V4.14.10 dev (#6674)

* feat: model config with brand-new price calculate machanism (#6616)

* fix: image read and json error (Agent) (#6502)

* fix:
1.image read
2.JSON parsing error

* dataset cite and pause

* perf: plancall second parse

* add test

---------

Co-authored-by: archer <545436317@qq.com>

* master message

* remove invalid code

* wip: model config

* feat: model config with brand-new price calculate machanism

* merge main branch

* ajust calculate way

* ajust priceTiers resolve procession

* perf: price config code

* fix: default price

* fix: test

* fix: comment

* fix test

---------

Co-authored-by: YeYuheng <57035043+YYH211@users.noreply.github.com>
Co-authored-by: archer <545436317@qq.com>

* wip: fix modal UI (#6634)

* wip: fix modal UI

* fix: maxInputToken set

* chore: add price unit for non llm models

* chore: replace question mark icon with beta tag (#6672)

* feat:rerank too long; fix:rerank ui(agent),embedding returns 0 (#6663)

* feat:rerank too long; fix:rerank ui(agent),embedding returns 0

* rerank

* fix:rerank function

* perf: rerank code

* fix rerank

* perf: model price ui

---------

Co-authored-by: archer <545436317@qq.com>

* remove llmtype field

* revert model init

* fix: filed

* fix: model select filter

* perf: multiple selector render

* remove invalid checker

* remove invalid i18n

* perf: model selector tip

* perf: model selector tip

* fix cr

* limit pnpm version

* fix: i18n

* fix action

* set default mintoken

* update i18n

* perf: usage push

* fix:rerank model ui (#6677)

* fix: tier match error

* fix: testr

---------

Co-authored-by: Ryo <whoeverimf5@gmail.com>
Co-authored-by: YeYuheng <57035043+YYH211@users.noreply.github.com>
This commit is contained in:
Archer
2026-03-30 10:05:42 +08:00
committed by GitHub
parent 03dd9c00a8
commit 3f4400a500
108 changed files with 3774 additions and 1469 deletions
@@ -13,12 +13,17 @@ description: 'FastGPT V4.14.10 更新说明'
1. 增加 OpenSandbox docker 部署方案及适配,并支持通过挂载 volumn 进行数据持久化。
2. 飞书发布渠道,支持流输出。
3. 目录最大上限,可通过环境变量配置。
4. rerank 模型上限配置,避免超出单条 document 上限导致 rerank 失败。
5. 增加 LLM 梯度计量计费模式,同时统一计费推送方式。
## ⚙️ 优化
1. 工作流 runtime,减少计算复杂。
2. 增加一些对于大变量的计算限制,避免计算复杂度过高导致线程阻塞。
3. 移除模型配置里“用于知识库文件处理”、“用于问题分类”等配置,统一增加“测试模型“标志。测试模型会有特殊标识,并且仅可在 ai chat 中使用,其余场景将会过滤。
## 🐛 修复
1. 子工作流的全局变量默认值未生效。
1. 子工作流的全局变量默认值未生效。
2. agent 模式下已配的 rerank 模型不显示。
3. bge-m3 embedding 向量模型输出都为 0 的问题。