v0.3.24

What's Changed

• add help wanted issue template by @crazywoola in #1199
• fix: provider spark free quota text by @zxhlyh in #1201
• Complete type defined. by @JunIce in #1200
• fix: transaction hangs due to message commit block during long LLM calls by @takatost in #1206
• fix: transaction not commit during long LLM calls by @takatost in #1213
• Feat/huggingface embedding support by @GarfieldDai in #1211
• Fix/qdrant data issue by @JohnJyong in #1203
• fix: dataset eslint error by @iamjoel in #1221
• Fix: add loading for dataset creation by @JzoNgKVO in #1214
• feat: batch run support retry errors and decrease rate limit times by @iamjoel in #1215
• Feat/api jwt by @zxhlyh in #1212
• fix: FATAL: role "root" does not exist. by @crazywoola in #1233
• Fix App logs page modal show different model icon. by @JunIce in #1224
• Fix: debounce of dataset creation by @JzoNgKVO in #1237
• Fix: dataset list refresh by @JzoNgKVO in #1216
• feat: support binding context var by @GarfieldDai in #1227
• fix: dataset query error. by @GarfieldDai in #1244
• Feat/dataset service api by @JohnJyong in #1245
• Feat/dataset support api service by @zxhlyh in #1240
• fix: wenxin model name invalid when llm call by @takatost in #1248
• fix: db session not commit before long llm call running by @takatost in #1251
• fix: dataset footer styles by @crazywoola in #1254
• Fix/dataset api key delete by @JohnJyong in #1255
• fix: doc by @zxhlyh in #1256
• Fix/notion sync by @zxhlyh in #1258
• feat: bump version to 0.3.24 by @takatost in #1262

Full Changelog: 0.3.23...0.3.24

Version 0.3.23

What's Changed

• feat: spark free quota verify by @zxhlyh in #1152
• feat: spark check by @takatost in #1134
• feat: add hosted moderation by @takatost in #1158
• chore: hover conversation show option button by @iamjoel in #1160
• Update README.md by @crazywoola in #1164
• add video to cn readme by @crazywoola in #1165
• fix LocalAI is missing in lang/en by @crazywoola in #1169
• Feat/add blocking mode resource return by @JohnJyong in #1171
• fix: correct invite url by @MatriQ in #1173
• fix: Improved fallback solution for avatar image loading failure by @CoolBoyLeo in #1172
• Mermaid front end rendering by @charli117 in #1166
• Restore the application template by @charli117 in #1174
• chore: Optimization model parameter description by @iamjoel in #1181
• fix: model param description error by @iamjoel in #1183
• fix: text error by @iamjoel in #1184
• feat(web): service request return generics type by @JunIce in #1157
• 1111 wrong embedding model displayed in datasets by @crazywoola in #1186
• fix: inference embedding validate by @takatost in #1187
• feat: add zhipuai by @takatost in #1188
• Feat/provider add zhipuai by @zxhlyh in #1192
• Feat/improve vector database logic by @JohnJyong in #1193
• feat: support gpt-3.5-turbo-instruct model by @takatost in #1195
• fix: gpt-3.5-turbo-instruct context size to 8192 by @takatost in #1196
• build: update devDependencies by @jiaowoxiaobala in #1125
• feat: bump version to 0.3.23 by @takatost in #1198

New Contributors

• @CoolBoyLeo made their first contribution in #1172
• @charli117 made their first contribution in #1166

Full Changelog: 0.3.22...0.3.23

Version 0.3.22

Notice: if you're upgrading from before version 0.3.21, make sure you've read the BREAKING CHANGE in version 0.3.21: https://github.com/langgenius/dify/releases/tag/0.3.21

What's Changed

• Feat:csv & docx support by @JzoNgKVO in #1139
• Feat:dataset retiever resource by @JohnJyong in #1123
• fix(web): complete some ts type by @JunIce in #1148
• Update issue template by @crazywoola in #1150
• Update bug_report.yml by @crazywoola in #1151
• feat: bump version to 0.3.22 by @takatost in #1153

Full Changelog: 0.3.21...0.3.22

Version 0.3.21

⚠️ [BREAKING CHANGE] ⚠️

• The fixed parameters of the Query content in the text generation application have been removed. Users can freely add any variables in the prompt, instead of appending the content of the query content at the end.
• The query parameter in the POST /v1/completion-messages endpoint of the App OpenAPI is expected to be completely removed in 3 months. Currently, it's optional and the parameter in API documentation has already been removed.

