RAGatm Open

Document formats

Every format identified in the Open edition code, with its actual status in the Compact profile as published. A format accepted at upload is not necessarily extracted; an extractor present is not necessarily enabled.

Scope: ragatm-open repository, main branch, verified on 11 November 2026. Statuses describe the published Compact profile; the "capabilities" columns describe what the extractor produces when enabled.

Four gates, not one

A file must pass four independent steps before it can feed an answer.

  1. Recognition at upload

    The UI offers the extensions announced by the server (capabilities API endpoint).

  2. Backend acceptance

    Content is identified by its binary signature; extension and declared type are never enough for a binary file. Macro-enabled files, Office 97-2003, OpenDocument and archives are refused with an explicit code.

  3. Extractor present

    Fifteen extractors are registered (text, Markdown, DOCX, CSV, HTML, PDF, image, Excel, PPTX, XML, DXF, IFC, DWG, EML, audio).

  4. Activation

    Environment flags and image build options decide what actually works: the Compact profile is built without OCR, IFC, DWG or audio.

Statuses

Enabled by default
Works in the Compact profile without extra settings.
Configurable
Present in the published code; requires a setting or environment variable.
Optional
Requires an optional component or dependency that is not included by default.
Experimental
Present but not stabilised; behaviour may change.
Disabled
Present in the code but disabled in the published Compact images.
Unavailable
Absent from the Open edition.
Not yet qualified
Present but not covered by a qualification campaign.

Catalogue

Search a format or filter by family and status. The table is complete without JavaScript; filters are a convenience.

Extraction: T text · Tb tables · I images · O OCR · L layout · G geometry · M metadata. Filled = yes, ringed = partial, empty = no.

