zlei6/ragflow
1
0
Fork 0
mirror of https://github.com/infiniflow/ragflow.git synced 2026-07-11 22:25:41 +08:00
Code Issues Packages Projects Releases Wiki Activity
7,298 Commits 2 Branches 69 Tags
0aa2993d55c2da8d43ccfe3fd1d968737701c541
Commit Graph

2215 Commits

Author SHA1 Message Date
balibabu
d317742975 Feat: Rewrite wiki template with reui (#16797) 2026-07-10 15:44:04 +08:00
Jack
7db39822db Feature: user select pipeline support (#16788) 
### Summary

Feature: user select pipeline support
 2026-07-10 14:30:28 +08:00
chanx
868e524f29 fix: pass ownerTenantId to LLMLabel and related components for improved model fetching (#16800) 2026-07-10 13:27:14 +08:00
chanx
8236f2cabb Fix: update LayoutRecognizeFormField to accept ownerTenantId and refactor model handling in LLM requests (#16781) 2026-07-10 09:36:25 +08:00
chanx
0095fa048f fix: Show full text on hover when text overflows the cards on the list page. (#16787) 2026-07-10 09:25:05 +08:00
balibabu
0083ad0deb Feat: Add a data compilation layer. (#16777) 
### Summary

Feat: Add a data compilation layer.
 2026-07-09 17:49:16 +08:00
chanx
a9420a7832 Fix: resolve shared embedding/LLM model selection errors (#16773) 2026-07-09 15:17:49 +08:00
Lynn
1430d0e431 Fix: provider name (#16733) 2026-07-09 10:19:10 +08:00
balibabu
575984877f Fix: Rapid clicking results in multiple message requests being sent. (#16739) 2026-07-09 09:57:54 +08:00
euvre
3ec9187cd2 fix(web): prevent 'last saved at' label from vertical stacking in agent home card (#16756) 2026-07-09 09:46:03 +08:00
chanx
080dd84fed Feat: apply prose typography styling to markdown preview (#16752) 2026-07-09 09:45:52 +08:00
chanx
36f053a248 fix: Fixed the empty state styling on the home page. (#16755) 2026-07-09 09:45:43 +08:00
euvre
70019810a1 fix(web): show memory owner name in shared memory card (#16751) 2026-07-08 20:13:07 +08:00
euvre
a41fef49d0 fix(web): hide folder tab in agent JSON import uploader (#16754) 2026-07-08 20:11:55 +08:00
chanx
3d167204e7 fix: issue with memory error message display (#16738) 2026-07-08 16:47:14 +08:00
chanx
b9432bb43f Feat: add filter in chat and search page (#16707) 2026-07-08 16:47:01 +08:00
euvre
699a25c19c fix(service): allow updating memory_type when memory is empty (#16668) 2026-07-08 10:03:58 +08:00
Lynn
0ae5961e1c Feat: v0.27.0 model provider (#16604) 2026-07-08 09:47:29 +08:00
chanx
5236c8f659 fix: update similarity threshold fallback to use nullish coalescing (#16700) 2026-07-07 17:03:03 +08:00
chanx
b7945f3a64 Fix: Referenced files not displaying. (#16696) 2026-07-07 16:29:17 +08:00
chanx
dd2f27d6a3 fix: Restrict the agent to using memory compatible with the embedding model. (#16699) 2026-07-07 16:28:58 +08:00
chanx
f082675e6f Fix: Prevent text overflow in confirm delete dialog (#16689) 2026-07-07 14:50:25 +08:00
chanx
5aa3e81a93 fix: remove duplicate error toast on memory update failure (#16690) 2026-07-07 14:50:09 +08:00
Wang Qi
a0bda639e0 Fix Agent Chat not working (#16688) 
Follow on this PR: #15484 it break the Agent chat
 2026-07-07 12:10:52 +08:00
天海蒼灆
318045dda5 feat(agent): support JSON object input on begin node (#16685) 
### Summary

Add object as a begin-node parameter type with JSON editor UI, webhook
schema support, and backend parsing in UserFillUp.

Co-authored-by: Cursor <cursoragent@cursor.com>
 2026-07-07 11:40:57 +08:00
Hz_
863b35db7f fix(go-agent-web): correct BGPT canvas form watcher usage (#16682) 
## Summary

Fixes a page crash when opening the BGPT node configuration in the
canvas.

## Root Cause

BGPT was using the tool-form watcher call pattern in a normal canvas
component form.

Tool forms use:

useWatchFormChange(form)

Canvas component forms use:

useWatchFormChange(node?.id, form)

Tool is not equal to component. The BGPT canvas component imported the
component-level hook but called it like a tool-form hook, so the form
argument became undefined and React Hook Form tried to read control from
a null context.

## Change

Updated the BGPT canvas form to pass the node id and form instance
correctly.