To ensure that the previous applications can still function properly after removing the fixed parameters in the query content of the text generation application, we need to execute a command to fix the data:

Migration Guide

1. Enter the Docker Container: Before running the migration script, you need to enter the Docker container where your application is running. Use the following command:

   docker exec -it docker-api-1 bash

2. Run the Migration Script: Once inside the container, execute the following migration command. The migration script can be executed multiple times without side effects:

   # --batch-size: Specifies the number of records migrated per batch, default is 500
   flask update_app_model_configs --batch-size=500

What's Changed

• feat: support optional query content by @iamjoel in #1097
• fix: in log page not show user query by @iamjoel in #1140
• fix: log page compatible old query by @iamjoel in #1141
• fix: remove the deprecated depends_on.condition format by @takatost in #1142
• feat: cache huggingface gpt2 tokenizer files by @takatost in #1138
• feat: set transformers offline default true by @takatost in #1143
• fix: TRANSFORMERS_OFFLINE orders in Dockerfile by @takatost in #1144
• feat: bump version to 0.3.21 by @takatost in #1145

Full Changelog: 0.3.20...0.3.21

Version 0.3.20

What's Changed

• feat: optimize generation of conversation title by @takatost in #1075
• feat: add visual studio code debug config. by @GarfieldDai in #1068
• feat: optimize conversation title generation output must be a valid JSON by @takatost in #1077
• feat: optimize high load sql query of document segment by @takatost in #1078
• chore(api): api image multistage build by @JunIce in #1069
• feat: optimize prompt by @takatost in #1080
• feat: optimize last_active_at update by @takatost in #1083
• fix: prompt of title generation by @takatost in #1084
• feat: support spark v2 validate by @takatost in #1086
• fix: ui in chat openning statement by @crazywoola in #1094
• fix: change frontend start script by @iamjoel in #1096
• fix: markdown code lang capitalization and line number color by @iamjoel in #1098
• feat: Optimize the description for Azure deployment name by @seewhy163 in #1091
• fix: optimize feedback and app icon by @jiaowoxiaobala in #1099
• chore: handle workspace apps scrollbar by @iamjoel in #1101
• fix: try to fix chunk load error by @iamjoel in #1109
• chore: enchancment frontend readme by @iamjoel in #1110
• fix(web): fix AppCard Menu popover open bug by @JunIce in #1107
• fix: optimize styles by @jiaowoxiaobala in #1112
• fix: self host embedding missing base url config by @iamjoel in #1116
• fix: shortening invite url by @MatriQ in #1100
• fix: click inner link no jump by @iamjoel in #1118
• fix: hf hosted inference check by @takatost in #1128
• feat: add health check and depend condition in docker compose by @deltaLRD in #1113
• feat: optimize celery start cmd by @takatost in #1129
• Feat/chat add origin by @zxhlyh in #1130
• fix: answer returned null when response_mode was blocking by @takatost in #1133
• feat: bump version to 0.3.20 by @takatost in #1135

New Contributors

• @GarfieldDai made their first contribution in #1068
• @seewhy163 made their first contribution in #1091
• @deltaLRD made their first contribution in #1113