FormatFamilyStatusExtractionFlag / dependencyDetails
Excel (.xlsx)
.xlsx		Office documentsEnabled by default
DOCUMENT_EXCEL_ENABLED
openpyxl (shipped)
Details
Limitations
Documentary ingestion only: cells rendered in windows (40 rows) with sheet/range provenance; no calculation or formula evaluation. Limits: 1,000,000 cells, 100 sheets, 256 columns. Images ignored.
Requirements
Compact stack as shipped (backend + Celery worker); no extra build argument.
Upload / backend / extractor / enabled
Yes / Yes / Yes / Yes
Pipeline (chunking, embeddings, retrieval, citations)
Yes / Yes / Yes / Yes
Tests
Unit tests present in the repository; format absent from the Compact qualification corpus.
Sources
Last verified
11 November 2026
PowerPoint (.pptx)
.pptx		Office documentsEnabled by default
DOCUMENT_PPTX_ENABLED
python-pptx (shipped)
Details
Limitations
One unit per slide (citations by slide number). Tables rendered as Markdown. Images: only author-written alt text is indexed, never OCR. Speaker notes off by default. 500-slide limit.
Requirements
Compact stack as shipped (backend + Celery worker); no extra build argument.
Upload / backend / extractor / enabled
Yes / Yes / Yes / Yes
Pipeline (chunking, embeddings, retrieval, citations)
Yes / Yes / Yes / Yes
Tests
Unit tests present in the repository; format absent from the Compact qualification corpus.
Sources
Last verified
11 November 2026
Word (.docx)
.docx		Office documentsEnabled by default
DOCUMENT_OFFICE_CHART_DATA_ENABLED
Details
Limitations
Document body only: headers, footers, footnotes and text boxes are not extracted. Tables linearised. Embedded images ignored. Citations without page number (unpaginated format).
Requirements
Compact stack as shipped (backend + Celery worker); no extra build argument.
Upload / backend / extractor / enabled
Yes / Yes / Yes / Yes
Pipeline (chunking, embeddings, retrieval, citations)
Yes / Yes / Yes / Yes
Tests
Unit tests present and format exercised by the Compact qualification corpus; detailed results not published in the repository.
Sources
Last verified
11 November 2026
CSV
.csv, .tsv		Structured dataEnabled by default
TABULAR_QUERY_ENABLED
Details
Limitations
Comma, semicolon or tab separators. Limits: 100,000 rows, 256 columns, 10,000 characters per cell. One "column: value" unit per row. No tabular computation by default.
Requirements
Compact stack as shipped (backend + Celery worker); no extra build argument.
Upload / backend / extractor / enabled
Yes / Yes / Yes / Yes
Pipeline (chunking, embeddings, retrieval, citations)
Yes / Yes / Yes / Yes
Tests
Unit tests present in the repository; format absent from the Compact qualification corpus.
Sources
Last verified
11 November 2026
HTML
.html, .htm		Text documentsEnabled by default
—
Details
Limitations
Treated as hostile: scripts, styles, iframes, SVG and forms removed; no external resource loaded. Table rows flattened to text. 2,000,000-character limit.
Requirements
Compact stack as shipped (backend + Celery worker); no extra build argument.
Upload / backend / extractor / enabled
Yes / Yes / Yes / Yes
Pipeline (chunking, embeddings, retrieval, citations)
Yes / Yes / Yes / Yes
Tests
Unit tests present in the repository; format absent from the Compact qualification corpus.
Sources
Last verified
11 November 2026
Markdown (.md)
.md, .markdown		Text documentsEnabled by default
—
Details
Limitations
Recognised by the .md/.markdown extension only. Headings, lists, tables and code fences kept as typed blocks; inline HTML neutralised; links never followed. 2,000,000-character limit.
Requirements
Compact stack as shipped (backend + Celery worker); no extra build argument.
Upload / backend / extractor / enabled
Yes / Yes / Yes / Yes
Pipeline (chunking, embeddings, retrieval, citations)
Yes / Yes / Yes / Yes
Tests
Unit tests present and format exercised by the Compact qualification corpus; detailed results not published in the repository.
Sources
Last verified
11 November 2026
PDF (text layer)
.pdf		Text documentsEnabled by default
DOCUMENT_PDF_NATIVE_FALLBACK_ENABLED
pdfplumber/pdfminer.six (shipped, extra [tables]); PyMuPDF (NOT shipped, AGPL)
Details
Limitations
