[ICC-237] OCR로 변경 완료

[GulSauce]

📢 설명

해당 Pull Request에 대해 간략하게 설명해주세요!

✅ 체크 리스트

• 리뷰어가 체크할 내용을 작성해주세요!

Summary by CodeRabbit

Release Notes

• New Features

  • Added PDF support with URL fetching and page selection capabilities
  • Enhanced quiz generation now embeds PDF content directly into prompts

• Improvements

  • Optimized batch processing pipeline with improved error and timeout handling
  • More robust API response extraction and parsing logic

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

[coderabbitai]

Caution

Review failed

The pull request is closed.

📝 Walkthrough

Walkthrough

This change refactors the request batching layer to use a new Responses API text extraction flow, restructures chunk generation to work with page counts instead of page lists with embedded text, and enhances the service layer with PDF fetching, page selection, and embedded PDF context in GPT prompts.

Changes

Cohort / File(s)	Summary
Request batch adapter app/adapter/request_batch.py	Replaced request_chat_completion_text with two new functions: request_responses_output_text (extracts text from Responses API output with fallback logic) and request_responses_batch (orchestrates batch requests with timeout handling and per-task exception recovery). Logic now traverses nested response structures to locate message dictionaries and extract text fields.
Service layer integration app/service/generate_service.py	Added PDF fetching, page count validation, and page selection logic. Introduces per-chunk PDF extraction and base64 embedding into GPT prompts. Switches to request_responses_batch for API calls and alters JSON output from response_format to text-based schema. Includes chunk expansion to ensure max 2 quizzes per chunk. Adds per-chunk PDF cache to avoid redundant extraction.
Chunk utility restructuring app/util/create_chunks.py	Renamed create_chunks to create_page_chunks with parameters changed from (pages: List[str], total_quiz_count, minimum_page_text_length_per_chunk, max_chunk_count) to (page_count: int, total_quiz_count, max_chunk_count). ChunkInfo model now contains only referenced_pages: List[int] and quiz_count: int (text field removed). Helper function signatures updated to accept page_count directly. Removed add_prefix_suffix function and text-based concatenation logic.

Sequence Diagram(s)

sequenceDiagram
    participant Client as Client/Service
    participant Adapter as Request Adapter
    participant OpenAI as OpenAI API
    participant Parser as Response Parser

    Client->>Adapter: request_responses_batch(requests, timeout)
    Adapter->>Adapter: Start async batch with wait_for
    loop For each request
        Adapter->>OpenAI: Submit GPT request
        OpenAI-->>Adapter: Return Responses object
        Adapter->>Parser: request_responses_output_text(response)
        Parser->>Parser: Try resp.output_text
        alt Found output_text
            Parser-->>Adapter: Return extracted text
        else Search resp.output list
            Parser->>Parser: Find dict with type="message"
            Parser->>Parser: Extract from text/output_text/value keys
            Parser-->>Adapter: Return first non-empty string or empty
        end
    end
    Adapter->>Adapter: Collect results (completed + cancelled tasks)
    Adapter-->>Client: Return List[Optional[str]]

Loading

sequenceDiagram
    participant Service as Generate Service
    participant PDF as PDF Fetcher
    participant Chunks as Chunk Creator
    participant Cache as PDF Cache
    participant Adapter as Request Adapter
    participant GPT as OpenAI API

    Service->>PDF: Fetch PDF from URL
    PDF-->>Service: PDF content + page_count
    Service->>Chunks: create_page_chunks(page_count, quiz_count, max_chunks)
    Chunks-->>Service: List[ChunkInfo] with referenced_pages
    Service->>Service: Expand chunks (ensure ≤2 quizzes per chunk)
    loop For each chunk
        Service->>Cache: Check if pages cached
        alt Not cached
            Service->>PDF: Extract base64 for referenced_pages
            PDF-->>Cache: Store extracted PDF
        else Cached
            Cache-->>Service: Return cached PDF
        end
        Service->>Service: Build prompt with embedded PDF + page hints
        Service->>Adapter: Add to batch request with PDF context
    end
    Service->>Adapter: request_responses_batch(all_requests, timeout)
    Adapter->>GPT: Submit batch with PDF-embedded prompts
    GPT-->>Adapter: Return responses
    Adapter->>Service: Return extracted text results
    Service->>Service: Parse JSON and map to chunks

Loading

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Poem

🐰 From batches old to responses new,
PDFs embedded in chunks of few,
Page counts dancing where text once stood,
The code hops faster—oh, so good!
hops twitch 📄✨

✨ Finishing touches

• 📝 Generate docstrings

---

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.}