Triage common

Triage common provides a source-neutral platform for importing messages, analysing them with AI, recording auditable results, and coordinating follow-up actions. It enables organisations to build consistent, transparent message-processing workflows across multiple communication channels while keeping source acquisition, AI processing, and task execution separate.

Features

Triage provides a configurable AI-powered workflow engine for processing inbound communications.

Features include:

• Source-neutral message processing architecture.
• AI-assisted analysis using Drupal AI providers.
• Complete audit history of AI requests and responses.
• Configurable workers and tasks that execute in a defined sequence.
• Needs identification and matching.
• Topic classification and topic maps.
• Recommended action generation.
• Executive summary creation.
• Queue-based processing through Service Queue.
• Assistant-based workflows for simulations, reporting, and guided interactions.
• Import and export of workers, tasks, and needs as YAML configuration.

Typical use cases include:

• Triage and prioritisation of customer enquiries.
• Processing support or service mailboxes.
• Identifying needs and recommended interventions.
• Classifying messages by topic or category.
• Creating summaries of large communication volumes.
• Supporting operational and casework teams.
• Providing auditable AI-assisted decision support.

Post-Installation

The recommended approach is to install the project's supplied recipe, which configures Triage, related provider modules, worker definitions, assistants, prompts, and supporting configuration.

If installing manually:

1. Enable the module and required dependencies.
2. Navigate to Configuration > Triage > Get started.
3. Configure an AI chat provider and model.
4. Configure OAuth and queue services.
5. Create a Triage queue connection.
6. Review and customise workers and tasks.
7. Install and configure a source module such as Triage Mail.

Once configured, imported messages progress through a series of AI-driven workers and tasks, with all actions, prompts, responses, and outcomes recorded for later review and auditing.

Additional Requirements

Required:

• Drupal 11.
• AI module with a configured chat provider and model.
• Entity Name Uniqueness.
• Key.
• OAuth Provider.
• Service Queue.
• Service Bus Provider.
• At least one source module to acquire messages.

The supplied source implementation is Triage Mail.

Recommended Modules/Libraries

Triage is designed to operate alongside:

• Triage Mail for Microsoft 365 and mailbox-based message acquisition.
• OAuth Provider for secure authentication and token management.
• Service Queue for asynchronous processing.
• Service Bus Provider for scalable queue transport.
• Drupal AI providers such as Azure AI Foundry or other supported AI services.

Similar Projects

Several Drupal projects provide AI integration, workflow management, inbox handling, or case management capabilities.

Triage differs by combining:

• Source-neutral message acquisition.
• Configurable AI processing pipelines.
• Full auditing of AI requests and outputs.
• Queue-based task orchestration.
• Needs analysis and action recommendations.
• Reusable assistant-driven workflows.

The architecture allows new communication channels to reuse the same processing, reporting, auditing, and workflow services without duplicating business logic.

Supporting this Module

Bug reports, feature requests, documentation improvements, and merge requests are welcome through the Drupal.org issue queue.

Real-world feedback helps improve processing models, workflow design, reporting capabilities, and AI-assisted analysis features.

Community Documentation

Documentation currently includes:

• Installation and onboarding guidance.
• Configuration service walkthroughs.
• Worker and task configuration examples.
• Assistant workflow configuration guidance.
• Queue and processing architecture documentation.
• Developer guidance for creating new source modules and outcome handlers.

Additional walkthroughs, training materials, example configurations, and implementation patterns will be added as the project evolves.

Architecture

Triage separates message acquisition, AI analysis, workflow execution, and queue transport into distinct layers. Source modules acquire messages, workers determine which processing tasks should apply, AI-powered tasks perform analysis, and queue providers distribute work for execution. Every request, response, task, and outcome is retained for auditing and operational review.

This architecture enables organisations to build scalable, transparent, and configurable AI-assisted processing solutions while maintaining clear separation between communication channels, processing rules, and delivery infrastructure.