pypdf engine with pdfminer/pdfplumber fallback. 500 pages and 180 s extraction budget maximum. Tables: only recognised table types are extracted by default; generic grids need a flag and a space allowlist. Embedded images are never extracted in Compact. Scanned PDFs fail without OCR.
Requirements
Compact stack as shipped (backend + Celery worker); no extra build argument.
Upload / backend / extractor / enabled
Yes / Yes / Yes / Yes
Pipeline (chunking, embeddings, retrieval, citations)
Yes / Yes / Yes / Yes
Tests
Unit tests present and format exercised by the Compact qualification corpus; detailed results not published in the repository.
Sources
Last verified
11 November 2026
Plain text (.txt)
.txt		Text documentsEnabled by default
—
Details
Limitations
Best-effort decoding (UTF-8, cp1252, latin-1); 5,000,000-character limit. Citations carry no page number.
Requirements
Compact stack as shipped (backend + Celery worker); no extra build argument.
Upload / backend / extractor / enabled
Yes / Yes / Yes / Yes
Pipeline (chunking, embeddings, retrieval, citations)
Yes / Yes / Yes / Yes
Tests
Unit tests present and format exercised by the Compact qualification corpus; detailed results not published in the repository.
Sources
Last verified
11 November 2026
MIME e-mail (.eml)
.eml		E-mailConfigurable
DOCUMENT_EML_ENABLED
Details
Limitations
Closed by default (eml_ingestion_disabled); even when open, upload is admitted only in listed spaces or with the global flag. Headers and body extracted; attachments only inventoried unless a dedicated flag is set (they then follow their own format’s pipeline). Limits: 25 MB, 20 attachments.
Requirements
DOCUMENT_EML_ENABLED=1 and a space allowlist (or DOCUMENT_EMAIL_GLOBAL_ENABLED=1); no rebuild.
Upload / backend / extractor / enabled
No / No / Yes / No
Pipeline (chunking, embeddings, retrieval, citations)
Yes / Yes / Yes / Yes
Tests
Unit tests present in the repository; format absent from the Compact qualification corpus.
Sources
Last verified
11 November 2026
Generic XML (.xml)
.xml		Structured dataConfigurable
DOCUMENT_XML_ENABLED
defusedxml 0.7.1 (shipped)
Details
Limitations
Closed by default (xml_ingestion_disabled); the published Compact configuration does not open it. Secure parsing (DTD, entities and external references refused; depth 64, 200,000 nodes). Structured units with element-path locators.
Requirements
DOCUMENT_XML_ENABLED=1 on backend and worker; no rebuild.
Upload / backend / extractor / enabled
No / No / Yes / No
Pipeline (chunking, embeddings, retrieval, citations)
Yes / Yes / Yes / Yes
Tests
Unit tests present in the repository; format absent from the Compact qualification corpus.
Sources
Last verified
11 November 2026
DXF drawing (.dxf)
.dxf		Technical formatsConfigurable
DOCUMENT_DXF_ENABLED
ezdxf 1.4.4 (shipped, MIT)
Details
Limitations
Closed by default (dxf_ingestion_disabled); the Compact configuration does not open it. Documentary support: annotations (TEXT, MTEXT, dimensions), block attributes, layers and declared references. Geometric entities are only counted: no coordinates indexed, no rendering, no interpretation of the drawing.
Requirements
DOCUMENT_DXF_ENABLED=1 on backend and worker; no rebuild (ezdxf shipped).
Upload / backend / extractor / enabled
No / No / Yes / No
Pipeline (chunking, embeddings, retrieval, citations)
Yes / Yes / Yes / Yes
Tests
Unit tests present in the repository; format absent from the Compact qualification corpus.
Sources
Last verified
11 November 2026
JPEG image
.jpg, .jpeg		ImagesOptional
DOCUMENT_IMAGE_INGEST_ENABLED
tesseract-ocr + pytesseract (build arg WITH_OCR=1)
Details
Limitations
Accepted at upload (DOCUMENT_IMAGE_INGEST_ENABLED=1 by default), but extraction relies on OCR only: in Compact (WITH_OCR=0) ingestion fails with ocr_engine_missing. No visual description by a vision model in the Open edition.
Requirements
Backend/worker image rebuilt with WITH_OCR=1 (Tesseract + fra/eng data); CPU time per page.
Upload / backend / extractor / enabled
Yes / Yes / Yes / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
Unit tests present in the repository; capability not exercised in Compact qualification (build option disabled).
Sources
Last verified
11 November 2026
PNG image
