Add openai integration for analysing documents

Author: benbaumann95

Diff for: Gemfile

@@ -44,6 +44,7 @@ gem "rgeo"
 gem "rgeo-geojson"
 gem "rswag-api"
 gem "rswag-ui"
+gem "ruby-openai"
 gem "sidekiq"
 gem "sidekiq-scheduler"
 gem "sprockets-rails"

Diff for: Gemfile.lock

@@ -259,6 +259,7 @@ GEM
     erubi (1.13.1)
     et-orbi (1.2.11)
       tzinfo
+    event_stream_parser (1.0.0)
     factory_bot (6.4.6)
       activesupport (>= 5.0.0)
     factory_bot_rails (6.4.3)
@@ -268,6 +269,8 @@ GEM
       i18n (>= 1.6, < 2)
     faraday (2.9.0)
       faraday-net_http (>= 2.0, < 3.2)
+    faraday-multipart (1.1.0)
+      multipart-post (~> 2.0)
     faraday-net_http (3.1.0)
       net-http
     ffi (1.17.1)
@@ -390,6 +393,7 @@ GEM
     minitest (5.25.4)
     msgpack (1.7.2)
     multi_test (1.1.0)
+    multipart-post (2.4.1)
     nenv (0.3.0)
     net-http (0.4.1)
       uri
@@ -568,6 +572,10 @@ GEM
       rack (>= 1.1)
       rubocop (>= 1.33.0, < 2.0)
       rubocop-ast (>= 1.31.1, < 2.0)
+    ruby-openai (8.0.0)
+      event_stream_parser (>= 0.3.0, < 2.0.0)
+      faraday (>= 1)
+      faraday-multipart (>= 1)
     ruby-progressbar (1.13.0)
     ruby-vips (2.1.4)
       ffi (~> 1.12)
@@ -731,6 +739,7 @@ DEPENDENCIES
   rswag-api
   rswag-specs
   rswag-ui
+  ruby-openai
   selenium-webdriver
   sidekiq
   sidekiq-scheduler

Diff for: app/controllers/documents_controller.rb

@@ -38,7 +38,7 @@ def create
     @document = @planning_application.documents.build(document_params)
 
     if @document.save
-      flash[:notice] = "#{@document.name} has been uploaded."
+      flash[:notice] = "#{@document.name} has been uploaded and queued for analysis."
       redirect_to planning_application_documents_path
     else
       render :new

Diff for: app/jobs/document_analyser_job.rb

@@ -0,0 +1,8 @@
+class DocumentAnalyserJob < ApplicationJob
+  queue_as :high_priority
+  retry_on(StandardError, attempts: 5, wait: 1.minute, jitter: 0)
+
+  def perform(document)
+    DocumentAnalyserService.new(document).call
+  end
+end

Diff for: app/models/document.rb

@@ -56,6 +56,8 @@ class NotArchiveableError < StandardError; end
   before_update :reset_replacement_document_validation_request_update_counter!, if: :owner_is_validation_request?
   after_update :audit_updated!
 
+  after_create :generate_ai_summary
+
   DRAWING_TAGS = %w[
     elevations.existing
     elevations.proposed
@@ -284,6 +286,11 @@ def tags(key)
     end
   end
 
+  def generate_ai_summary
+    # Call the open ai service
+    DocumentAnalyserJob.perform_later(self)
+  end
+
   def name
     file.filename.to_s if file.attached?
   end

