test(captain): add behavior specs for chunk services

enterprise/app/services/captain/documents/context_generation_service.rb

@@ -3,14 +3,14 @@ class Captain::Documents::ContextGenerationService < Llm::BaseAiService
 
   MAX_DOCUMENT_CHARACTERS = 20_000
   MAX_CHUNK_CHARACTERS = 6_000
-  DEFAULT_MODEL = 'gpt-4.1-mini'.freeze
+  DEFAULT_MODEL = 'gpt-4.1'.freeze
 
-  def initialize(document_content:, chunk_content:, account_id:, model: DEFAULT_MODEL)
+  def initialize(document_content:, chunk_content:, account_id:, model: nil)
     super()
     @document_content = document_content.to_s
     @chunk_content = chunk_content.to_s
     @account_id = account_id
-    @model = model
+    @model = resolve_model(model)
   end
 
   def generate
@@ -74,4 +74,12 @@ class Captain::Documents::ContextGenerationService < Llm::BaseAiService
       ]
     }
   end
+
+  def resolve_model(model)
+    return model if model.present?
+
+    context_model = InstallationConfig.find_by(name: 'CAPTAIN_CONTEXT_GENERATION_MODEL')&.value.presence
+    open_ai_model = InstallationConfig.find_by(name: 'CAPTAIN_OPEN_AI_MODEL')&.value.presence
+    context_model || open_ai_model || DEFAULT_MODEL
+  end
 end

spec/enterprise/services/captain/documents/chunk_embedding_service_spec.rb

@@ -0,0 +1,53 @@
+require 'rails_helper'
+
+RSpec.describe Captain::Documents::ChunkEmbeddingService do
+  let(:embedding_service) { instance_double(Captain::Llm::EmbeddingService) }
+
+  before do
+    allow(Captain::Llm::EmbeddingService).to receive(:new).with(account_id: 7).and_return(embedding_service)
+    allow(embedding_service).to receive(:get_embedding).and_return([0.1, 0.2, 0.3])
+  end
+
+  describe '#embed' do
+    it 'embeds combined context and content when both are present' do
+      service = described_class.new(account_id: 7)
+
+      result = service.embed(content: 'Starts at $19', context: 'Pricing page')
+
+      expect(result).to eq([0.1, 0.2, 0.3])
+      expect(embedding_service).to have_received(:get_embedding).with("Pricing page\n\nStarts at $19")
+    end
+
+    it 'returns empty embedding when both content and context are blank' do
+      service = described_class.new(account_id: 7)
+
+      result = service.embed(content: '', context: nil)
+
+      expect(result).to eq([])
+      expect(embedding_service).not_to have_received(:get_embedding)
+    end
+  end
+
+  describe '#build_record_attributes' do
+    it 'returns attributes ready for persisting a document chunk record' do
+      document = instance_double(Captain::Document, id: 10, assistant_id: 11, account_id: 7)
+      chunk = { position: 3, content: 'Chunk body', token_count: 42 }
+
+      service = described_class.new(account_id: 7)
+      result = service.build_record_attributes(document: document, chunk: chunk, context: 'Chunk context')
+
+      expect(result).to include(
+        document_id: 10,
+        assistant_id: 11,
+        account_id: 7,
+        position: 3,
+        content: 'Chunk body',
+        token_count: 42,
+        context: 'Chunk context',
+        embedding: [0.1, 0.2, 0.3]
+      )
+      expect(result[:created_at]).to be_present
+      expect(result[:updated_at]).to be_present
+    end
+  end
+end

spec/enterprise/services/captain/documents/chunking_service_spec.rb

@@ -0,0 +1,57 @@
+require 'rails_helper'
+
+RSpec.describe Captain::Documents::ChunkingService do
+  describe '#chunk' do
+    it 'returns no chunks for blank content' do
+      result = described_class.new('').chunk
+
+      expect(result).to eq([])
+    end
+
+    it 'returns ordered chunks with position and token count' do
+      content = <<~TEXT
+        # Pricing
+        Starts at $19 per agent.
+
+        ## Limits
+        Includes 100 contacts.
+
+        ## Support
+        Email support is available 24/7.
+      TEXT
+
+      result = described_class.new(
+        content,
+        target_tokens: 8,
+        min_tokens: 4,
+        max_tokens: 10,
+        overlap_tokens: 0
+      ).chunk
+
+      expect(result.size).to be >= 2
+      expect(result.map { |chunk| chunk[:position] }).to eq((0...result.size).to_a)
+      expect(result).to all(include(:content, :token_count, :position))
+      expect(result).to all(satisfy { |chunk| chunk[:token_count].positive? })
+      expect(result.first[:content]).to include('Section: Pricing')
+    end
+
+    it 'adds overlap from previous chunk to the next chunk' do
+      content = <<~TEXT
+        one two three four five six seven eight nine ten
+
+        alpha beta gamma delta epsilon zeta eta theta iota kappa
+      TEXT
+
+      result = described_class.new(
+        content,
+        target_tokens: 5,
+        min_tokens: 3,
+        max_tokens: 5,
+        overlap_tokens: 2
+      ).chunk
+
+      expect(result.size).to eq(2)
+      expect(result.last[:content]).to include('nine ten')
+    end
+  end
+end