.png		ImagesOptional
DOCUMENT_IMAGE_INGEST_ENABLED
tesseract-ocr + pytesseract (build arg WITH_OCR=1)
Details
Limitations
Accepted at upload (DOCUMENT_IMAGE_INGEST_ENABLED=1 by default), but extraction relies on OCR only: in Compact (WITH_OCR=0) ingestion fails with ocr_engine_missing. No visual description by a vision model in the Open edition.
Requirements
Backend/worker image rebuilt with WITH_OCR=1 (Tesseract + fra/eng data); CPU time per page.
Upload / backend / extractor / enabled
Yes / Yes / Yes / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
Unit tests present in the repository; capability not exercised in Compact qualification (build option disabled).
Sources
Last verified
11 November 2026
Scanned / image-only PDF (OCR)
.pdf		ImagesOptional
DOCUMENT_OCR_ENABLED
tesseract-ocr (+fra/eng data), pytesseract, pypdfium2
Details
Limitations
Upload accepted (valid PDF) but ingestion fails in Compact (pdf_scanned_requires_ocr). With an image rebuilt with OCR: local Tesseract only, 50 pages maximum, 60 s per page, fra+eng languages; table reconstruction from OCR is experimental and not qualified.
Requirements
Backend/worker image rebuilt with WITH_OCR=1 (Tesseract + fra/eng data); CPU time per page.
Upload / backend / extractor / enabled
Yes / Yes / Yes / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
Unit tests present in the repository; capability not exercised in Compact qualification (build option disabled).
Sources
Last verified
11 November 2026
TIFF image
.tif, .tiff		ImagesOptional
DOCUMENT_IMAGE_INGEST_ENABLED
tesseract-ocr + pytesseract (build arg WITH_OCR=1)
Details
Limitations
Accepted at upload (DOCUMENT_IMAGE_INGEST_ENABLED=1 by default), but extraction relies on OCR only: in Compact (WITH_OCR=0) ingestion fails with ocr_engine_missing. No visual description by a vision model in the Open edition.
Requirements
Backend/worker image rebuilt with WITH_OCR=1 (Tesseract + fra/eng data); CPU time per page.
Upload / backend / extractor / enabled
Yes / Yes / Yes / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
Unit tests present in the repository; capability not exercised in Compact qualification (build option disabled).
Sources
Last verified
11 November 2026
WEBP image
.webp		ImagesOptional
DOCUMENT_IMAGE_INGEST_ENABLED
tesseract-ocr + pytesseract (build arg WITH_OCR=1)
Details
Limitations
Accepted at upload (DOCUMENT_IMAGE_INGEST_ENABLED=1 by default), but extraction relies on OCR only: in Compact (WITH_OCR=0) ingestion fails with ocr_engine_missing. No visual description by a vision model in the Open edition.
Requirements
Backend/worker image rebuilt with WITH_OCR=1 (Tesseract + fra/eng data); CPU time per page.
Upload / backend / extractor / enabled
Yes / Yes / Yes / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
Unit tests present in the repository; capability not exercised in Compact qualification (build option disabled).
Sources
Last verified
11 November 2026
DWG drawing (.dwg)
.dwg		Technical formatsOptional
DOCUMENT_DWG_ENABLED
GNU LibreDWG 0.14 dwg2dxf (GPLv3+, separate executable)
Details
Limitations
Build option disabled in Compact (WITH_DWG=0). Converted to a temporary DXF by LibreDWG (GPL, compliance review required by the documentation) then same documentary scope as DXF; AutoCAD 2000-2018 versions only.
Requirements
Separate worker container built with WITH_DWG=1 (LibreDWG compiled), ingest_dwg queue, DOCUMENT_DWG_ENABLED=1.
Upload / backend / extractor / enabled
No / No / Yes / No
Pipeline (chunking, embeddings, retrieval, citations)
Yes / Yes / Yes / Yes
Tests
Unit tests present in the repository; capability not exercised in Compact qualification (build option disabled).
Sources
Last verified
11 November 2026
IFC model (.ifc)
.ifc		Technical formatsOptional
DOCUMENT_IFC_ENABLED
ifcopenshell 0.8.5 (+shapely, lark, isodate)
Details
Limitations
Build option disabled in Compact (WITH_IFC=0) and flag closed. Semantic extraction only: spatial hierarchy, elements, properties, quantities, materials; no geometry, no rendering. Requires a dedicated worker.
Requirements
Separate worker container built with WITH_IFC=1 (IfcOpenShell), ingest_ifc queue, DOCUMENT_IFC_ENABLED=1.
Upload / backend / extractor / enabled
