AIService

Kind: Service

Source: atloria-monorepo/apps/api/src/ai/ai.service.ts

AIService is a NestJS backend service that centralizes access to configured AI providers and document-generation capabilities. It resolves the active provider, exposes provider discovery, and supports document enhancement, outlining, writing improvement, summarization, text generation, and streaming completions.

Methods

Dependencies

• PrismaService
• ConfigService
• AzureClaudeProvider
• AzureOpenAIProvider
• AzureResponsesProvider
• AzureResponsesProvider

Where it refuses work

• AIService stops the work with NotFoundException when !document — “Document not found”, in 2 places.
• AIService stops the work with NotFoundException when !provider.

Diagram

mermaid
sequenceDiagram
    participant Client as API Controller
    participant Service as AIService
    participant Provider as AIProvider
    participant Model as AI Model API

    Client->>Service: generateOutline(document)
    Service->>Service: getProvider()
    Service->>Provider: generateText(prompt/options)
    Provider->>Model: completion request
    Model-->>Provider: generated text
    Provider-->>Service: outline result
    Service-->>Client: Promise<string>

    Client->>Service: streamCompletion(prompt)
    Service->>Provider: streamCompletion(prompt)
    Provider-->>Service: AsyncIterableIterator<string>
    Service-->>Client: stream tokens

Usage

ts
import { Injectable } from '@nestjs/common';
import { AIService } from './ai.service';

@Injectable()
export class DocumentService {
  constructor(private readonly aiService: AIService) {}

  async createOutline(content: string): Promise<string> {
    return this.aiService.generateOutline(content);
  }

  async improveContent(content: string): Promise<string> {
    return this.aiService.improveWriting(content);
  }

  async streamResponse(prompt: string): Promise<void> {
    for await (const chunk of this.aiService.streamCompletion(prompt)) {
      process.stdout.write(chunk);
    }
  }

  getAvailableProviders() {
    return this.aiService.listProviders();
  }
}

AI Coding Instructions

• Use AIService as the application-level integration point rather than calling an AIProvider directly from controllers or feature services.
• Call getProvider() only when provider-specific behavior is necessary; prefer high-level methods such as summarizeDocument() or improveWriting() for document workflows.
• Preserve streaming behavior by consuming streamCompletion() with for await...of; do not eagerly collect chunks unless a complete response is required.
• Handle provider failures, missing configuration, and malformed AI output at the controller or calling-service boundary.
• Add new AI capabilities as focused service methods that delegate to the active provider and return consistently typed results.

Relationships

• DEPENDS_ON → PrismaService
• DEPENDS_ON → configservice
• DEPENDS_ON → AzureClaudeProvider
• DEPENDS_ON → AzureOpenAIProvider
• DEPENDS_ON → azureresponsesprovider
• DEPENDS_ON → azureresponsesprovider

Referenced By

• AIController (DEPENDS_ON)
• AIModule (MODULE_PROVIDES)
• AIModule (MODULE_EXPORTS)
• CodeAnalysisService (DEPENDS_ON)
• PlaywrightGeneratorService (DEPENDS_ON)
• UIAnalyzerService (DEPENDS_ON)
• EntityExtractorService (DEPENDS_ON)
• PublicProjectController (DEPENDS_ON)