url
• crypto
• phone_number

Behavior:

• Selected patterns are either redacted or warned based on the configured action.

3.8.9 Classifier Tests

Enables built in content classifiers for safety and quality.

Available Tests:

• prompt_injection_detection
• toxic_detection
• language_detection
• bias_detection

Purpose:

• Detects unsafe, malicious, or policy violating content.
• Improves trust and governance across model usage.

3.8.10 Automated Configuration

Choose For Me automatically selects recommended defaults based on environment and use case.

Use Case:

• Quick setup for new deployments.
• Safe baseline configuration for most users.

The Controls section provides centralized enforcement of security, compliance, and safety policies, ensuring consistent and governed model behavior across all applications and agents.

3.9 Templates

The Templates section enables custom templates to be created for accelerated bot and agent development.

Template management provides options to build new templates or edit and view existing ones. Custom templates streamline the creation process by providing pre-configured structures, default settings, and reusable components. This significantly reduces development time when building multiple bots or agents with similar configurations or workflow patterns.

3.10 Connections

The Connections screen allows backend API endpoints to be configured for Model HQ connectivity.

3.10.1 API Name

A descriptive label to identify the connection within the UI can be provided.

Example: Model HQ Server

3.10.2 IP Address

The address of the API server should be specified.

Example: 127.0.0.1 for local server

3.10.3 Port

The port where the server is running should be entered.

Example: 52640

3.10.4 Secret Key

An optional key used to authenticate requests between Model HQ and the server can be provided.

3.10.5 Protocol

The connection protocol should be selected based on the deployment environment.

• HTTP for local or internal setups
• HTTPS for secure or remote connections

3.10.6 Transfer local credentials

If enabled, local app credentials are copied to the server.
Use only in trusted environments.

3.10.7 Activate Connection

Turns the connection on or off.

Only active connections are used.

3.11 Reset

The Reset section provides options to reset the application or specific configurations.

The following reset options are available:

• Default Bots: Restores default bot configurations
• Default Agents: Restores default agent configurations
• Reset Model Catalog: Refreshes the model catalog to default state
• Delete Models: Removes downloaded models from local storage
• Clear All: Clears all loaded state for agents, bots, and sources

Reset operations should be performed with caution. Once reset, models, custom applications, and other Model HQ-related files will be deleted and will need to be re-created or re-downloaded.

3.12 Theme toggler

The theme toggler provides a quick switch between light and dark display modes for the interface.

Conclusion

This document described the comprehensive configuration options available in Model HQ's Configure interface. The configuration system enables fine-grained control over application behavior, visual appearance, RAG performance, database connectivity, safety controls, and enterprise integrations. Key configuration areas include App settings for controlling feature visibility and runtime modes, UI customization for branding and appearance, RAG parameters for optimizing document retrieval and context building, Server settings for headless deployments, Controls for safety and compliance enforcement, and Connections for backend API integration. Understanding and properly configuring these settings enables organizations to optimize Model HQ for their specific use cases—whether prioritizing security in air-gapped environments, customizing branding for enterprise deployments, tuning RAG performance for document-heavy workflows, or establishing robust safety controls for production systems. The flexibility provided by these configuration options allows Model HQ to adapt from development environments to production deployments while maintaining consistent behavior, appearance, and security posture across different scenarios.