No / No / Yes / No
Pipeline (chunking, embeddings, retrieval, citations)
Yes / Yes / Yes / Yes
Tests
Unit tests present in the repository; capability not exercised in Compact qualification (build option disabled).
Sources
Last verified
11 November 2026
JSON (as plain text)
.json		Text documentsNot yet qualified
—
Details
Limitations
No dedicated extractor: the file is classified as plain text and indexed as is (markup or structure included). Not tested; retrieval quality not guaranteed.
Requirements
Compact stack as shipped (backend + Celery worker); no extra build argument.
Upload / backend / extractor / enabled
No / Yes / No / Not applicable
Pipeline (chunking, embeddings, retrieval, citations)
Yes / Yes / Yes / Yes
Tests
No dedicated test identified.
Sources
Last verified
11 November 2026
RTF (as plain text)
.rtf		Text documentsNot yet qualified
—
Details
Limitations
No dedicated extractor: the file is classified as plain text and indexed as is (markup or structure included). Not tested; retrieval quality not guaranteed.
Requirements
Compact stack as shipped (backend + Celery worker); no extra build argument.
Upload / backend / extractor / enabled
No / Yes / No / Not applicable
Pipeline (chunking, embeddings, retrieval, citations)
Yes / Yes / Yes / Yes
Tests
No dedicated test identified.
Sources
Last verified
11 November 2026
SVG (as plain text)
.svg		Text documentsNot yet qualified
—
Details
Limitations
No dedicated extractor: the file is classified as plain text and indexed as is (markup or structure included). Not tested; retrieval quality not guaranteed.
Requirements
Compact stack as shipped (backend + Celery worker); no extra build argument.
Upload / backend / extractor / enabled
No / Yes / No / Not applicable
Pipeline (chunking, embeddings, retrieval, citations)
Yes / Yes / Yes / Yes
Tests
No dedicated test identified.
Sources
Last verified
11 November 2026
YAML (as plain text)
.yaml, .yml		Text documentsNot yet qualified
—
Details
Limitations
No dedicated extractor: the file is classified as plain text and indexed as is (markup or structure included). Not tested; retrieval quality not guaranteed.
Requirements
Compact stack as shipped (backend + Celery worker); no extra build argument.
Upload / backend / extractor / enabled
No / Yes / No / Not applicable
Pipeline (chunking, embeddings, retrieval, citations)
Yes / Yes / Yes / Yes
Tests
No dedicated test identified.
Sources
Last verified
11 November 2026
EPUB e-book
.epub		ArchivesDisabled
—
Details
Limitations
Unrecognised ZIP archive: refused (archive_not_supported). No EPUB extractor.
Requirements
Not applicable: the file is refused at upload.
Upload / backend / extractor / enabled
No / No / No / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
The refusal is covered by unit tests.
Sources
Last verified
11 November 2026
ZIP archive
.zip		ArchivesDisabled
—
Details
Limitations
Any PK archive not recognised as Office Open XML or OpenDocument is refused (archive_not_supported). Upload files individually.
Requirements
Not applicable: the file is refused at upload.
Upload / backend / extractor / enabled
No / No / No / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
The refusal is covered by unit tests.
Sources
Last verified
11 November 2026
BMP image
.bmp		ImagesDisabled
DOCUMENT_IMAGE_FORMATS
tesseract-ocr + pytesseract (build arg WITH_OCR=1)
Details
Limitations
Detected by signature but refused (image_unsupported_format): not in the default DOCUMENT_IMAGE_FORMATS list. Adding it also requires an OCR-enabled image build.
Requirements
DOCUMENT_IMAGE_FORMATS allowlist change and image built with WITH_OCR=1.
Upload / backend / extractor / enabled
No / No / Yes / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
The refusal is covered by unit tests.
Sources
Last verified
11 November 2026
GIF image
.gif		ImagesDisabled
DOCUMENT_IMAGE_FORMATS
tesseract-ocr + pytesseract (build arg WITH_OCR=1)
Details
Limitations
Detected by signature but refused (image_unsupported_format): not in the default DOCUMENT_IMAGE_FORMATS list. Adding it also requires an OCR-enabled image build.
Requirements
DOCUMENT_IMAGE_FORMATS allowlist change and image built with WITH_OCR=1.