Diff for: app/services/document_analyser_service.rb

@@ -0,0 +1,133 @@
+require "openai"
+require "base64"
+
+class DocumentAnalyserService
+  attr_reader :document, :client
+
+  def initialize(document)
+    @document = document
+    @client = OpenAI::Client.new(
+      access_token: "XXX",
+      request_timeout: 120
+    )
+  end
+
+  def call
+    return unless document.file.attached?
+
+    file_path = download_file
+    return unless file_path
+
+    image = build_image(file_path)
+    user_content = user_message_tags(image)
+
+    ai_summary = generate_ai_summary(user_content)
+
+    document_tags = generate_document_tags(user_content)
+  
+    document.update!(tags: document_tags) if document_tags.any?
+    
+    document.update!(ai_summary: ai_summary) if ai_summary.present?
+  rescue StandardError => e
+    Rails.logger.error("Error analysing the document with id: #{document.id}: #{e.message}")
+    false
+  end
+
+  private
+
+  def download_file
+    # Create a temporary file with the original filename and extension
+    temp_file = Tempfile.new([document.file.filename.base, document.file.filename.extension_with_delimiter], binmode: true)
+
+    # Download the file locally in binary mode
+    begin
+      document.file.blob.download do |chunk|
+        temp_file.write(chunk.force_encoding("BINARY"))
+      end
+    rescue => e
+      Rails.logger.error "Failed to download file: #{e.message}"
+      return nil
+    end
+
+    temp_file.close
+    temp_file.path
+  end
+
+  def build_image(file_path)
+    image_data = File.binread(file_path)
+    base64_image = Base64.strict_encode64(image_data)
+  end
+
+  def generate_ai_summary(user_content)    
+    response = client.chat(
+      parameters: {
+        model: "gpt-4o-mini",
+        messages: [
+          { 
+            role: "system", 
+            content: "You are an expert, with experience in UK housing/planning, in interpreting and extracting a concise summary/description of a document" 
+          },
+          {
+            role: "user",
+            content: user_content
+          }
+        ]
+      }
+    )
+
+    response.dig("choices", 0, "message", "content")
+  end
+
+  def user_message(base64_image)
+    [
+      {
+        type: "text",
+        text: "Analyse this image/document and provide a short description/summary of its content in no more than 2 sentences."    
+      },
+      {
+        type: "image_url",
+        image_url: { 
+          url: "data:image/jpeg;base64,#{base64_image}"
+        }
+      }
+    ]
+  end
+
+  def generate_document_tags(user_content)
+    response = client.chat(
+      parameters: {
+        model: "gpt-4o-mini",
+        messages: [
+          { 
+            role: "system", 
+            content: "You are an AI expert in reviewing and tagging documents with expertise in the UK housing/planning system. When tagging documents use ONLY the following predefined options:
+            - **DRAWING TAGS**: #{Document::DRAWING_TAGS.join(', ')}
+            - **EVIDENCE TAGS**: #{Document::EVIDENCE_TAGS.join(', ')}
+            - **SUPPORTING DOCUMENT TAGS**: #{Document::SUPPORTING_DOCUMENT_TAGS.join(', ')}" 
+          },
+          {
+            role: "user",
+            content: user_content
+          }
+        ]
+      }
+    )
+    response.dig("choices", 0, "message", "content").split(", ").map(&:strip)
+  end
+
+  def user_message_tags(base64_image)
+    [
+      {
+        type: "text",
+        text: "Analyse this image/document and assign relevant document tags from the supplied lists. 
+        ONLY choose tags from these lists. You may choose more than one tag if relevant. Provide a comma separated list of ONLY the tags."    
+      },
+      {
+        type: "image_url",
+        image_url: { 
+          url: "data:image/jpeg;base64,#{base64_image}"
+        }
+      }
+    ]
+  end
+end

Diff for: app/views/documents/edit.html.erb

@@ -73,6 +73,12 @@
     <p class="govuk-body">
       <%= created_by(@document) %>
     </p>
+
+    <p class="govuk-body govuk-!-margin-bottom-0">
+      <strong>AI summary:</strong> <%= @document.ai_summary %>
+    </p>
+    
+
   </div>
   <div class="govuk-grid-column-full">
     <%= render "edit_and_upload" %>

Diff for: db/migrate/20250319102739_add_ai_summary_to_documents.rb

@@ -0,0 +1,5 @@
+class AddAiSummaryToDocuments < ActiveRecord::Migration[7.2]
+  def change
+    add_column :documents, :ai_summary, :text
+  end
+end

Diff for: db/schema.rb

@@ -10,7 +10,7 @@
 #
 # It's strongly recommended that you check this file into your version control system.
 
-ActiveRecord::Schema[7.2].define(version: 2025_03_17_171204) do
+ActiveRecord::Schema[7.2].define(version: 2025_03_19_102739) do
   # These are extensions that must be enabled in order to support this database
   enable_extension "btree_gin"
   enable_extension "plpgsql"
@@ -402,6 +402,7 @@
     t.bigint "owner_id"
     t.string "tags", default: [], array: true
     t.bigint "document_checklist_items_id"
+    t.text "ai_summary"
     t.index ["api_user_id"], name: "ix_documents_on_api_user_id"
     t.index ["document_checklist_items_id"], name: "ix_documents_on_document_checklist_items_id"
     t.index ["evidence_group_id"], name: "ix_documents_on_evidence_group_id"