### What problem does this PR solve?
fix issues：https://github.com/infiniflow/ragflow/issues/10402
As the newly distributed embedding models support vector dimensions max
to 4096, while current OpenSearch's max dimension support is 1536.
As I tested, the 4096-dimensions vector will be treated as a float type
which is unacceptable in OpenSearch.
Besides, OpenSearch supports max to 16000 dimensions by defalut with the
vector engine(Faiss). According to:
https://docs.opensearch.org/2.19/field-types/supported-field-types/knn-methods-engines/
I added max to 10240 dimensions support for OpenSearch, as I think will
be sufficient in the future.
As I tested , it worked well on my own server (treated as knn_vector)by
using qwen3-embedding:8b as the embedding model:
<img width="1338" height="790" alt="image"
src="https://github.com/user-attachments/assets/a9b2d284-fcf6-4cea-859a-6aadccf36ace"
/>
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
By the way, I will still focus on the stuff about
Elasticsearch/Opensearch as search engines and vector databases.
Co-authored-by: 张雨豪 <zhangyh80@chinatelecom.cn>
### What problem does this PR solve?
Fix: Added table of contents extraction functionality and optimized form
item layout #9869
- Added `EnableTocToggle` component to toggle table of contents
extraction on and off
- Added multiple parser configuration components (such as naive, book,
laws, etc.), displaying different parser components based on built-in
slicing methods
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
### What problem does this PR solve?
Feat: Modify the default style of the agent node anchor #9869
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
issue:
[#6571](https://github.com/infiniflow/ragflow/issues/6571)
change:
include author, journal name, volume, issue, page, and DOI in PubMed
search results
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
Refactor(login): Refactor the login page and add dynamic background and
highlight effects #9869
### Type of change
- [x] Refactoring
### What problem does this PR solve?
Fix: Fixed the issue where the connection lines of placeholder nodes in
the agent canvas could not be displayed #9869
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
### What problem does this PR solve?
Gemini 2.5 Flash Models use reasoning by default. There is currently no
way to disable this behaviour. This leads to very long response times (>
1min). The default behaviour should be, that reasoning is disabled and
configurable
issue #10474
### Type of change
- [X] Bug Fix (non-breaking change which fixes an issue)
### What problem does this PR solve?
Fixes the return type annotation for the `get_urls` function in
`download_deps.py`
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
### What problem does this PR solve?
Fix: Agent.reset() argument wrong #10463 & Unable to converse with agent
through Python API. #10415
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
### What problem does this PR solve?
Feat: Adjust the style of the delete button on the agent side #9869
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
Google Cloud model does not work correctly with gemini-2.5 models
Close#10408
### Type of change
- [X] Bug Fix (non-breaking change which fixes an issue)
---------
Co-authored-by: Kevin Hu <kevinhu.sh@gmail.com>
### What problem does this PR solve?
Fix: Fixed the issue where swagger apidocs could not be opened
properly(#9522)
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
Co-authored-by: virgilwong <hyhvirgil@gmail.com>
### What problem does this PR solve?
Feat: Added toc enhance field to chat and retrieval operator
configuration #10436
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
Fix: Bug fixes#9869
- Adjusted the breadcrumb display logic on the data flow results page
- Added the default display of "Local Upload" to the Source field in the
dataset overview table
- Replaced the original Mindmap Task field with the GraphRAG Task field
on the dataset settings page
- Optimized the build button status check criteria and adjusted the
progress information display logic
- Introduced a Tooltip in the parsing status cell component and removed
redundant Button components
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
### What problem does this PR solve?
issue:
[#6193](https://github.com/infiniflow/ragflow/issues/6193)
change:
support qwq reasoning models with non-stream output
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
Fix: Fixed an issue where parser configurations could be added
infinitely #9869
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
### What problem does this PR solve?
Fix: Fixed the issue where the operator added by clicking the plus sign
in the data flow would overlap with the original section #9886
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
### What problem does this PR solve?
Feat: Add prompts for toc relevance check according to #10436
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
Feat: add total in List dataset API, solved #10360
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
- Running DeepDoc OCR on large PDFs inside the GPU docker-compose setup
would intermittently fail with
[ONNXRuntimeError] ... p2o.Clip.6 ... Available memory of 0 is smaller
than requested bytes ...
- Root cause: load_model() in deepdoc/vision/ocr.py treated
device_id=None as-is.
torch.cuda.device_count() > device_id then raised a TypeError, the
helper returned False, and ONNXRuntime quietly fell back to
CPUExecutionProvider with
the hard-coded 512 MB limit, which then triggered the allocator failure.
- Environment where this reproduces: Windows 11, AMD 5900x, 64 GB RAM,
RTX 3090 (24 GB), docker-compose-gpu.yml from upstream, default DeepDoc
+ GraphRAG
parser settings, ingesting heavy PDF such as 《内科学》（第10版）.pdf (~180 MB).
Fixes:
- Normalize device_id to 0 when it is None before calling any CUDA APIs,
so the GPU path is considered available.
- Allow configuring the CUDA provider’s memory cap via
OCR_GPU_MEM_LIMIT_MB (default 2048 MB) and expose
OCR_ARENA_EXTEND_STRATEGY; the calculated byte
limit is logged to confirm the effective settings.
After the change, ragflow_server.log shows for example
load_model ... uses GPU (device 0, gpu_mem_limit=21474836480,
arena_strategy=kNextPowerOfTwo) and the same document finishes OCR
without allocator errors.
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
### What problem does this PR solve?
Feat: Importing data flow files from the list page #9869
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
Feat: Replace the collapse icon #9869
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
issue:
[#10417](https://github.com/infiniflow/ragflow/issues/10417)
change:
Adjusted the `searxng_url` priority logic to ensure the
frontend-provided URL takes precedence over the model’s default
configuration. This allows user-specified SearXNG endpoints to be
correctly applied during execution, improving flexibility across
different environments.
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
### What problem does this PR solve?
Feat: Modify the translation file of the workflow #9869
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
**Adds a new feature that enables the LLM to extract a structured table
of contents (TOC) directly from plain text.**
_This implementation prioritizes efficiency over reasoning — the model
runs in a strictly deterministic mode (thinking disabled) to minimize
latency.
As a result, overall performance may be less optimal, but the extraction
speed and consistency are guaranteed._
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### Related issues
#10078
### What problem does this PR solve?
Integrate DeerAPI provider.
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
- [x] Documentation Update
Co-authored-by: DeerAPI <tensor.null@gmail.com>
### What problem does this PR solve?
Hi team, @ZhenhangTung @KevinHuSh @cike8899
About #10384 , I've completed the UI optimization adjustments for the
Agent page according to our previous discussions and the design draft
sketches provided by @Naomi. The main modifications include:
1. Adjusted the style and content of placeholder-node.
2. Adjusted the location of the dropdown (to the right of the
placeholder-node) .
3. Adjusted the tooltip position spacing when the mouse hovers in the
dropdown menu.
4. Hides the thick scroll bar on the dropdown component.
5. Highlight the connection line when dragging to generate a
placeholder-node
<img width="1323" height="509" alt="Image"
src="https://github.com/user-attachments/assets/0d366f7f-477d-4c00-bb58-d5d58b3a745f"
/>
Please review the related code modifications when you have time. Let me
know if further adjustments are needed!
Thanks！
### Type of change
- [x] Other (please describe): UI Enhancement
---------
Co-authored-by: leonlai <leonlai@futurefab.ai>
### What problem does this PR solve?
- Admin service support SHOW SERVICE <id>.
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
issue: #10241