Upload / backend / extractor / enabled
No / No / Yes / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
The refusal is covered by unit tests.
Sources
Last verified
11 November 2026
Excel 97-2003 (.xls)
.xls		Office documentsDisabled
—
Details
Limitations
OLE2 container detected by signature and refused (legacy_office_format). Save in Office Open XML format. Detection is content-based: a plain text file merely renamed with this extension is processed as plain text.
Requirements
Not applicable: the file is refused at upload.
Upload / backend / extractor / enabled
No / No / No / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
The refusal is covered by unit tests.
Sources
Last verified
11 November 2026
Macro-enabled Excel (.xlsm)
.xlsm		Office documentsDisabled
—
Details
Limitations
Detected by content (vbaProject or macroEnabled content type), even when renamed, and refused (macro-enabled document). Save without macros.
Requirements
Not applicable: the file is refused at upload.
Upload / backend / extractor / enabled
No / No / No / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
The refusal is covered by unit tests.
Sources
Last verified
11 November 2026
Macro-enabled PowerPoint (.pptm)
.pptm		Office documentsDisabled
—
Details
Limitations
Detected by content (vbaProject or macroEnabled content type), even when renamed, and refused (macro-enabled document). Save without macros.
Requirements
Not applicable: the file is refused at upload.
Upload / backend / extractor / enabled
No / No / No / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
The refusal is covered by unit tests.
Sources
Last verified
11 November 2026
Macro-enabled Word (.docm)
.docm		Office documentsDisabled
—
Details
Limitations
Detected by content (vbaProject or macroEnabled content type), even when renamed, and refused (macro-enabled document). Save without macros.
Requirements
Not applicable: the file is refused at upload.
Upload / backend / extractor / enabled
No / No / No / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
The refusal is covered by unit tests.
Sources
Last verified
11 November 2026
OpenDocument presentation (.odp)
.odp		Office documentsDisabled
—
Details
Limitations
OpenDocument format detected by the archive "mimetype" entry and refused. Export to Office Open XML (or CSV for a spreadsheet).
Requirements
Not applicable: the file is refused at upload.
Upload / backend / extractor / enabled
No / No / No / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
The refusal is covered by unit tests.
Sources
Last verified
11 November 2026
OpenDocument spreadsheet (.ods)
.ods		Office documentsDisabled
—
Details
Limitations
OpenDocument format detected by the archive "mimetype" entry and refused. Export to Office Open XML (or CSV for a spreadsheet).
Requirements
Not applicable: the file is refused at upload.
Upload / backend / extractor / enabled
No / No / No / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
The refusal is covered by unit tests.
Sources
Last verified
11 November 2026
OpenDocument text (.odt)
.odt		Office documentsDisabled
—
Details
Limitations
OpenDocument format detected by the archive "mimetype" entry and refused. Export to Office Open XML (or CSV for a spreadsheet).
Requirements
Not applicable: the file is refused at upload.
Upload / backend / extractor / enabled
No / No / No / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
The refusal is covered by unit tests.
Sources
Last verified
11 November 2026
PowerPoint 97-2003 (.ppt)
.ppt		Office documentsDisabled
—
Details
Limitations
OLE2 container detected by signature and refused (legacy_office_format). Save in Office Open XML format. Detection is content-based: a plain text file merely renamed with this extension is processed as plain text.
Requirements
Not applicable: the file is refused at upload.
Upload / backend / extractor / enabled
No / No / No / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
The refusal is covered by unit tests.
Sources
Last verified
11 November 2026
Word 97-2003 (.doc)
.doc		Office documentsDisabled
—
Details
Limitations
OLE2 container detected by signature and refused (legacy_office_format). Save in Office Open XML format. Detection is content-based: a plain text file merely renamed with this extension is processed as plain text.