## Validation

Ran ESLint for the changed file:
npx eslint src/pages/agent/form/bgpt-form/index.tsx

<img width="1369" height="1184" alt="image"
src="https://github.com/user-attachments/assets/a40c5202-7394-4f26-9da2-08329dcc7fbf"
/>
 2026-07-07 11:08:50 +08:00
Rodger Blom
d8cefcf052 feat: add native Dutch language support for BM25 tokenization (#14140) 
## Summary
- Add language-aware Snowball stemmer to `RagTokenizer` supporting 16
languages (Dutch, German, French, Spanish, etc.)
- Thread the KB `language` parameter through the full tokenization
pipeline (14 parser modules + task executor)
- Add Dutch to the frontend language lists and cross-language form

## Problem
RAGFlow uses the English Porter stemmer + WordNet lemmatizer for **all**
BM25 tokenization, regardless of the knowledge base language setting.
This produces incorrect stems for non-English text. For example:

| Dutch word | Dutch stemmer | English Porter |
|---|---|---|
| documenten | document | documenten (unchanged!) |
| gebruikers | gebruiker | gebruik (over-stemmed) |
| instellingen | instell | instellingen (unchanged!) |

This degrades BM25 recall for any non-English knowledge base.

## Solution
NLTK already ships Snowball stemmers for 16 languages. This PR:

1. **`rag/nlp/rag_tokenizer.py`**: Overrides `tokenize()` with
`set_language()` and `_normalize_token()` that selects the correct NLTK
Snowball stemmer. Falls back to Porter for unmapped languages (Chinese,
Japanese, Korean, etc. — these use character-based tokenization anyway).
2. **`rag/nlp/__init__.py`** + **14 `rag/app/*.py` parsers** +
**`rag/svr/task_executor.py`**: Threads the `language` parameter through
`tokenize()`, `tokenize_chunks()`, `tokenize_table()`, and all callers.
3. **Frontend**: Adds Dutch (`Nederlands`) to `LanguageList`,
`LanguageMap`, `LanguageAbbreviationMap`, `LanguageTranslationMap`,
cross-language form field, and `en.ts` locale.

## Backward Compatibility
- Default language is `"English"`, preserving existing behavior for all
current users
- Languages without a Snowball stemmer mapping fall back to Porter (no
change)
- No new dependencies — NLTK Snowball is already bundled
 2026-07-06 23:39:56 +08:00
Wang Qi
57625c919a Fix Tag weight should be greater than 0 (#16657) 2026-07-06 16:06:27 +08:00
balibabu
290ab557a5 Fix: Layout of the agent prompt dropdown menu is messed up. (#16653) 2026-07-06 14:48:42 +08:00
chanx
c9f064d5fd fix(metadata): inline value edits not persisted to backend (#16655) 2026-07-06 13:04:04 +08:00
monsterDavid
7da4f200e5 fix(agent): enable MCP file preview via doc_id (#15399) 
## Summary
MCP-wrapped agents could only force-download files looked up by
`doc_id`. This adds an explicit preview path and inline response headers
for previewable file types.

- **New** `GET /api/v1/agents/attachments/{attachment_id}/preview` —
inline preview for PDFs, images, and other safe types (pass `ext` and/or
`mime_type`)
- **Improved** `GET /api/v1/documents/{doc_id}/preview` — sets inline
disposition using the document filename
- **Improved** attachment download routing — resolves `mime_type` /
`ext` query params (no default `markdown`), supports
`disposition=inline`
- **DocGenerator output** — includes URL-encoded `preview_url` for MCP
clients
- **Legacy `/document/download/...` aliases** — still use download
semantics; MCP clients should call `/preview` explicitly

Fixes #15398

## Test plan
- [x] `pytest test/unit_test/api/utils/test_file_response_headers.py`
(6/6)

---------

Co-authored-by: MkDev11 <mkdev11@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Ling Qin <qinling0210@163.com>
 2026-07-03 19:56:01 +08:00
Jin Hai
1880e65e99 Go: refactor (#16602) 
### Summary

