adm_ai_rag_collection_files_v

Detailed view of the files in RAG collections with document information, chunk statistics, extraction method, latest queue job and indexing status per file. AI Pack.

Overview

Group: Administrative reporting

Category: RAG collections - Monitoring of RAG collections: files, chunks, queue and search log

Use Cases

See which files of a collection are indexed, pending or failed
Find the error that kept a file out of the index
Check which reader produced the text of a file

Columns

The view contains 26 columns providing comprehensive information:

Column Name	Description
`rag_collection_file_id`	Primary key of the RAG collection file link.
`rag_collection_id`	Foreign key to the RAG collection.
`collection_name`	Name of the RAG collection.
`is_active`	Y if the file is currently active in the collection, N if inactive.
`version_id`	Foreign key to the document version.
`document_id`	Foreign key to the document.
`document_name`	Name of the document.
`folder_path`	Path of the folder containing the document.
`file_path`	Full path to the file (folder path + document name).
`file_mime_type`	MIME type of the document.
`version_number`	Version number of the document.
`file_size`	Size of the file in bytes.
`version_created_date`	Timestamp when the document version was created.
`version_created_by`	User who created the document version.
`added_to_collection_date`	Timestamp when the file was added to the RAG collection.
`added_to_collection_by`	User who added the file to the RAG collection.
`chunk_count`	Total number of chunks for this file.
`embedded_chunk_count`	Number of chunks that have been embedded.
`pending_chunk_count`	Number of chunks waiting to be embedded.
`latest_job_status`	Status of the most recent job for this file.
`latest_job_type`	Type of the most recent job: EMBEDDING or REMOVE.
`latest_error_message`	Why the file failed: the error message of the most recent FAILED queue job, otherwise the chunk_error or extract_error stored on the file. Can be set while the file is still being retried (status is not ERROR until the attempt budget is used up).
`extraction_method`	Which reader produced the text that is indexed for this version: FILTER (built-in Oracle Text reader), LLM, OCR, or UNKNOWN (text from before 26.2, which has no source). Null while the version has no text yet. The text belongs to the version, not to the collection: the first collection that indexes a version decides it.
`requested_extraction_method`	The reader this collection asks for this file: LLM or OCR if its llm_text_extraction or ocr_text_extraction section applies (mime type in apply_to, and the annotation if one is set), otherwise FILTER.
`extraction_mismatch`	Y if the file has text from another reader than the collection asks for (extraction_method differs from requested_extraction_method). N if they match, there is no text yet, or the source is unknown. The text is kept: other collections may already point into it by offset.
`file_status`	Overall file status, first match wins: INACTIVE (not part of the collection any more); ERROR (a FAILED queue job, or text extraction or chunking has used up its attempt budget); PROCESSING (a job is running); PENDING (a job is waiting); INCOMPLETE (some chunks are not embedded); EMPTY (the file has no text, so no chunks; wins over the attempt budget but not over a FAILED job); NOT_INDEXED (no chunks yet, text extraction or chunking has still to run); INDEXED (chunks exist and all are embedded).

Description: Detailed view of files in RAG collections including document information, chunk statistics, and indexing status for each file.

Key Features

One row per file version in a collection
Chunk statistics and a file status
Latest job status, type and error message
Extraction method stored and extraction method requested