spec/enterprise/services/captain/documents/context_generation_service_spec.rb

@@ -0,0 +1,108 @@
+require 'rails_helper'
+require 'ostruct'
+
+RSpec.describe Captain::Documents::ContextGenerationService do
+  let(:chat) { instance_double(RubyLLM::Chat) }
+  let(:response) { instance_double(RubyLLM::Message, content: "  Pricing page context.  \n") }
+
+  before do
+    InstallationConfig.find_or_initialize_by(name: 'CAPTAIN_OPEN_AI_API_KEY').update!(value: 'test-key')
+    InstallationConfig.where(name: %w[CAPTAIN_CONTEXT_GENERATION_MODEL CAPTAIN_OPEN_AI_MODEL]).delete_all
+
+    allow(RubyLLM).to receive(:chat).and_return(chat)
+    allow(chat).to receive(:with_temperature).and_return(chat)
+    allow(chat).to receive(:with_instructions).and_return(chat)
+    allow(chat).to receive(:ask).and_return(response)
+  end
+
+  describe '#generate' do
+    it 'returns stripped context text from the model response' do
+      service = described_class.new(
+        document_content: 'Pricing docs for all plans',
+        chunk_content: 'Business plan starts at $19/agent/month.',
+        account_id: 1
+      )
+
+      result = service.generate
+
+      expect(result).to eq('Pricing page context.')
+    end
+
+    it 'uses explicit model when provided' do
+      service = described_class.new(
+        document_content: 'Doc text',
+        chunk_content: 'Chunk text',
+        account_id: 1,
+        model: 'gpt-4.1-mini'
+      )
+
+      expect(RubyLLM).to receive(:chat).with(model: 'gpt-4.1-mini').and_return(chat)
+
+      service.generate
+    end
+
+    it 'uses CAPTAIN_CONTEXT_GENERATION_MODEL when configured' do
+      create(:installation_config, name: 'CAPTAIN_CONTEXT_GENERATION_MODEL', value: 'gpt-4.1-mini')
+
+      service = described_class.new(
+        document_content: 'Doc text',
+        chunk_content: 'Chunk text',
+        account_id: 1
+      )
+
+      expect(RubyLLM).to receive(:chat).with(model: 'gpt-4.1-mini').and_return(chat)
+
+      service.generate
+    end
+
+    it 'uses CAPTAIN_OPEN_AI_MODEL when context model is not configured' do
+      create(:installation_config, name: 'CAPTAIN_OPEN_AI_MODEL', value: 'gpt-4o-mini')
+
+      service = described_class.new(
+        document_content: 'Doc text',
+        chunk_content: 'Chunk text',
+        account_id: 1
+      )
+
+      expect(RubyLLM).to receive(:chat).with(model: 'gpt-4o-mini').and_return(chat)
+
+      service.generate
+    end
+
+    it 'falls back to the service default model when no model config exists' do
+      service = described_class.new(
+        document_content: 'Doc text',
+        chunk_content: 'Chunk text',
+        account_id: 1
+      )
+
+      expect(RubyLLM).to receive(:chat).with(model: 'gpt-4.1').and_return(chat)
+
+      service.generate
+    end
+
+    it 'returns empty string when chunk content is blank' do
+      service = described_class.new(
+        document_content: 'Doc text',
+        chunk_content: '',
+        account_id: 1
+      )
+
+      expect(service.generate).to eq('')
+      expect(RubyLLM).not_to have_received(:chat)
+    end
+
+    it 'returns empty string when LLM call fails' do
+      error = RubyLLM::Error.new(OpenStruct.new(body: {}), 'request failed')
+      allow(chat).to receive(:ask).and_raise(error)
+
+      service = described_class.new(
+        document_content: 'Doc text',
+        chunk_content: 'Chunk text',
+        account_id: 1
+      )
+
+      expect(service.generate).to eq('')
+    end
+  end
+end