Requirements
Not applicable: the file is refused at upload.
Upload / backend / extractor / enabled
No / No / No / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
The refusal is covered by unit tests.
Sources
Last verified
11 November 2026
Executable
.exe, .dll, .so, .sh, .class		OtherDisabled
—
Details
Limitations
Executable signature detected and refused.
Requirements
Not applicable: the file is refused at upload.
Upload / backend / extractor / enabled
No / No / No / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
The refusal is covered by unit tests.
Sources
Last verified
11 November 2026
Unrecognised binary
OtherDisabled
—
Details
Limitations
Unidentified binary file: refused (unsupported_file_type).
Requirements
Not applicable: the file is refused at upload.
Upload / backend / extractor / enabled
No / No / No / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
The refusal is covered by unit tests.
Sources
Last verified
11 November 2026
FLAC audio
.flac		AudioUnavailable
DOCUMENT_AUDIO_INGEST_ENABLED
ffmpeg/ffprobe (build arg WITH_AUDIO=1) + Gateway ASR backend (not included in ai-gateway-open)
Details
Limitations
The audio path exists in the code (dedicated endpoint, 600 MB ceiling, 12 h maximum, timestamped transcript cited by time range), but the Compact profile refuses the upload (audio_ingestion_disabled) and AI Gateway Open ships no speech-to-text engine: without an operator-provided ASR service, audio cannot be processed.
Requirements
Audio worker built with WITH_AUDIO=1 (ffmpeg), ingest_audio queue, audio ingestion flag and space allowlist, AND a compatible speech-to-text (ASR) service behind the gateway, not provided by AI Gateway Open.
Upload / backend / extractor / enabled
No / No / Yes / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
Unit tests present in the repository; capability not exercised in Compact qualification (build option disabled).
Sources
Last verified
11 November 2026
M4A audio
.m4a		AudioUnavailable
DOCUMENT_AUDIO_INGEST_ENABLED
ffmpeg/ffprobe (build arg WITH_AUDIO=1) + Gateway ASR backend (not included in ai-gateway-open)
Details
Limitations
The audio path exists in the code (dedicated endpoint, 600 MB ceiling, 12 h maximum, timestamped transcript cited by time range), but the Compact profile refuses the upload (audio_ingestion_disabled) and AI Gateway Open ships no speech-to-text engine: without an operator-provided ASR service, audio cannot be processed.
Requirements
Audio worker built with WITH_AUDIO=1 (ffmpeg), ingest_audio queue, audio ingestion flag and space allowlist, AND a compatible speech-to-text (ASR) service behind the gateway, not provided by AI Gateway Open.
Upload / backend / extractor / enabled
No / No / Yes / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
Unit tests present in the repository; capability not exercised in Compact qualification (build option disabled).
Sources
Last verified
11 November 2026
MP3 audio
.mp3		AudioUnavailable
DOCUMENT_AUDIO_INGEST_ENABLED
ffmpeg/ffprobe (build arg WITH_AUDIO=1) + Gateway ASR backend (not included in ai-gateway-open)
Details
Limitations
The audio path exists in the code (dedicated endpoint, 600 MB ceiling, 12 h maximum, timestamped transcript cited by time range), but the Compact profile refuses the upload (audio_ingestion_disabled) and AI Gateway Open ships no speech-to-text engine: without an operator-provided ASR service, audio cannot be processed.
Requirements
Audio worker built with WITH_AUDIO=1 (ffmpeg), ingest_audio queue, audio ingestion flag and space allowlist, AND a compatible speech-to-text (ASR) service behind the gateway, not provided by AI Gateway Open.
Upload / backend / extractor / enabled
No / No / Yes / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
Unit tests present in the repository; capability not exercised in Compact qualification (build option disabled).
Sources
Last verified
11 November 2026
OGG audio
.ogg		AudioUnavailable
DOCUMENT_AUDIO_INGEST_ENABLED
ffmpeg/ffprobe (build arg WITH_AUDIO=1) + Gateway ASR backend (not included in ai-gateway-open)
Details
Limitations
