ragflow [Question]: In v0.19.0, In addition to the General chunking method, other methods(like Presentation) still cannot be parsed using the VLM model, and this problem does not seem to be solved.

Self Checks

[x] I have searched for existing issues search for existing issues, including closed ones.
[x] I confirm that I am using English to submit this report (Language Policy).
[x] Non-english title submitions will be closed directly ( 非英文标题的提交将会被直接关闭 ) (Language Policy).
[x] Please do not modify this template :) and fill in all the required fields.

Describe your problem

Actually, #8109 still hasn't been resolved.

Here is Log:

2025-06-11 23:13:04,980 INFO 2025-06-11 23:13:04,986 INFO 2025-06-11 23:13:04,988 INFO 2025-06-11 23:13:04,996 INFO 2025-06-11 23:13:04,997 INFO 2025-06-11 23:13:04,997 INFO 2025-06-11 23:13:05,026 INFO 2025-06-11 23:13:05,027 INFO 2025-06-11 23:13:06,139 INFO 2025-06-11 23:13:06,154 INFO 2025-06-11 23:13:06,225 INFO 2025-06-11 23:13:06,229 INFO 2025-06-11 23:13:08,158 INFO 2025-06-11 23:13:08,163 INFO 2025-06-11 23:13:08,163 INFO 2025-06-11 23:13:15,208 INFO 2025-06-11 23:13:15,232 INFO 2025-06-11 23:13:15,233 INFO 2025-06-11 23:13:15,253 INFO 2025-06-11 23:13:15,287 INFO 2025-06-11 23:13:15,300 INFO 2025-06-11 23:13:19,496 INFO 2025-06-11 23:13:19,566 INFO 2025-06-11 23:13:19,576 INFO 2025-06-11 23:13:19,583 INFO 2025-06-11 23:13:19,583 INFO 2025-06-11 23:13:19,586 INFO 2025-06-11 23:13:19,587 INFO 2025-06-11 23:13:19,587 INFO 2025-06-11 23:13:19,598 INFO 2025-06-11 23:13:21,431 INFO 2025-06-11 23:13:21,435 WARNING 2025-06-11 23:13:21,530 INFO 2025-06-11 23:13:21,531 INFO 2025-06-11 23:13:21,570 INFO 2025-06-11 23:13:21,669 INFO 2025-06-11 23:13:21,670 INFO 2025-06-11 23:13:21,701 INFO 2025-06-11 23:13:21,809 INFO 2025-06-11 23:13:21,813 INFO 2025-06-11 23:13:22,002 INFO 2025-06-11 23:13:22,108 INFO 2025-06-11 23:13:22,110 INFO 2025-06-11 23:13:22,246 INFO 2025-06-11 23:13:22,354 INFO 2025-06-11 23:13:22,356 INFO 2025-06-11 23:13:22,568 INFO 2025-06-11 23:13:22,678 INFO 2025-06-11 23:13:22,680 INFO 2025-06-11 23:13:23,024 INFO 2025-06-11 23:13:23,030 INFO 2025-06-11 23:13:23,136 INFO 2025-06-11 23:13:23,137 INFO 2025-06-11 23:13:23,796 INFO 2025-06-11 23:13:23,891 INFO 2025-06-11 23:13:23,892 INFO 2025-06-11 23:13:23,925 INFO 2025-06-11 23:13:24,019 INFO 2025-06-11 23:13:24,021 INFO 2025-06-11 23:13:24,049 INFO 2025-06-11 23:13:24,133 INFO 2025-06-11 23:13:24,134 INFO 2025-06-11 23:13:24,169 INFO 2025-06-11 23:13:24,170 INFO 2025-06-11 23:13:24,175 INFO 2025-06-11 23:13:24,180 INFO 2025-06-11 23:13:24,258 INFO 2025-06-11 23:13:24,407 INFO 2025-06-11 23:13:24,407 INFO 2025-06-11 23:13:24,411 INFO 2025-06-11 23:13:24,463 INFO 2025-06-11 23:13:25,739 INFO 2025-06-11 23:13:25,792 INFO 2025-06-11 23:13:25,792 INFO 2025-06-11 23:13:25,796 INFO 2025-06-11 23:13:25,816 INFO 2025-06-11 23:13:25,819 INFO 2025-06-11 23:13:25,839 INFO 2025-06-11 23:13:25,851 INFO 2025-06-11 23:13:25,853 INFO 2025-06-11 23:13:25,859 INFO 2025-06-11 23:13:25,859 INFO 2025-06-11 23:13:25,859 INFO 2025-06-11 23:13:27,619 INFO 2025-06-11 23:13:27,620 INFO 2025-06-11 23:13:27,623 INFO 2025-06-11 23:13:27,627 INFO 2025-06-11 23:13:27,630 INFO 2025-06-11 23:13:27,632 INFO 2025-06-11 23:13:27,666 INFO 2025-06-11 23:13:27,666 INFO 2025-06-11 23:13:30,883 INFO 2025-06-11 23:13:30,883 INFO 2025-06-11 23:13:30,888 INFO 2025-06-11 23:13:30,889 INFO 2025-06-11 23:13:30,907 INFO 2025-06-11 23:13:30,909 INFO 2025-06-11 23:13:30,924 INFO 2025-06-11 23:13:30,924 INFO 18 172.24.0.6 - - [11/Jun/2025 23:13:04] "GET /v1/user/info HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:04] "GET /v1/tenant/list HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:04] "GET /v1/user/tenant_info HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:04] "GET /v1/kb/detail?kb_id=5c23770246ce11f08b8b0242ac180006 HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:04] "POST /v1/document/list?kb_id=5c23770246ce11f08b8b0242ac180006&keywords=&page_size=10&page=1 HTTP/1.1" 200 - 18 HEAD http://es01:9200/ragflow_ff5921f8f59b11ef88e50242ac120006 [status:200 duration:0.003s] 18 POST http://es01:9200/ragflow_ff5921f8f59b11ef88e50242ac120006/_search [status:200 duration:0.005s] 18 172.24.0.6 - - [11/Jun/2025 23:13:05] "GET /v1/kb/5c23770246ce11f08b8b0242ac180006/knowledge_graph HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:06] "GET /v1/kb/detail?kb_id=5c23770246ce11f08b8b0242ac180006 HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:06] "GET /v1/user/tenant_info HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:06] "GET /v1/llm/list HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:06] "POST /v1/kb/list HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:08] "GET /v1/user/tenant_info HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:08] "GET /v1/kb/detail?kb_id=5c23770246ce11f08b8b0242ac180006 HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:08] "POST /v1/document/list?kb_id=5c23770246ce11f08b8b0242ac180006&keywords=&page_size=10&page=1 HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:15] "POST /v1/document/upload HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:15] "POST /v1/document/list?kb_id=5c23770246ce11f08b8b0242ac180006&keywords=&page_size=10&page=1 HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:15] "POST /v1/document/list?kb_id=5c23770246ce11f08b8b0242ac180006&keywords=&page_size=10&page=1 HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:15] "POST /v1/document/run HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:15] "GET /v1/document/image/5c23770246ce11f08b8b0242ac180006-thumbnail_98f3019a46d611f090cc0242ac180006.png HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:15] "POST /v1/document/list?kb_id=5c23770246ce11f08b8b0242ac180006&keywords=&page_size=10&page=1 HTTP/1.1" 200 - 31 handle_task begin for task {"id": "99020a8246d611f0aba00242ac180006", "doc_id": "98f3019a46d611f090cc0242ac180006", "from_page": 0, "to_page": 10, "retry_count": 0, "kb_id": "5c23770246ce11f08b8b0242ac180006", "parser_id": "presentation", "parser_config": {"layout_recognize": "Qwen2.5-VL-7B-Instruct___VLLM@VLLM", "auto_keywords": 0, "auto_questions": 0, "raptor": {"use_raptor": false}, "graphrag": {"use_graphrag": false}}, "name": "test_doc.pdf", "type": "pdf", "location": "test_doc.pdf", "size": 1442605, "tenant_id": "ff5921f8f59b11ef88e50242ac120006", "language": "English", "embd_id": "bge-m3@Xinference", "pagerank": 0, "kb_parser_config": {"layout_recognize": "Qwen2.5-VL-7B-Instruct___VLLM@VLLM", "auto_keywords": 0, "auto_questions": 0, "raptor": {"use_raptor": false}, "graphrag": {"use_graphrag": false}}, "img2txt_id": "Qwen2.5-VL-7B-Instruct___VLLM@VLLM", "asr_id": "", "llm_id": "DeepSeek-R1-Distill-Qwen-32B___OpenAI-API@OpenAI-API-Compatible", "update_time": 1749654795251, "task_type": ""} 31 HTTP Request: POST http://10.133.34.19:9998/v1/embeddings "HTTP/1.1 200 OK" 31 HEAD http://es01:9200/ragflow_ff5921f8f59b11ef88e50242ac120006 [status:200 duration:0.003s] 31 From minio(0.006399601930752397) test_doc.pdf/test_doc.pdf 31 load_model /ragflow/rag/res/deepdoc/det.onnx reuses cached model 31 load_model /ragflow/rag/res/deepdoc/rec.onnx reuses cached model 31 load_model /ragflow/rag/res/deepdoc/layout.onnx reuses cached model 31 load_model /ragflow/rag/res/deepdoc/tsr.onnx reuses cached model 31 set_progress(99020a8246d611f0aba00242ac180006), progress: None, progress_msg: 23:13:19 