Full Changelog: 0.3.19...0.3.20

Version 0.3.19

What's Changed

• document limit by @JohnJyong in #999
• Feat/vector db manage by @JohnJyong in #997
• fix: styles by @crazywoola in #1005
• refactor:cache in place & function name by @yezhwi in #1001
• fix: loading and calc rem by @crazywoola in #1006
• embedding model available check by @JohnJyong in #1009
• Fix/filter empty segment by @JohnJyong in #1004
• update dataset index struct by @JohnJyong in #1012
• fix: xinference last token being ignored by @UranusSeven in #1013
• fix update document index technique by @JohnJyong in #1018
• feat: maintenance notice by @zxhlyh in #1016
• fix document estimate parameter by @JohnJyong in #1019
• feat: update contributing by @takatost in #1020
• fix: maintenance notice by @zxhlyh in #1025
• fix: dataset query truncated by @takatost in #1026
• feat: hf inference endpoint stream support by @takatost in #1028
• Fix: eslint by @JzoNgKVO in #1030
• Fix/upload document limit by @zxhlyh in #1033
• Fix/model selector by @zxhlyh in #1032
• fix: pub generate message text return null by @takatost in #1037
• fix notion import limit check by @JohnJyong in #1042
• chore: perfect type definition by @jiaowoxiaobala in #1003
• Fix/ignore economy dataset by @JohnJyong in #1043
• fix: dataset desc by @takatost in #1045
• chore: match new dataset tool format by @iamjoel in #1044
• fix openpyxl dimensions error by @wangbooth in #1041
• add qdrant migration by @JohnJyong in #1046
• recreate qdrant vector by @JohnJyong in #1049
• feat: add LocalAI local embedding model support by @takatost in #1021
• 773 change embed title welcome to use by @crazywoola in #1053
• fix: handle AttributeError for datasets and index by @yezhwi in #1052
• Fix: disable operations of dataset when embedding unavailable by @JzoNgKVO in #1055
• feat: support rename conversation by @iamjoel in #1056
• feat: code support copy by @iamjoel in #1057
• fix error weaviate vector by @JohnJyong in #1058
• Feat/support to invite multiple users by @MatriQ in #1011
• refactor(file extractor): file extractor by @yezhwi in #1059
• chore: doc issue by @crazywoola in #1062
• fix: csv import error by @takatost in #1063
• feat: optimize anthropic connection pool by @takatost in #1066
• feat: disable anthropic retry by @takatost in #1067
• feat: siderbar operation support portal by @iamjoel in #1061
• feat: decrease claude model request time by set max top_k to 10 by @takatost in #1071
• fix: segments query missing idx hit by @takatost in #1072
• fix: summary no docs by @takatost in #1073
• feat: bump version to 0.3.19 by @takatost in #1074

New Contributors

• @UranusSeven made their first contribution in #1013
• @wangbooth made their first contribution in #1041

Full Changelog: 0.3.18...0.3.19

Version 0.3.18

What's Changed

• feat: remove unnecessary prompt of baichuan by @takatost in #993
• fix: safetensor arm complie error by @takatost in #996
• fix style in app share by @crazywoola in #995
• fix: apps loading issue by @MatriQ in #994
• feat: optimize xinference request max token key and stop reason by @takatost in #998
• feat: bump version to 0.3.18 by @takatost in #1000

Full Changelog: 0.3.17...0.3.18

Version 0.3.17

What's Changed

• feat: adjust hf max tokens by @takatost in #979
• feat: add baichuan prompt by @takatost in #985
• chore: update lintrc config by @MatriQ in #986
• feat: optimize baichuan prompt by @takatost in #988
• feat: optimize xinference stream by @takatost in #989
• fix: xinference-chat-stream-response by @takatost in #991
• feat: bump version to 0.3.17 by @takatost in #992

Full Changelog: 0.3.16...0.3.17

Version 0.3.16

What's Changed

• normalize embedding by @JohnJyong in #974
• feat: optimize hf inference endpoint by @takatost in #975
• fix: add hf task field by @takatost in #976
• Fix:condition for dataset availability check by @yezhwi in #973
• feat: upgrade xinference to 0.2.1 which support stream response by @takatost in #977

New Contributors

• @yezhwi made their first contribution in #973

Full Changelog: 0.3.15...0.3.16

Version 0.3.15

New features

• feat: add spark v2 support by @takatost in #885
• feat: server xinference support by @takatost in #927
• feat: add openllm support by @takatost in #928

