Commit Graph
578 Commits
Author SHA1 Message Date
Jin Hai 834a994b24 add more glm models (#22087)
Signed-off-by: Jin Hai <haijin.chn@gmail.com>
2026-11-06 14:45:47 +08:00
Caldalis abafb52286 feat(parser): add MiniCPM-OCR as a PDF parser (#21785) 2026-11-06 14:29:41 +08:00
jimmyplup 19e5abc971 feat: add Synap as a model provider (#21818) 2026-11-03 10:33:46 +08:00
fenglingxiansheng 14336bba2d fix: SoMark web page is missing (#21710) 2026-11-02 12:05:44 +08:00
kk710 89199670f7 feat(vastbase): support Vastbase G100 as the metadata database in Go (#20913) 2026-11-02 10:16:16 +08:00
Aditya Batra 6ec4e3d80b feat: add Wally as a model provider (#21719) 2026-10-31 10:16:04 +08:00
Jin Hai c81b2f0052 Add audit log stub (#21712)
Signed-off-by: Jin Hai <haijin.chn@gmail.com>
2026-10-30 11:30:47 +08:00
Nicho Deng 321dfe84ee feat: add cheapaimodel as a model provider (#21612) 2026-10-29 20:33:24 +08:00
ricardasbitinas-jpg fdf9db6fcd feat: add Hostinger Router as a model provider (#21419) 2026-10-29 16:47:41 +08:00
Anand B a5ad435e17 Fix: remove retired OpenAI models from the preset model list (#21685) 2026-10-29 16:38:34 +08:00
Tony Davis f64d6699a0 feat: add Obscurify.ai as a model provider (#21639) 2026-10-29 13:52:19 +08:00
xugangqiang 9742be8b5b refactor(server): collapse duplicate ingestor concurrency resolver and drop redundant Max prefix (#21588) 2026-10-28 15:30:36 +08:00
JinJing-tomo e0a52fc856 feat: add WorldRouter model provider (#21603) 2026-10-28 15:20:33 +08:00
Lynn 61eea6e8b0 Fix: remove sso related code (#21584) 2026-10-28 12:11:47 +08:00
CloudGPU 0cfa864cc6 feat: add CloudGPU as a model provider (#21543) 2026-10-27 16:07:26 +08:00
Nirupam Bhowmick 396525c7e0 feat: add Runcrate as a model provider (#21446) 2026-10-26 12:12:33 +08:00
Caobaining1 5a6b094069 feat(decision-model): using Jev-style decision-model for agentic-rag (#21250) 2026-10-21 17:24:42 +08:00
Tim Wang aae7f15080 feat(models): add New API model driver (#21226) 2026-10-21 15:28:40 +08:00
Lawrence Li d893751b7d fix(models): restore gpustack rerank and audio support (#21233)
ref gpustack/gpustack#6515
2026-10-21 15:07:17 +08:00
Sbaaoui Idriss 2bbb984b5f feat(models): add Radeon Cloud provider for GO (#21163) 2026-10-21 10:04:24 +08:00
Studio e2afaaf3bb feat: add AI Stupid Level as a model provider (#21185) 2026-10-20 15:24:12 +08:00
Adam Dalloul ccd2657dc9 feat: add EmpirioLabs AI as a model provider (#21192) 2026-10-20 13:49:56 +08:00
Yingfeng 7eb673bfc3 Fix claim/evidence usage in agentic RAG (#21170) 2026-10-19 21:03:36 +08:00
Joseph Perla f9138d2c39 feat: add TrustedRouter as a model provider (#21105) 2026-10-19 19:47:29 +08:00
Xavierando ea0d7f21ae feat(go): complete the Mistral /v1/ocr client and list mistral-ocr-latest as an ocr model (#20531) 2026-10-16 18:30:39 +08:00
qinling0210 54cb269d0d feat(infinity): implement SearchByRegexp for the agent grep tool (#21068) 2026-10-16 17:15:53 +08:00
britz ee12f193ed fix(aliyun): transcribe audio via DashScope multimodal-generation (#20975) 2026-10-15 13:43:23 +08:00
xugangqiang fa1a071967 refactor(config): resolve CLI/env/file precedence in one place with final-value validation (#20934) 2026-10-15 10:01:57 +08:00
Lynn bf6ec9c839 Fix: handle redundant v1 in base_url (#20952) 2026-10-14 19:27:22 +08:00
Caobaining1 25481cdff3 Feat/migrates navigation tools form old mode to agentic mode (#20735) 2026-10-13 13:46:16 +08:00
Demon Route 7730e07714 Add Demon Route as an LLM provider (#20853) 2026-10-13 10:37:26 +08:00
Wang Qi c89a98f347 Fix Agentic RAG does not do web search (#20793) 2026-10-12 16:14:08 +08:00
Sbaaoui Idriss 2d1385eba5 refactor(sandbox): remove obsolete configuration (#20809) 2026-10-12 15:16:54 +08:00
Zhichang Yu 9b5d73bd7c Go: handle llama-server over-limit rejections in embed and rerank (#20784) 2026-10-12 10:55:05 +08:00
Sbaaoui Idriss f6477b1977 fix(chat): preserve images across reasoning and conversation history (#20728) 2026-10-12 10:15:41 +08:00
Jin Hai bc2934b89d Go: fix xunfei model (#20757)
Signed-off-by: Jin Hai <haijin.chn@gmail.com>
2026-10-11 08:53:26 +08:00
Sourav Rajvi 38e63566be feat(go): MonkeyOCR env keys and tenant provisioning (#20468) 2026-10-10 17:12:18 +08:00
Caobaining1 c5c75b2c5d fix(agentic_rag): number citations from a per-turn evidence registry (#20677) 2026-10-09 19:47:00 +08:00
Lynn 3cbde9ad8e Fix: remove rank in mws config (#20676) 2026-10-09 19:11:54 +08:00
adws 83c6ebdf2b fix(wiki): remove duplicate Markdown storage (#20597) 2026-10-09 15:48:06 +08:00
Caobaining1 d50f1f02d6 feat(agentic_rag): add metadata_search tool to eino smart-reasoning path (#20645) 2026-10-09 14:52:44 +08:00
qinling0210 ca22c25245 fix(infinity): restore deprecated knowledge-compilation columns in the chunk mapping (#20646) 2026-10-09 14:12:39 +08:00
kk710 ef150bc2cd feat(vastbase): add Vastbase G100 document engine support with go (#20419) 2026-10-08 20:02:37 +08:00
buua436 fceacbfb27 feat: add SiliconFlow model catalog entries (#20596) 2026-10-08 17:45:20 +08:00
Wang Qi f307aba34f Refactor: knowledge compilation field name refactor (#20587) 2026-10-08 16:05:27 +08:00
Felix 536f8fa7a6 feat: add Opper as a model provider (#20577) 2026-10-08 11:52:15 +08:00
Zhichang Yu 98b48a0857 Port agentic RAG to Go, expose it as a chat mode, and add per-dialog failover (#20503)
## Background

