Automatically detects document type and chooses best extraction method:
Academic papers → Nougat (equations)
Documents with visuals → Multimodal extraction
General documents → Marker
Usage
extract_pdf_smart(
file_path,
doc_type = "auto",
vision_provider = "ollama",
vision_model = NULL,
api_key = NULL,
envname = "langgraph-env"
)Arguments
- file_path
Character string path to PDF file
- doc_type
Character: "auto" (default), "academic", or "general"
- vision_provider
Character: "ollama" (default) or "openai"
- vision_model
Character: Model name for vision analysis
- api_key
Character: API key for cloud providers
- envname
Character: Python environment name
Examples
if (FALSE) { # \dontrun{
# Auto-detect and extract
result <- extract_pdf_smart("document.pdf")
# Feed to text analysis
corpus <- prep_texts(result$combined_text)
topics <- fit_semantic_model(corpus, k = 10)
# Force academic extraction (with equations)
result <- extract_pdf_smart("paper.pdf", doc_type = "academic")
} # }