Files

LiteLLM routing - The LLM orchestration layer that selects providers, builds messages, calls the API client, parses responses, and logs payloads.
Services - Service-layer map for upload orchestration, PDF/OCR processing, LiteLLM routing, storage backends, parallel execution, PII masking, and related helpers.
Parallel processing - Parallel PDF page execution, page classification, and database coordination used when the extraction pipeline splits work across pages.
PDF processing - PDF and OCR pipeline, page classification, confidence handling, mixed-content branching, and persistence hooks used by document extraction.
PII masking - Text and image masking helpers, Presidio integration, and upload logging support for privacy-sensitive document processing.
Prompt generation - Prompt-generation helpers and the boundary between schema-aware prompts, user prompts, and downstream LLM execution.
Storage - Storage-provider selection, cached backend instances, validation of provider-specific configuration, and the supported file backends.
Upload processing - Background document-processing orchestration, validation, strategy selection, repository access, task behavior, and response building for queued uploads.