Sector Classifier

API-first Rails app that classifies free-text questions into industries per project/tenant. Includes:

• Projects with rate limits and API keys.
• Shared industry catalog and per-project associations.
• Classification endpoint with logging.
• Admin JSON endpoints secured with optional HTTP basic auth (set ADMIN_USER / ADMIN_PASSWORD).

Setup

git clone <repo>
cd sector-classifier
DISABLE_SPRING=1 BUNDLE_USER_HOME=./.bundle bundle install
DISABLE_SPRING=1 BUNDLE_USER_HOME=./.bundle rails db:create db:migrate db:seed
DISABLE_SPRING=1 BUNDLE_USER_HOME=./.bundle rails server

Postgres config is in config/database.yml (env vars: PGUSER, PGPASSWORD, PGHOST, PGPORT).

API

• Health: GET /api/v1/health
• Classify: POST /api/v1/classify with header X-Api-Key and body { "text": "I need a hotel booking system" }
  • Returns matched industry { id, name, code, description, score }
  • Logs every call in classification_requests.

Admin JSON (optional HTTP Basic with ADMIN_USER/ADMIN_PASSWORD)

• Projects: GET/POST/PATCH/DELETE /admin/projects
• API keys: GET/POST/PATCH/DELETE /admin/projects/:project_id/api_keys
• Industries catalog: GET/POST/PATCH/DELETE /admin/industries
• Project industries: GET/POST/DELETE /admin/projects/:project_id/industries

Seed data

• Project: Leadhub (status: active)
• Industries: Construction, Healthcare, IT, Hospitality, Real Estate
• Default API key attached to Leadhub (token printed in DB, generated automatically)

Bulk import & embeddings

• Fast import (CSV or JSON) to a project: DISABLE_SPRING=1 BUNDLE_USER_HOME=./.bundle rails import:industries[PROJECT_ID,/absolute/path/to/file.csv]
  • The file must have name,code,description.
  • Imports also link the industries to the given project; names must be unique per project (case-insensitive).
• Rebuild embeddings (all or per project): DISABLE_SPRING=1 BUNDLE_USER_HOME=./.bundle rails embeddings:rebuild[PROJECT_ID]

Import file formats (share with clients)

• CSV headers: name,code,description,external_id,search_terms,category
  • external_id recommended if you want to round-trip the source ID.
  • search_terms can be JSON array or plain text; stored as array.
  • Example:

    name,code,description,external_id,search_terms,category
    "Hotel Services","HOTEL","Hotel and accommodation","49213","[""hotel"",""lodging""]","Hospitality"
    "AI Consulting","AI","AI & data projects","AI-001","[""machine learning"",""ml""]","Technology"
    

• JSON array; camelCase accepted:

  [
    {
      "name": "Hotel Services",
      "code": "HOTEL",
      "description": "Hotel and accommodation",
      "industryId": "49213",
      "searchTerms": ["hotel", "lodging"],
      "category": "Hospitality"
    },
    {
      "name": "AI Consulting",
      "code": "AI",
      "description": "AI & data projects",
      "industryId": "AI-001",
      "searchTerms": ["machine learning", "ml"],
      "category": "Technology"
    }
  ]

API response shape

• POST /api/v1/classify and POST /api/v1/semantic_search return industry fields: id, name, code, description, external_id, category, search_terms (plus score).