FastGPT/dataset.md at a8d456f4487f668c4a92e22c4c302ba1d78a4ba9

mirror of https://github.com/labring/FastGPT.git synced 2025-07-23 05:12:39 +00:00

Files

Archer 10d8c56e23 V4.8.18 feature (#3565 )

* feat: org CRUD (#3380)

* feat: add org schema

* feat: org manage UI

* feat: OrgInfoModal

* feat: org tree view

* feat: org management

* fix: init root org

* feat: org permission for app

* feat: org support for dataset

* fix: disable org role control

* styles: opt type signatures

* fix: remove unused permission

* feat: delete org collaborator

* perf: Team org ui (#3499)

* perf: org ui

* perf: org ui

* feat: org auth for app & dataset (#3498)

* feat: auth org resource permission

* feat: org auth support for app & dataset

* perf: org permission check (#3500)

* i18n (#3501)

* name

* i18n

* feat: support dataset changeOwner (#3483)

* feat: support dataset changeOwner

* chore: update dataset change owner api

* feat: permission manage UI for org (#3503)

* perf: password check;perf: image upload check;perf: sso login check (#3509)

* perf: password check

* perf: image upload check

* perf: sso login check

* force show update notification modal & fix login page text (#3512)

* fix login page English text

* update notification modal

* perf: notify account (#3515)

* perf(plugin): improve searXNG empty result handling and documentation (#3507)

* perf(plugin): improve searXNG empty result handling and documentation

* 修改了文档和代码部分无搜索的结果的反馈

* refactor: org pathId (#3516)

* optimize payment process (#3517)

* feat: support wecom sso (#3518)

* feat: support wecom sso

* chore: remove unused wecom js-sdk dependency

* fix qrcode script (#3520)

* fix qrcode script

* i18n

* perf: full text collection and search code;perf: rename function (#3519)

* perf: full text collection and search code

* perf: rename function

* perf: notify modal

* remove invalid code

* perf: sso login

* perf: pay process

* 4.8.18 test (#3524)

* perf: remove local token

* perf: index

* perf: file encoding;perf: leave team code;@c121914yu perf: full text search code (#3528)

* perf: text encoding

* perf: leave team code

* perf: full text search code

* fix: http status

* perf: embedding search and vector avatar

* perf: async read file (#3531)

* refactor: team permission  manager (#3535)

* perf: classify org, group and member

* refactor: team per manager

* fix: missing functions

* 4.8.18 test (#3543)

* perf: login check

* doc

* perf: llm model config

* perf: team clb config

* fix: MemberModal UI (#3553)

* fix: adapt MemberModal title and icon

* fix: adapt member modal

* fix: search input placeholder

* fix: add button text

* perf: org permission (#3556)

* docs:用户答疑的官方文档补充 (#3540)

* docs:用户答疑的官方文档补充

* 问题回答的内容修补

* share link random avatar (#3541)

* share link random avatar

* fix

* delete unused code

* share page avatar (#3558)

* feat: init 4818

* share page avatar

* feat: tmp upgrade code (#3559)

* feat: tmp upgrade code

* fulltext search test

* update action

* full text tmp code (#3561)

* full text tmp code

* fix: init

* fix: init

* remove tmp code

* remove tmp code

* 4818-alpha

* 4.8.18 test (#3562)

* full text tmp code

* fix: init

* upgrade code

* account log

* account log

* perf: dockerfile

* upgrade code

* chore: update docs app template submission (#3564)

---------

Co-authored-by: a.e. <49438478+I-Info@users.noreply.github.com>
Co-authored-by: Finley Ge <32237950+FinleyGe@users.noreply.github.com>
Co-authored-by: heheer <heheer@sealos.io>
Co-authored-by: Jiangween <145003935+Jiangween@users.noreply.github.com>

2025-01-11 15:15:38 +08:00

2.2 KiB

Raw Blame History

title, description, icon, draft, toc, weight

title	description	icon	draft	toc	weight
知识库使用问题	常见知识库使用问题	quiz	false	true	910

上传的文件内容出现中文乱码

将文件另存为 UTF-8 编码格式。

知识库配置里的文件处理模型是什么？与索引模型有什么区别？

文件处理模型：用于数据处理的【增强处理】和【问答拆分】。在【增强处理】中，生成相关问题和摘要，在【问答拆分】中执行问答对生成。
索引模型：用于向量化，即通过对文本数据进行处理和组织，构建出一个能够快速查询的数据结构。

基于知识库的查询，但是问题相关的答案过多。ai回答到一半就不继续回答。

FastGPT回复长度计算公式:

最大回复=min(配置的最大回复（内置的限制），最大上下文（输入和输出的总和）-历史记录)

18K模型->输入与输出的和

输出增多->输入减小

所以可以：

检查配置的最大回复（回复上限）
减小输入来增大输出，即减小历史记录，在工作流其实也就是“聊天记录”

配置的最大回复：

私有化部署的时候，后台配模型参数，可以在配置最大上文时候，预留一些空间，比如 128000 的模型，可以只配置 120000, 剩余的空间后续会被安排给输出

受到模型上下文的限制，有时候达不到聊天记录的轮次，连续对话字数过多就会报上下文不够的错误。