This branch started as a focused fix to agentic RAG regexp retrieval
semantics (`f80556585`) and grew into the full agentic RAG path. The
title no longer describes the contents, so it has been rewritten.

The PR now covers three largely independent lines of work:

### 1. The agentic RAG is reachable from the UI

`internal/agentic_rag` (the eino-ADK ReAct explorer) was already built
and wired, but only reachable by hand-crafting an `agent_mode` kwarg. It
is now the sixth option in the chat mode selector (`reasoning` level 5).

One subtlety worth stating plainly: **levels 1-4 and level 5 are not the
same agent.** Levels 1-4 go through `internal/rag/agentic-rag` (the
harness graph) with a depth chosen by `harnessModeForLevel`; level 5
switches engines outright to `internal/agentic_rag`. That is why level 5
must never reach `harnessModeForLevel` — its `level >= 4` case would
silently answer "ultra" for a level outside its domain.

### 2. Per-dialog failover chain

`agenticModelChain` resolved exactly one model and the caller then used
`chain[0]`, so a "chain" was never more than a single element. A dialog
can now configure an ordered list of fallback models in Chat Settings,
handed to `NewFailoverEinoChatModel` (sticky cursor plus a 30s
full-chain cooldown).

The list lives in the dialog's own `llm_setting.failover_llm_ids`, so no
new table is involved. A member that no longer resolves is skipped with
a warning rather than failing the turn.

Also removed: `tenant_model_group` / `tenant_model_group_mapping`, which
nothing ever read (the DAOs were constructed but never called, and no
frontend or Python code referenced the concept). Their removal takes an
explicit drop migration with it, plus the account-deletion cascade that
queried them.

### 3. A hung MiniMax stream (independent of the agentic work)

With any mode selected, a chat rendered its whole answer and then sat on
"thinking" forever. Root cause is `minimax.go:256`: MiniMax sends `data:
[DONE]` but leaves the HTTP connection open, and the code waited for the
scanner goroutine's EOF *after* `HandleStreamingResponse` had already
returned. That receive can only end when `streamCallTimeout` (20
minutes) expires.

Diagnosed by capturing a real SSE stream (the complete answer arrives,
the terminal `final: true` never does) and a goroutine dump (6 requests
parked in `chan receive`).

## Two review findings fixed on the way through

- **KB-scope authorization**: the agentic branch bypassed quote
resolution, and an empty KB scope made `buildBoolQueryFromCondition`
drop the `kb_id` filter — so a citation could resolve a chunk belonging
to a different KB in the same tenant. The agentic branch now requires a
non-empty scope and otherwise falls through to the regular path.
- **Stale documentation**: `agentic-rag-failover-groups.md` described
the "automatically include every tenant model" strategy that upstream
had already removed. It was rewritten for the per-dialog scope and then
dropped entirely, since the design now lives in the code it describes.

## Verification

- `bash build.sh --test`: `admin`, `dao`, `service`, `service/dataset`
and `entity/models` all pass
- The MiniMax fix was verified end-to-end against a live server: before,
the turn hung indefinitely; after, it completes in **1.9s** with `final:
true` present
- Frontend: 9 tests added; type-check and lint clean on the touched
files

## Not included

- **Attachment support in agentic mode.** Text attachments could be
appended safely, but images have no safe fix: the agent's toolset is
built around corpus retrieval and has no image input channel. Fixing
only the text path would leave the feature half-supported and harder to
diagnose than now. Planned as a follow-up PR, with the design synced
here first.
- Tool-calling is not enforced as a group constraint. `is_tools` is a
provider-declared flag rather than a measured capability (187 of 659
chat models do not declare it), so gating on it would reject working
configurations while admitting broken ones.
2026-10-02 23:00:16 +08:00
buua436 c8a00c6f6d refactor: use model tokenizer for rerank (#20422) 2026-09-30 14:30:32 +08:00
qinling0210 02569aee9d fix(mappings): support any embedding dimension on Elasticsearch (#20368) 2026-09-29 17:46:45 +08:00
xugangqiang 18d9750981 feat(config): make ingestor worker count (K) and page concurrency (N) configurable (#20121) 2026-09-29 14:19:24 +08:00