계승 v 2.1.1

[GulSauce]

📢 설명

해당 Pull Request에 대해 간략하게 설명해주세요!

✅ 체크 리스트

• 리뷰어가 체크할 내용을 작성해주세요!

Summary by CodeRabbit

• New Features

  • Quiz generation endpoint now streams results incrementally, improving responsiveness for large documents.
  • Standardized error response format with detailed status codes and messages.

• Improvements

  • Concurrent chunk processing accelerates quiz generation performance.
  • Optimized API request timeout handling with reduced latency.
  • Enhanced PDF chunking strategy with improved page allocation and lookahead optimization.

_{✏️ Tip: You can customize this high-level summary in your review settings.}

[coderabbitai]

Caution

Review failed

The pull request is closed.

📝 Walkthrough

Walkthrough

Consolidates OpenAI client management by removing duplicate singleton implementations, eliminates batch processing wrapper, refactors quiz generation to support streaming responses with concurrent chunk processing, and introduces standardized error response models.

Changes

Cohort / File(s)	Summary
GPT Client & Adapter Consolidation app/adapter/request_batch.py, app/client/oepn_ai.py, app/adapter/request_to_gpt.py	Removed batch processing module and duplicate singleton client. Replaced with cached get_gpt_client() returning AsyncOpenAI and new request_to_gpt_returning_text() function that applies per-call timeout via OpenAI's with_options method instead of external timeout handling.
Response Data Models app/dto/response/error_response.py, app/dto/response/generate_response.py, app/dto/model/problem_set.py	Added new ErrorResponse DTO for standardized error payloads. Refactored GenerateResponse to inherit from new ProblemSetDTO base, renamed ProblemResponse to ProblemDTO, added type: Literal["quiz"] field. Minor formatting adjustment to Selection.correct description.
Service Layer Refactor app/service/generate_service.py, app/service/explanation_service.py	Major streaming refactor in GenerateService.generate: introduced async chunk processing, concurrent task creation, per-chunk streaming results as JSONL, and error response streaming. Added process_single_chunk() helper. Migrated to new GPT function, reduced explanation timeout from 60 to 30 seconds.
Chunking & Utility Updates app/util/create_chunks.py, app/util/gpt_utils.py	Simplified chunk distribution logic in create_chunks.py: replaced specialized handlers with unified approach accepting page_numbers list, added gpt_content field to ChunkInfo. Added new enforce_additional_properties_false() utility to recursively enforce schema constraints for GPT responses.
Router & Endpoint Changes app/router/generate_router.py	Removed FastAPI dependency injection; endpoints now call service methods directly as static/class methods. generate endpoint now returns StreamingResponse wrapping JSONL output instead of single GenerateResponse object.

Sequence Diagram

sequenceDiagram
    participant Client as Client
    participant Router as Router Handler
    participant Service as GenerateService
    participant Chunk as process_single_chunk
    participant GPT as GPT Client
    participant Stream as StreamingResponse

    Client->>Router: POST /generate (GenerateRequest)
    Router->>Service: GenerateService.generate(request)
    
    Service->>Service: Create page chunks<br/>with gpt_content

    par Concurrent Chunk Processing
        Service->>Chunk: Task 1: process_single_chunk()
        Chunk->>GPT: request_to_gpt_returning_text()
        GPT-->>Chunk: Response (text)
        Chunk->>Chunk: Parse & validate response
        Chunk-->>Service: ProblemSetDTO
        
        Service->>Chunk: Task N: process_single_chunk()
        Chunk->>GPT: request_to_gpt_returning_text()
        GPT-->>Chunk: Response (text)
        Chunk-->>Service: ProblemSetDTO
    end

    Service->>Service: Stream results as JSONL<br/>on task completion
    Service-->>Stream: Emit ProblemSetDTO lines
    alt Task Failure
        Service->>Stream: Emit ErrorResponse line
    end
    Stream-->>Router: JSONL Stream
    Router-->>Client: StreamingResponse

Loading

Estimated code review effort

🎯 4 (Complex) | ⏱️ ~45 minutes

Possibly related PRs

• [ICC-237] OCR로 변경 완료 #106: Adds request_responses_output_text and request_responses_batch functions to app/adapter/request_batch.py — directly inverse to this PR's removal of the same module and functions.
• [ICC-241] 비동기 퀴즈 응답 #112: Modifies the same adapter and client code paths, removing app/client/oepn_ai.py and refactoring app/adapter/request_to_gpt.py with cached get_gpt_client and new request function.
• 계승: v 1.5.3 (#87) #88: Updates batch request timeout handling and singleton OpenAI client implementation in the same adapter/client layers affected by this PR.

Poem

🐰 Hop-hop, streaming quiz chunks through the air!
Concurrent tasks dance without a care,
No more batches—just nimble flows,
Error schemas locked down in a row,
The client's cached, the chunks are neat,
Generation's refactor—a hoppy treat! ✨

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch main

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}