27 lines
725 B
Markdown
27 lines
725 B
Markdown
|
```r title="R"
|
||
|
|
library(kreuzberg)
|
||
|
|
|
||
|
|
document_id <- "doc-001"
|
||
|
|
|
||
|
|
config <- list(
|
||
|
|
chunking = list(
|
||
|
|
max_characters = 512L,
|
||
|
|
overlap = 50L,
|
||
|
|
embedding = list(
|
||
|
|
model = list(type = "preset", name = "balanced"),
|
||
|
|
normalize = TRUE,
|
||
|
|
batch_size = 32L
|
||
|
|
)
|
||
|
|
)
|
||
|
|
)
|
||
|
|
|
||
|
|
json <- extract_file_sync("document.pdf", "application/pdf", config)
|
||
|
|
result <- jsonlite::fromJSON(json, simplifyVector = FALSE)
|
||
|
|
|
||
|
|
# Each chunk has $content, $embedding, and $metadata. Pass these directly
|
||
|
|
# to a vector database client (pgvector, Qdrant, Pinecone, etc.) along with
|
||
|
|
# the document_id stored as a metadata field.
|
||
|
|
cat(sprintf("document_id: %s\n", document_id))
|
||
|
|
cat(sprintf("chunks ready for upsert: %d\n", length(result$chunks)))
|
||
|
|
```
|