Create a new DocumentProcessor.
The processor dependencies.
Delete all chunks for a file from the vector store and remove metadata.
The file to delete.
Move a file's vector points from old path to new path without re-embedding. Re-applies inference rules against the new path.
The original file path.
The new file path.
Process a file through the full pipeline: extract, hash, chunk, embed, upsert.
The file to process.
Process a metadata update: merge into enrichment store, update Qdrant
payloads (no re-embed). The stored file_path (casing included) and
matched_rules are kept.
The file whose metadata to update.
The new metadata to merge.
The merged payload, or null if the file is not indexed.
Re-apply inference rules to a file without re-embedding (see
reapplyFileRules: system fields and the stored file_path are
re-asserted; interpolation failures are recorded as an issue).
The file to update.
The merged metadata, or null if the file is not indexed.
Render a file through the rule engine without embedding. Returns rendered content, renderAs, matched rules, and metadata.
The file to render.
The render result.
Source operations bound to the processor's current rules and stores.
Update compiled inference rules, template engine, and custom map lib.
The newly compiled rules.
OptionaltemplateEngine: TemplateEngine
Optional updated template engine.
OptionalcustomMapLib: Record<string, (...args: unknown[]) => unknown>
Optional updated custom JsonMap lib functions.
Core document processing pipeline.
Handles extracting text, computing embeddings, and syncing with the vector store.