feat(api): add server-side PDF text extraction endpoint #2014
Conversation
Add POST /extract/pdf endpoint to handle PDF text extraction on the server. This allows the React SDK to offload PDF processing, reducing client bundle size by ~2MB (removal of pdfjs-dist dependency). Changes: - Add pdf-parse dependency (lightweight, 25KB) - Create ExtractPdfDto and ExtractPdfResponseDto following NestJS patterns - Add extractPdfText endpoint in ExtractorController with file validation - Add unit tests for PDF file validation (size/type checks) - Follow existing audio.controller.ts pattern for consistency 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
Resolve package-lock.json conflicts by regenerating from package.json. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
…tion feat(api): add server-side PDF text extraction endpoint
|
@BurhanCantCode is attempting to deploy a commit to the tambo ai Team on Vercel. A member of the Team first needs to authorize it. |
|
Hey @BurhanCantCode! Thanks for the PR. I reviewed the server-side PDF extraction implementation. After discussing internally, we do need to update our Tambo Cloud API, but we don't need any new API routes. Instead, we need to extend our current image handling to support more file types, using S3 instead of storing in the database. Many of the Frontier models just support passing file types. File types to support (from OpenAI, Anthropic, and Gemini): Common across all three providers:
Additional formats:
Implementation approach:
No need for pdf-parse or any extraction logic - the LLM providers handle that themselves. Here are the relevant docs for supported file types: Could you refactor this to follow the S3 upload pattern instead of adding a new extraction endpoint? |
Summary
Adds a new
POST /extract/pdfendpoint to handle PDF text extraction on the server-side. This enables the React SDK to offload PDF processing to the backend, reducing client bundle size by ~2MB (removal of pdfjs-dist dependency).Changes
pdf-parsedependency (lightweight, 25KB vs 2MB pdfjs-dist)ExtractPdfDtoandExtractPdfResponseDtofollowing NestJS patternsextractPdfTextendpoint inExtractorControllerwith file validationaudio.controller.tspattern for consistencyAPI Endpoint
POST /extract/pdf Content-Type: multipart/form-data Request:
file: PDF file (max 50MB)
Response: { "text": "extracted text content...", "pages": 5 }
Testing
Related
This follows:
✅ Conventional Commits format (feat(api):)
✅ Clear summary of changes
✅ Technical details about the implementation
✅ Testing information
✅ Links to related PRs