1. update doc
2. refactor route code

---------

Signed-off-by: Jin Hai <haijin.chn@gmail.com>
 2026-07-03 17:00:43 +08:00
chanx
79518973e5 Fix: optimize folder data handling in MoveDialog component (#16580) 2026-07-03 16:13:57 +08:00
euvre
4effd057f0 i18n: localize visual input file label in agent form (#16594) 2026-07-03 15:31:27 +08:00
euvre
7b341539e7 fix: prevent exporting empty MCP server selection (#16589) 2026-07-03 14:22:17 +08:00
euvre
918229613a fix: prevent duplicate 'skills' and '.knowledgebase' folders caused by race conditions (#16568) 2026-07-03 12:06:45 +08:00
Lynn
bc54903bf6 Fix: display model_id in memory_list (#16567) 2026-07-02 20:28:27 +08:00
chanx
9a6d30bfe6 Fix: send agent log date filters as local wall-clock strings (#16575) 2026-07-02 20:23:15 +08:00
chanx
16b8c79a2b Fix: hide model settings button and related functionality (#16563) 2026-07-02 17:49:52 +08:00
chanx
2ef78189ce Fix: pass mcp to useExportMcp for correct JSON export filename (#16564) 2026-07-02 17:49:46 +08:00
chanx
c44d56f1bb Fix: enhance reference handling in SessionChat component (#16571) 2026-07-02 17:48:48 +08:00
euvre
3195d6fa89 fix: improve Normal role badge visibility with proper styling (#16528) 2026-07-02 10:47:01 +08:00
euvre
81cfcdf2d3 feat(frontend): add AuthenticatedImg component for authorized image requests (#16525) 2026-07-01 17:02:44 +08:00
qinling0210
7862f69f39 Implement chat completions in go (#16491) 
### Summary

POST   /api/v1/chat/completions
 2026-07-01 15:52:52 +08:00
connerlambden
9bf57600cf feat(agent): add BGPT structured literature evidence search tool (#16050) 
## Summary

Adds a first-class **BGPT** Agent tool (backend + UI) in response to
[#15997](https://github.com/infiniflow/ragflow/issues/15997#issuecomment-4703864227).

BGPT calls `POST https://bgpt.pro/api/mcp-search` and returns structured
study evidence from full-text papers — not just titles/abstracts. Each
result is formatted for RAGFlow citations with:

- methods
- sample size / population
- results
- limitations
- conflicts of interest
- data availability
- study blind spots
- `how_to_falsify`

## Why this shape

- Mirrors existing literature tools (`PubMed`, `ArXiv`) and HTTP tools
(`SearXNG`).
- Works on the free tier (no API key required for first 50 results).
- Optional `api_key` and `days_back` in the node/tool config.
- Surfaces both `formalized_content` and raw `json` outputs (like
SearXNG).

## Files

- `agent/tools/bgpt.py` — REST client + evidence formatter
- Frontend: Operator enum, forms, tool picker, canvas accordion, en/zh
locales, icon

## Demo / docs

Runnable claim-interrogation demo:
https://github.com/connerlambden/bgpt-mcp/blob/main/EVIDENCE_DEMO.md

## Test plan

- [ ] Add BGPT node on Agent canvas, run query `GLP-1 alcohol craving`,
verify `formalized_content` includes limitations/COI fields
- [ ] Add BGPT as Agent sub-tool under Search, verify tool-calling works
- [ ] Confirm empty query / try-run returns gracefully
- [ ] Optional: paid-tier `api_key` path

---------

Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Zhichang Yu <yuzhichang@gmail.com>
 2026-07-01 13:52:24 +08:00
Harsh Kashyap
572f1ea9f4 fix(web): sanitize agent rerun modal HTML against stored XSS (#16516) 2026-07-01 13:38:31 +08:00
Lynn
400476f0b3 Feat: SoMark (#16482) 
Follow #15486
Co-authored-by: limuting <limuting233@gmail.com>
Co-authored-by: lutianyi <lutianyi233@163.com>
Co-authored-by: justinychuang <huangyicheng@soulcode.cn>
Co-authored-by: maybehokori <138367708+maybehokori@users.noreply.github.com>
 2026-07-01 13:29:28 +08:00
Lynn
b53b693f22 Fix: CI (#16504) 
### Summary

Fix race condition in parallel lefthook hooks causing ETXTBSY error
 2026-06-30 22:14:11 +08:00
Wang Qi
d2ecd57c59 Fix: UI cannot start up (#16497) 2026-06-30 17:09:09 +08:00
Zhichang Yu
4c54cefd29 Port 14 upstream agent security / correctness fixes to Go canvas (#16455) 
Mirrors 14 merged upstream PRs into the Go agent port.

PRs ported:
  - #15609 ExeSQL SSRF guard + DNS pin
  - #15436 HTTP timeout on external API tools
  - #16363 be_output restore + DeepL error path
  - #15644 switch no longer matches empty condition
  - #15374 session_id bind to path agent_id (DAO idor guard)
  - #16169 sandbox artifact ownership gate
  - #15457 tenant ownership on agentbots
  - #15145 rerun agent document access check
- #15446 thinking switch (component portion; provider policy lives in
internal/llm)
  - #15426 Invoke URL/proxy SSRF + DNS pin + no-redirects
  - #15238 agentbot thinking-logs beta endpoint
  - #14589 UserFillUp SSE event propagation
  - #14890 anonymous webhook opt-in
- #15068 PipelineChunker new component (text/file_ref/parser_id
dispatch; file-format extraction is a follow-up)

40 files, +2355 / -58 lines. 33 new tests, all targeted package suites
pass (1721 + 4 skipped); 1 pre-existing flaky test unrelated.
 2026-06-30 16:28:48 +08:00