AI System¶

The AI system helps triage maintenance requests. It classifies issues, assigns priority, and supports image-based intake.

What It Does¶

Classifies maintenance issues from text.
Predicts priority for faster routing.
Supports images and videos when media is attached.

How AI Is Used In The System¶

User message or media
  -> text model and/or vision model
  -> fusion engine
  -> final category and priority
  -> ticket creation or chat reply
  -> AI log saved for review

Models¶

Text Model¶

Runs in chatbot.py.
Uses OpenAI when available.
Falls back to local DistilBERT models when needed.
Returns category, priority, and confidence.

Vision Model¶

Runs in vision_ai.py.
Analyzes uploaded images and videos.
Uses a YOLO model plus metadata mapping for labels.

Fusion¶

Runs in fusion_engine.py.
Combines text and vision output into one result.
Uses the stronger priority when the two models disagree.

AI classification and fusion pipeline Caption: AI pipeline view showing classification flow, confidence handling, and model outputs.

AI and dispatch workflow diagram Caption: Request intake moving through AI, ticket creation, and dispatch output.

Logging And Feedback¶

AI output is stored in ai_logs and on the ticket record.
Logged fields include category, confidence, auto-assignment, and model version.
Feedback can be sent to /ai/logs/{id}/feedback to record the actual category later.

POST /chat
POST /upload-image
PATCH /ai/logs/{id}/feedback
GET /ai/status