What's Changed

• feat: support app rename and make app card ui better by @iamjoel in #766
• fix: promptEng menu in wrong place by @iamjoel in #864
• fix: get_num_tokens_from_messages params error by @takatost in #866
• fix multi thread app context by @JohnJyong in #868
• feat: optimize error record in agent by @takatost in #869
• Update README_CN.md by @Panmuse in #867
• Update README.md by @Panmuse in #865
• feat: optimize override app model config convert by @takatost in #874
• feat: optimize model when app create by @takatost in #875
• Feat/customizable file upload config by @MatriQ in #818
• build: improve dockerfile by @MatriQ in #851
• feat: optimize prompt of suggested_questions_after_answer by @takatost in #881
• fix: web dockfile by @zxhlyh in #883
• fix: wenxin error not raise when stream mode by @takatost in #884
• feat: add spark v2 support by @takatost in #885
• feat: support show model display name by @iamjoel in #887
• fix: error raise status code not exist by @takatost in #888
• Feat/pay modal by @zxhlyh in #889
• fix: max token not exist in generate summary when calc rest tokens by @takatost in #891
• feat: claude paid optimize by @takatost in #890
• Fix/embedding chat by @zxhlyh in #899
• fix entrypoint script line endings by @JunIce in #900
• fix: sensitive_word_avoidance npe by @takatost in #902
• fix: react router agent direct output by @takatost in #904
• fix: universal chat when default model invalid by @takatost in #905
• feat: optimize output parse by @takatost in #906
• Feat/segment add tag by @zxhlyh in #907
• Feature/mutil embedding model by @JohnJyong in #908
• feat: optimize service api authorization header invalid error by @takatost in #910
• feat: remove unuse envs by @takatost in #912
• fix: run model serval ex not return by @takatost in #915
• fix: run extra model serval ex not return by @takatost in #916
• fix: azure embedding model credentials include base_model_name is invalid for openai sdk by @takatost in #917
• Fix/price calc by @lingfengchencn in #862
• fix: embedding price config by @takatost in #918
• feat: record price unit in messages by @takatost in #919
• fix: embedding get price definition missing by @takatost in #922
• fix: replicate text generation model validate by @takatost in #923
• feat: optimize generate name length by @takatost in #924
• fix: document paused at format error by @takatost in #925
• Fix Readme.md typo error. by @BenjaminX in #926
• feat: server xinference support by @takatost in #927
• feat: add openllm support by @takatost in #928
• fix: xinference max_tokens alisa error by @takatost in #929
• feat: add xinference embedding model support by @takatost in #930
• fix: remove openllm pypi package because of this package too large by @takatost in #931
• fix: web reader tool missing nodejs by @takatost in #932
• feat: optimize web reader summary in 3.5 by @takatost in #933
• Feat/token support by @JohnJyong in #909
• Feat/add document status tooltip by @zxhlyh in #937
• fix: xinference chat support by @takatost in #939
• fix: login improvement by @MatriQ in #941
• fix: default language improvement in activation page by @MatriQ in #942
• chore: perfect type definition by @jiaowoxiaobala in #940
• fix: add flex-wrap to categories container by @MatriQ in #944
• fix: openllm generate cutoff by @takatost in #945
• fix: provider model not delete when reset key pair by @takatost in #946
• Fix/speech to text button by @zxhlyh in #947
• update document segment by @JohnJyong in #948
• fix: segment error tip & save segment disable when loading by @zxhlyh in #949
• Feat/model providers by @zxhlyh in #951
• fix: xinference chat completion error by @takatost in #952
• chore: create btn to first place by @iamjoel in #953
• chore: obsolete info api use new api by @iamjoel in #954
• fix: spark provider field name by @zxhlyh in #955
• fix: embedding init err by @takatost in #956
• update document segment by @JohnJyong in #950
• fix: web reader tool retrieve content empty by @takatost in #957
• Feat/add xinference openllm provider by @zxhlyh in #958
• feat: bump version to 0.3.15 by @takatost in #959
• fix: user activation request 404 issue by @MatriQ in #963
• update doc by @crazywoola in #965
• fix: activation page reload issue after activating by @MatriQ in #964

Full Changelog: 0.3.14...0.3.15