Page(1~11): OCR started 31 images dedupe_chars cost 1.8335928990272805s 31 Miss outlines 31 __ocr detecting boxes of a image cost (0.09450889506842941s) 31 __ocr sorting 53 chars cost 0.0006220480427145958s 31 __ocr recognize 8 boxes cost 0.03930688195396215s 31 __ocr detecting boxes of a image cost (0.09745385695714504s) 31 __ocr sorting 43 chars cost 0.0005265669897198677s 31 __ocr recognize 7 boxes cost 0.031114830053411424s 31 __ocr detecting boxes of a image cost (0.10635609098244458s) 31 __ocr sorting 409 chars cost 0.0029008679557591677s 31 __ocr recognize 37 boxes cost 0.18889174505602568s 31 __ocr detecting boxes of a image cost (0.10438321705441922s) 31 __ocr sorting 181 chars cost 0.0016334600513800979s 31 __ocr recognize 29 boxes cost 0.13544121000450104s 31 __ocr detecting boxes of a image cost (0.10657399310730398s) 31 __ocr sorting 204 chars cost 0.001860317075625062s 31 __ocr recognize 34 boxes cost 0.2115059489151463s 31 __ocr detecting boxes of a image cost (0.10764243407174945s) 31 __ocr sorting 268 chars cost 0.001782367005944252s 31 __ocr recognize 37 boxes cost 0.3447964320657775s 31 set_progress(99020a8246d611f0aba00242ac180006), progress: 0.36, progress_msg: 31 __ocr detecting boxes of a image cost (0.10478150693234056s) 31 __ocr sorting 103 chars cost 0.0009583050850778818s 31 __ocr recognize 38 boxes cost 0.6587542770430446s 31 __ocr detecting boxes of a image cost (0.09396304795518517s) 31 __ocr sorting 68 chars cost 0.0006430470384657383s 31 __ocr recognize 5 boxes cost 0.03250362700782716s 31 __ocr detecting boxes of a image cost (0.09265734406653792s) 31 __ocr sorting 208 chars cost 0.001447585062123835s 31 __ocr recognize 8 boxes cost 0.027606691932305694s 31 __ocr detecting boxes of a image cost (0.08356666704639792s) 31 __ocr sorting 11 chars cost 0.00021717208437621593s 31 __ocr recognize 5 boxes cost 0.03483505605254322s 31 images 10 pages cost 2.735132099944167s 31 set_progress(99020a8246d611f0aba00242ac180006), progress: None, progress_msg: 23:13:24 Page(1~11): Page 0~10: OCR finished (4.58s) 31 set_progress(99020a8246d611f0aba00242ac180006), progress: 0.9, progress_msg: 23:13:24 Page(1~11): Page 0~10: Parsing finished 31 Chunking(4.682046517031267) test_doc.pdf/test_doc.pdf done 31 MINIO PUT(test_doc.pdf) cost 0.148 s 31 Build document test_doc.pdf: 4.83s 31 set_progress(99020a8246d611f0aba00242ac180006), progress: None, progress_msg: 23:13:24 Page(1~11): Generate 10 chunks 31 HTTP Request: POST http://10.133.34.19:9998/v1/embeddings "HTTP/1.1 200 OK" 31 HTTP Request: POST http://10.133.34.19:9998/v1/embeddings "HTTP/1.1 200 OK" 31 set_progress(99020a8246d611f0aba00242ac180006), progress: 0.72, progress_msg: 31 Embedding chunks (1.38s) 31 set_progress(99020a8246d611f0aba00242ac180006), progress: None, progress_msg: 23:13:25 Page(1~11): Embedding chunks (1.38s) 31 PUT http://es01:9200/ragflow_ff5921f8f59b11ef88e50242ac120006/_bulk?refresh=false&timeout=60s [status:200 duration:0.016s] 31 set_progress(99020a8246d611f0aba00242ac180006), progress: 0.81, progress_msg: 31 PUT http://es01:9200/ragflow_ff5921f8f59b11ef88e50242ac120006/_bulk?refresh=false&timeout=60s [status:200 duration:0.015s] 31 PUT http://es01:9200/ragflow_ff5921f8f59b11ef88e50242ac120006/_bulk?refresh=false&timeout=60s [status:200 duration:0.010s] 31 Indexing doc(test_doc.pdf), page(0-10), chunks(10), elapsed: 0.06 31 set_progress(99020a8246d611f0aba00242ac180006), progress: 1.0, progress_msg: 23:13:25 Page(1~11): Indexing done (0.06s). Task done (6.36s) 31 Chunk doc(test_doc.pdf), page(0-10), chunks(10), token(1540), elapsed:6.36 31 handle_task done for task {"id": "99020a8246d611f0aba00242ac180006", "doc_id": "98f3019a46d611f090cc0242ac180006", "from_page": 