Taxonomy Manager

Create and manage shot taxonomies for the Media Processing Node

Taxonomy Manager

Define required photo checklists for media workflows

What Is a Taxonomy?

A taxonomy is a predefined checklist of required shots for a specific use case. It tells the Media Processing Node’s Coverage Scoring section what photos to expect and how to recognize them.

Each taxonomy includes:

• Required Shots: Must be captured before proceeding
• Conditional Shots: Required only when a specific condition is met
• AI Descriptions: How Claude identifies each shot type
• Extraction Links: Which shots should trigger OCR data extraction

Taxonomies are managed separately from workflows, allowing you to reuse the same shot requirements across multiple workflows and clients.

---

Accessing the Taxonomy Manager

Navigate to Settings → Taxonomy Manager to view, create, and edit taxonomies.

The List View displays all available taxonomies with:

• Taxonomy name and icon
• Industry category
• Shot count (required vs. conditional)
• Type badge (Plura Template or Custom)

Use the filters to narrow by industry or type, or search by name.

---

Creating a Taxonomy

1. Click Create Taxonomy

   Or clone an existing template to start with a base configuration.

2. Set Basic Info

   • Name: Descriptive name (e.g., “Auto Damage - Standard”)
   • Industry: Category for filtering (Auto, Homeowners, Healthcare, Logistics)
   • Icon: Visual identifier for the list view
   • Description: Optional context for your team
3. Add Shot Requirements

   Define each photo you need. See Shot Configuration below.

4. Configure Conditions

   For conditional shots, specify when they’re required.

5. Link Extraction Fields

   Connect shots to extraction fields for OCR (e.g., VIN plate → VIN extraction).

6. Save

   Your taxonomy is now available in the Media Processing Node dropdown.

---

Shot Configuration

Each shot in a taxonomy has the following fields:

Field	Purpose	Example
Shot ID	Variable name for code/triggers	front_full
Display Name	Human-readable label shown to users	”Front of Vehicle”
AI Description	How Claude recognizes this shot type	”Full front view showing hood, grille, headlights, and front bumper”
Distance	Expected framing	Wide / Mid / Close-up
Requirement	When this shot is needed	Required / Conditional / Optional
Condition	Logic for conditional shots	airbags_deployed == true
Extract Field	Link to extraction field	VIN, License Plate, etc.

Writing Good AI Descriptions

The AI Description is critical — it’s what Claude uses to determine if a photo matches a shot type.

Good descriptions include:

• What should be visible in the frame
• Relative positioning and framing
• Key identifying features

Example — Front of Vehicle:

Full front view showing hood, grille, headlights, and front bumper. 
Vehicle should fill most of the frame. 
Shot from directly in front, not at an angle.

Example — VIN Plate:

Close-up of the VIN plate on the dashboard (visible through windshield) 
or on the door jamb sticker. All 17 characters must be clearly readable.

Be specific about what makes this shot different from similar shots. For corners, specify which corner and the expected angle (e.g., “45-degree angle showing both front and driver side”).

---

Conditional Shots

Conditional shots are only required when a specific condition is true. This keeps the shot list manageable while ensuring you capture situational documentation.

Condition Syntax

Conditions use the same variable syntax as Decision Triggers:

{variable} == {value}
{variable} != {value}
{variable} > {value}

Common Conditions

Shot	Condition	When Required
Interior / Airbags	airbags_deployed == true	When airbags deployed
Glass Close-up	glass_damaged == true	When glass is damaged
Undercarriage	flood_damage == true	For flood claims
Roof Damage	hail_reported == true	For hail claims

Setting Condition Variables

Condition variables can be set by:

1. Data Extraction in an earlier node
2. API Call that returns claim details
3. User Input captured in the conversation
4. Workflow Variables passed from the trigger

---

Plura Template Library

Plura ships with pre-built taxonomies for common use cases. These can be used as-is or cloned and customized.

