## Summary
- **Collapsible thinking**: Replace `<section>` with `<details>` for
`<think>` content, so model thinking output is collapsed by default
(click to expand). Works for all models that output `<think>` tags
(Qwen3, DeepSeek, Gemini, Claude, etc.).
- **Fix double thinking tags**: When reasoning/deep research mode is
enabled in knowledge base chat, both the retrieval progress and model
thinking were wrapped in `<think>` tags, producing two "Thinking..."
blocks. Now retrieval progress uses a dedicated `<retrieving>` tag
rendered as a separate "Retrieving..." collapsible with a distinct green
accent.
### Before
- Thinking content displayed as flat gray-bordered `<section>`,
occupying significant screen space
- Deep research + model thinking both use `<think>` → two identical
"Thinking..." blocks
### After
- Thinking content collapsed by default in a `<details>` element, click
"Thinking..." to expand
- Deep research shows "Retrieving..." (green border), model thinking
shows "Thinking..." (gray border)
## Changes
**Backend (`api/db/services/dialog_service.py`)**
- Deep research callback: replace `start_to_think`/`end_to_think` marker
flags with direct `<retrieving>`/`</retrieving>` answer text
**Frontend**
- `web/src/utils/chat.ts`: `replaceThinkToSection()` now uses
`<details>` instead of `<section>`; add new
`replaceRetrievingToSection()`
- 4 tsx files: import and pipe `replaceRetrievingToSection`, whitelist
`details`, `summary`, `retrieving` in DOMPurify `ADD_TAGS`
- 4 less files: `section.think` → `details.think` with `<summary>`
styles; add `details.retrieving` with green accent; dark mode and RTL
variants
## Test plan
- [ ] Open a chat WITHOUT knowledge base, ask a question to a model with
thinking (e.g. Qwen3) → thinking content should be collapsed by default,
click "Thinking..." to expand
- [ ] Open a chat WITH knowledge base and reasoning enabled, ask a
question → "Retrieving..." (green) shows retrieval progress,
"Thinking..." (gray) shows model thinking, each independently
collapsible
- [ ] Verify dark mode renders correctly for both collapsible blocks
- [ ] Verify RTL layout renders correctly
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-authored-by: wanghualoong <wanghualoong@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Fixes#14562
## Problem
LLMs like DeepSeek V4 Flash and Qwen3-MAX return \\( and \\[
(double backslash) in LaTeX output. The preprocessLaTeX() function
only handled single backslash delimiters, so equations showed as raw
text.
HTML entities like < and > were also not decoded.
## Solution
Added normalization step before existing delimiter conversion:
- \\( → \( and \\[ → \[
- < → < and > → > and & → &
---------
Co-authored-by: Vivek <viveksantoshkumardubey@email.com>
### What problem does this PR solve?
Fix: LaTeX formulas cannot be displayed on the chat page.
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
### What problem does this PR solve?
This PR adds comprehensive **Right-to-Left (RTL) language support**,
primarily targeting Arabic and other RTL scripts (Hebrew, Persian, Urdu,
etc.).
Previously, RTL content had multiple rendering issues:
- Incorrect sentence splitting for Arabic punctuation in citation logic
- Misaligned text in chat messages and markdown components
- Improper positioning of blockquotes and “think” sections
- Incorrect table alignment
- Citation placement ambiguity in RTL prompts
- UI layout inconsistencies when mixing LTR and RTL text
This PR introduces backend and frontend improvements to properly detect,
render, and style RTL content while preserving existing LTR behavior.
#### Backend
- Updated sentence boundary regex in `rag/nlp/search.py` to include
Arabic punctuation:
- `،` (comma)
- `؛` (semicolon)
- `؟` (question mark)
- `۔` (Arabic full stop)
- Ensures citation insertion works correctly in RTL sentences.
- Updated citation prompt instructions to clarify citation placement
rules for RTL languages.
#### Frontend
- Introduced a new utility: `text-direction.ts`
- Detects text direction based on Unicode ranges.
- Supports Arabic, Hebrew, Syriac, Thaana, and related scripts.
- Provides `getDirAttribute()` for automatic `dir` assignment.
- Applied dynamic `dir` attributes across:
- Markdown rendering
- Chat messages
- Search results
- Tables
- Hover cards and reference popovers
- Added proper RTL styling in LESS:
- Text alignment adjustments
- Blockquote border flipping
- Section indentation correction
- Table direction switching
- Use of `<bdi>` for figure labels to prevent bidirectional conflicts
#### DevOps / Environment
- Added Windows backend launch script with retry handling.
- Updated dependency metadata.
- Adjusted development-only React debugging behavior.
---
### Type of change
- [x] Bug Fix (non-breaking change which fixes RTL rendering and
citation issues)
- [x] New Feature (non-breaking change which adds RTL detection and
dynamic direction handling)
---------
Co-authored-by: 6ba3i <isbaaoui09@gmail.com>
Co-authored-by: Ahmad Intisar <ahmadintisar@Ahmads-MacBook-M4-Pro.local>
Co-authored-by: Ahmad Intisar <168020872+ahmadintisar@users.noreply.github.com>
Co-authored-by: Liu An <asiro@qq.com>
## Summary
Fixes the initial enabled/disabled state of chat variable checkboxes by
correcting a helper function that previously always returned .
## Problem
in had two statements:
Because of the early , the function always returned , so all chat
variable checkboxes were initially disabled regardless of the field.
This also made the helper inconsistent with , which enables all fields
by default except .
## Fix
Update to use the same condition as :
This ensures:
- All chat variable checkboxes are enabled by default
- remains the only field disabled by default
- Behavior is consistent between the helper and the checkbox map
initialization in .
No API or backend changes are involved; this is a small, isolated
frontend bugfix.
### What problem does this PR solve?
- Use negative lookbehind (?<![a-zA-Z]) so \] and \) inside commands
(e.g. \right], \big)) are not treated as block/inline delimiters
- Use greedy matching to capture up to the last valid delimiter, fixing
truncated formulas (e.g. C_{seq}(y|x) = \frac{1}{|y|} ...)
- Add unit tests for preprocessLaTeX
Closes#13134
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
### What problem does this PR solve?
Feat: Users can chat directly without first creating a conversation.
#11768
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
Feat: Delete useless knowledge base, chat, and search files. #10427
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
Feat: Fixed the issue where the parameters could not be set after
switching the large model parameter template. #8282
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
Feat: Reset the default values of large model parameters
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
Feat: Display inline (non-quoted) images in the chat and search modules
#7623
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
Feat: Disable Max_token by default #5283
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
Feat: Write the thinking style in the MarkdownContent layer #4930
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
Feat: Display Think for Deepseek R1 model #4903
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
[remarkjs/react-markdown/issues/785](https://github.com/remarkjs/react-markdown/issues/785)
Fix: Fixed an issue where math formulas could not be displayed correctly
#4405
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
### What problem does this PR solve?
Fix: In order to distinguish the keys of a pair of messages, add a
prefix to the id when rendering the message. #4409
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
### What problem does this PR solve?
Feat: Exclude reference from the data returned by the conversation/get
interface #3909
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
fix: Fixed an issue where the first message would be displayed when
sending the second message #2625
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
- [ ] New Feature (non-breaking change which adds functionality)
- [ ] Documentation Update
- [ ] Refactoring
- [ ] Performance Improvement
- [ ] Other (please describe):
### What problem does this PR solve?
feat: Select derived messages from backend #2088
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
feat: Regenerate chat message #2088
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
feat: Send message with uuid #2088
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
feat: Fetch conversation list by @tanstack/react-query
feat: Displays error message that task_executor does not exist
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