The audio path exists in the code (dedicated endpoint, 600 MB ceiling, 12 h maximum, timestamped transcript cited by time range), but the Compact profile refuses the upload (audio_ingestion_disabled) and AI Gateway Open ships no speech-to-text engine: without an operator-provided ASR service, audio cannot be processed.
Requirements
Audio worker built with WITH_AUDIO=1 (ffmpeg), ingest_audio queue, audio ingestion flag and space allowlist, AND a compatible speech-to-text (ASR) service behind the gateway, not provided by AI Gateway Open.
Upload / backend / extractor / enabled
No / No / Yes / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
Unit tests present in the repository; capability not exercised in Compact qualification (build option disabled).
Sources
Last verified
11 November 2026
WAV audio
.wav		AudioUnavailable
DOCUMENT_AUDIO_INGEST_ENABLED
ffmpeg/ffprobe (build arg WITH_AUDIO=1) + Gateway ASR backend (not included in ai-gateway-open)
Details
Limitations
The audio path exists in the code (dedicated endpoint, 600 MB ceiling, 12 h maximum, timestamped transcript cited by time range), but the Compact profile refuses the upload (audio_ingestion_disabled) and AI Gateway Open ships no speech-to-text engine: without an operator-provided ASR service, audio cannot be processed.
Requirements
Audio worker built with WITH_AUDIO=1 (ffmpeg), ingest_audio queue, audio ingestion flag and space allowlist, AND a compatible speech-to-text (ASR) service behind the gateway, not provided by AI Gateway Open.
Upload / backend / extractor / enabled
No / No / Yes / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
Unit tests present in the repository; capability not exercised in Compact qualification (build option disabled).
Sources
Last verified
11 November 2026
WEBM audio
.webm		AudioUnavailable
DOCUMENT_AUDIO_INGEST_ENABLED
ffmpeg/ffprobe (build arg WITH_AUDIO=1) + Gateway ASR backend (not included in ai-gateway-open)
Details
Limitations
The audio path exists in the code (dedicated endpoint, 600 MB ceiling, 12 h maximum, timestamped transcript cited by time range), but the Compact profile refuses the upload (audio_ingestion_disabled) and AI Gateway Open ships no speech-to-text engine: without an operator-provided ASR service, audio cannot be processed.
Requirements
Audio worker built with WITH_AUDIO=1 (ffmpeg), ingest_audio queue, audio ingestion flag and space allowlist, AND a compatible speech-to-text (ASR) service behind the gateway, not provided by AI Gateway Open.
Upload / backend / extractor / enabled
No / No / Yes / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
Unit tests present in the repository; capability not exercised in Compact qualification (build option disabled).
Sources
Last verified
11 November 2026
Outlook (.msg)
.msg		E-mailUnavailable
DOCUMENT_MSG_ENABLED
extract-msg (GPL)
Details
Limitations
Detected by signature but always refused: the .msg parser (GPL dependency) was removed from the Open edition. Export the message as .eml.
Requirements
Not achievable in the Open edition.
Upload / backend / extractor / enabled
No / No / No / No
Pipeline (chunking, embeddings, retrieval, citations)
No / No / No / No
Tests
The refusal is covered by unit tests.
Sources
Last verified
11 November 2026

Worth knowing

  • Scanned PDFs and images are accepted at upload but fail at extraction in Compact (no OCR in the published image). Enabling OCR requires rebuilding the image with WITH_OCR=1.
  • XML, DXF and EML ship with their parsers but are closed by a flag in the published Compact configuration; the repository README lists them as verified, which the published configuration does not reflect.
  • Audio has a complete path on the application side, but AI Gateway Open contains no speech-to-text engine: without an operator-provided ASR service, audio remains unavailable.
  • Unrecognised text files (JSON, YAML, RTF, SVG…) are indexed as plain text, with no dedicated extractor and no test.
  • PDF tables are extracted by default only for recognised table types; generic grids require a flag and a space allowlist.
  • The Compact qualification exercised TXT, Markdown, text PDF and DOCX; XLSX, PPTX, CSV and HTML have unit tests but are not in the qualification corpus.