0, "to_page": 10, "retry_count": 0, "kb_id": "5c23770246ce11f08b8b0242ac180006", "parser_id": "presentation", "parser_config": {"layout_recognize": "Qwen2.5-VL-7B-Instruct___VLLM@VLLM", "auto_keywords": 0, "auto_questions": 0, "raptor": {"use_raptor": false}, "graphrag": {"use_graphrag": false}}, "name": "test_doc.pdf", "type": "pdf", "location": "test_doc.pdf", "size": 1442605, "tenant_id": "ff5921f8f59b11ef88e50242ac120006", "language": "English", "embd_id": "bge-m3@Xinference", "pagerank": 0, "kb_parser_config": {"layout_recognize": "Qwen2.5-VL-7B-Instruct___VLLM@VLLM", "auto_keywords": 0, "auto_questions": 0, "raptor": {"use_raptor": false}, "graphrag": {"use_graphrag": false}}, "img2txt_id": "Qwen2.5-VL-7B-Instruct___VLLM@VLLM", "asr_id": "", "llm_id": "DeepSeek-R1-Distill-Qwen-32B___OpenAI-API@OpenAI-API-Compatible", "update_time": 1749654795251, "task_type": ""} 18 172.24.0.6 - - [11/Jun/2025 23:13:27] "GET /v1/tenant/list HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:27] "GET /v1/user/info HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:27] "GET /v1/user/tenant_info HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:27] "POST /v1/document/list?kb_id=5c23770246ce11f08b8b0242ac180006&keywords=&page_size=10&page=1 HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:27] "GET /v1/kb/detail?kb_id=5c23770246ce11f08b8b0242ac180006 HTTP/1.1" 200 - 18 HEAD http://es01:9200/ragflow_ff5921f8f59b11ef88e50242ac120006 [status:200 duration:0.003s] 18 POST http://es01:9200/ragflow_ff5921f8f59b11ef88e50242ac120006/_search [status:200 duration:0.005s] 18 172.24.0.6 - - [11/Jun/2025 23:13:27] "GET /v1/kb/5c23770246ce11f08b8b0242ac180006/knowledge_graph HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:30] "GET /v1/tenant/list HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:30] "GET /v1/user/info HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:30] "GET /v1/user/tenant_info HTTP/1.1" 200 - 18 HEAD http://es01:9200/ragflow_ff5921f8f59b11ef88e50242ac120006 [status:200 duration:0.003s] 18 172.24.0.6 - - [11/Jun/2025 23:13:30] "POST /v1/document/list?kb_id=5c23770246ce11f08b8b0242ac180006&keywords=&page_size=10&page=1 HTTP/1.1" 200 - 18 172.24.0.6 - - [11/Jun/2025 23:13:30] "GET /v1/kb/detail?kb_id=5c23770246ce11f08b8b0242ac180006 HTTP/1.1" 200 - 18 POST http://es01:9200/ragflow_ff5921f8f59b11ef88e50242ac120006/_search [status:200 duration:0.005s] 18 172.24.0.6 - - [11/Jun/2025 23:13:30] "GET /v1/kb/5c23770246ce11f08b8b0242ac180006/knowledge_graph HTTP/1.1" 200 -

Jun 11 '25 07:06 Dreamcatcher-wind

You are correct—this is a known limitation. As of v0.19.0, the Presentation chunking method (for both PDF and PPTX) does not support parsing via VLM (Vision Language Model) pipelines. Only the General and some other PDF chunking methods allow you to select a vision model for layout recognition or OCR. The Presentation method still relies on classic extraction (python-pptx, Aspose.Slides, or standard OCR for PDFs) and ignores any VLM configuration, even if you set layout_recognize to a VLM model in the parser config. There have been no changes in v0.19.0 to address this, and the architecture currently does not route Presentation files through VLM-based parsing logic. This is consistent with both the code and the documentation [source] [source].

If you need VLM-based parsing, you will need to use a chunking method that explicitly supports it (like General for PDFs). There is no workaround for Presentation chunking at this time.

To reply, just mention @dosu.

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}

Jun 11 '25 07:06 dosubot[bot]

I did not quite follow. Why did you think it's not been resolved?

Jun 12 '25 05:06 KevinHuSh