Auto Insurance

Taxonomy	Shots	Description
Auto Damage - Standard	12 (9 req, 3 cond)	Standard FNOL documentation
Auto Damage - Total Loss	15 (12 req, 3 cond)	Extended for total loss evaluation
VIN Verification - Quick	4 (4 req)	Quick VIN + plate capture

Homeowners Insurance

Taxonomy	Shots	Description
Water Damage - Standard	14 (10 req, 4 cond)	Water intrusion documentation
Fire Damage - Standard	16 (12 req, 4 cond)	Fire and smoke damage
Roof Damage - Standard	10 (8 req, 2 cond)	Roof condition assessment

Healthcare

Taxonomy	Shots	Description
Insurance Card	2 (2 req)	Front and back of insurance card
Driver’s License	2 (2 req)	Front and back of ID

Logistics

Taxonomy	Shots	Description
Package Intake	6 (4 req, 2 cond)	Shipment receiving
Damage Documentation	8 (6 req, 2 cond)	Damage claims

---

Example: Auto Damage - Standard

Here’s the complete breakdown of the Auto Damage - Standard taxonomy:

Required Shots (9)

Shot ID	Display Name	AI Description	Distance
front_full	Front of Vehicle	Full front view showing hood, grille, headlights, bumper	Wide
rear_full	Rear of Vehicle	Full rear view showing trunk/tailgate, taillights, bumper	Wide
left_side	Left Side	Full driver side from front wheel to rear wheel	Wide
right_side	Right Side	Full passenger side from front wheel to rear wheel	Wide
corner_fl	Front-Left Corner	45° angle showing front and driver side	Wide
corner_fr	Front-Right Corner	45° angle showing front and passenger side	Wide
corner_rl	Rear-Left Corner	45° angle showing rear and driver side	Wide
corner_rr	Rear-Right Corner	45° angle showing rear and passenger side	Wide
damage_closeup	Damage Close-up	Close-up of primary damage area, minimum 2 photos	Close-up

Required with Extraction (3)

Shot ID	Display Name	Extract Field	Validation
vin_plate	VIN Plate	VIN Number	VIN Checksum
license_plate	License Plate	Plate Number	US Plate Format
odometer	Odometer	Mileage	Number (0-500000)

Conditional Shots (3)

Shot ID	Display Name	Condition
interior_airbags	Interior / Airbags	airbags_deployed == true
glass_damage	Glass Close-up	glass_damaged == true
undercarriage	Undercarriage	flood_damage == true

---

Taxonomy Versioning

When you update a taxonomy, existing in-flight conversations continue using the version they started with. New conversations use the updated taxonomy.

Significant taxonomy changes should be tested before deployment. Consider cloning the taxonomy and testing the new version in a staging workflow before updating the production taxonomy.

---

JSON Import/Export

For power users, taxonomies can be exported and imported as JSON.

Export

Click Export JSON on any taxonomy detail view to download the configuration.

Import

Use Create Taxonomy → Import JSON to create a new taxonomy from a JSON file.

JSON Structure

{
  "name": "Auto Damage - Standard",
  "industry": "auto_insurance",
  "icon": "🚗",
  "shots": [
    {
      "id": "front_full",
      "display_name": "Front of Vehicle",
      "ai_description": "Full front view showing hood, grille...",
      "distance": "wide",
      "requirement": "required",
      "extract_field": null
    },
    {
      "id": "vin_plate",
      "display_name": "VIN Plate",
      "ai_description": "Close-up of VIN plate...",
      "distance": "closeup",
      "requirement": "required",
      "extract_field": "vin"
    },
    {
      "id": "interior_airbags",
      "display_name": "Interior / Airbags",
      "ai_description": "Interior shot showing deployed airbags...",
      "distance": "wide",
      "requirement": "conditional",
      "condition": "airbags_deployed == true",
      "extract_field": null
    }
  ]
}

---

Best Practices

Start with Templates

Clone a Plura template and customize rather than building from scratch.

Be Specific in Descriptions

Detailed AI descriptions reduce misclassification and improve coverage accuracy.

Use Conditional Shots

Don’t overload required shots — use conditions to keep the flow manageable.

Link Extraction Fields

Connect shots to extraction when you need OCR data from that specific photo.

---

Next Steps

• Media Processing Node — Use taxonomies in workflows
• Dynamic Data Extraction — Configure extraction fields
• Decision Triggers — Route based on coverage