v0.15.3

[laipz8200]

✨ What’s New in v0.15.3? ✨

Hey folks, it's time for another update, and we've got some exciting additions and improvements rolling out with version v0.15.3. Let’s dive into the highlights:

🚀 New Model Support

• DeepSeek Models: DeepSeek R1 is now supported across several platforms, including SiliconFlow, Azure AI Foundry, Ollama, Volcengine, OpenRouter, Nvidia Catalog, and more. This opens up new possibilities for leveraging AI reasoning and insights across diverse environments.

• Gemini 2.0 Series: Introducing the latest models in the Gemini 2.0 series, such as Gemini 2.0 Flash 001 and Gemini 2.0 Pro Exp. These models deliver exceptional performance and efficiency, perfect for high-precision, fast-processing tasks.

⚙️ Enhancements

• Firecrawl Upgrade: Updated to the v1 API by @ftonato, providing improved performance and more refined results.

• GLM-4-AIR-0111 Support: Dive into advanced natural language processing tasks more effectively with this addition by @Jhvcc.

• Agent Thinking Content Display: Visualize agent thinking paths with DeepSeek’s R1 model, providing better understanding and debugging capabilities, courtesy of @hjlarry.

🛠️ Bug Fixes

• SSRF Proxy File Descriptor Leak Fix: Eliminated file descriptor leaks to enhance performance in concurrent environments, thanks to @ZuzooVn.

• Page Crash on String Variable Property Access: Resolved issues where improper string handling could cause unintentional crashes, fixed by @iamjoel.

Keep exploring and integrating these new capabilities into your workflows. As always, happy building, and we look forward to hearing how you leverage these updates in your projects! 🔍🚀

---

Upgrade Guide

Docker compose deployments

Warning

The files in the docker-legacy directory will soon stop being maintained and will be removed from the repository. If you are still using them, please switch to the new version as soon as possible.

1. Back up your customized docker-compose YAML file (optional)

   cd docker
   cp docker-compose.yaml docker-compose.yaml.$(date +%s).bak

2. Get the latest code from the main branch

   git checkout main
   git pull origin main

3. Stop the service，Command, please execute in the docker directory

   docker compose down

4. Back up data

   tar -cvf volumes-$(date +%s).tgz volumes

5. Upgrade services

   docker compose up -d

Source Code deployments

1. Stop API server, Worker and Web frontend Server.

2. Get the latest code from the release branch:

   git checkout 0.15.3

3. Update Python dependencies:

   cd api
   poetry install

4. Then, let's run the migration script:

   poetry run flask db upgrade

5. Finally, run API server, Worker and Web frontend Server again.

---

What's Changed

• feat: add pg vector index by @huangzhuo1949 in feat: add pg vector index #12338
• feat: add giteeAI risk control identification. by @jiandanfeng in feat: add giteeAI risk control identification. #12946
• fix: tools translate search by @Jhvcc in fix: tools translate search #12950
• fix: get property of string type variable caused page crash by @iamjoel in fix: get property of string type variable caused page crash #12969
• refactor: Update Firecrawl to use v1 API by @ftonato in refactor: Update Firecrawl to use v1 API #12574
• Feat/prompt editor dark theme by @zxhlyh in Feat/prompt editor dark theme #12976
• feat: add glm-4-air-0111 by @Jhvcc in feat: add glm-4-air-0111 #12997
• fix: wrong zh_Hans translation: Ohio by @iwaim in fix: wrong zh_Hans translation: Ohio #13006
• Change typo in translation by @y10ab1 in Change typo in translation #13004
• chore: add Japanese translation: model_providers/bedrock by @iwaim in chore: add Japanese translation: model_providers/bedrock #13016
• feat: add agent thinking content display of deepseek R1 by @hjlarry in feat: add agent thinking content display of deepseek R1 #12949
• support bedrock kb: retrieve and generate by @warren830 in support bedrock kb: retrieve and generate #13027
• fix: Unexpected tag creation when pressing enter during tag conversion by @miya in fix: Unexpected tag creation when pressing enter during tag conversion #13041
• Update Novita AI predefined models by @jasonhp in Update Novita AI predefined models #13045
• fix: adjust iteration node dark style by @douxc in fix: adjust iteration node dark style #13051
• feat: add inner API to create workspace without requiring email by @zhangx1n in feat: add inner API to create workspace without requiring email #13021
• fix:Improve MIME type detection for remote URL uploads using python-magic by @yjc980121 in fix:Improve MIME type detection for remote URL uploads using python-magic #12693
• fix: "Model schema not found" error only in agents ("Model schema not found" error only in agents #12655) by @jiandanfeng in fix: "Model schema not found" error only in agents (#12655) #12760
• fix qwen-vl blocking mode by @hieheihei in fix qwen-vl blocking mode #13052
• Set default LOG_LEVEL to INFO for celery workers and beat by @osaimi in Set default LOG_LEVEL to INFO for celery workers and beat #13066
• Fixed Novita AI color and added DeepSeek R1 model by @jasonhp in Fixed Novita AI color and added DeepSeek R1 model #13074
• fix: fix the formatter is not applied on log file by @acelyc111 in fix: fix the formatter is not applied on log file #12704
• refactor: reduce duplciate code by inheritance by @acelyc111 in refactor: reduce duplciate code by inheritance #13073
• refactor: Update Firecrawl API parameters and default settings by @ftonato in refactor: Update Firecrawl API parameters and default settings #13082
• add tongyi qwen2.5-14b/7b-instruct-1m model by @hieheihei in add tongyi qwen2.5-14b/7b-instruct-1m model #13089
• refactor: avoid to use extra space when finding model by name by @acelyc111 in refactor: avoid to use extra space when finding model by name #13043
• feat: switch to chat messages before regenerated by @xuzuodong in feat: switch to chat messages before regenerated #11301
• fix: deduct LLM quota after processing invoke result by @laipz8200 in fix: deduct LLM quota after processing invoke result #13075
• test(huggingface_hub): Skip the failed test temporarily. by @laipz8200 in test(huggingface_hub): Skip the failed test temporarily. #13142
• chore: fix wrong VectorType match case by @tmokmss in chore: fix wrong VectorType match case #13123
• feature. add feat to modify metadata via dataset api by @aplio in feature. add feat to modify metadata via dataset api #13116
• pass model and cost so that langfuse can show cost by @obadakhalili in pass model and cost so that langfuse can show cost #13117
• feat: the think tag display of siliconflow's deepseek r1 by @hjlarry in feat: the think tag display of siliconflow's deepseek r1 #13153
• chore: upgrade libldap2 by @ZuzooVn in chore: upgrade libldap2 #13158
• nitpick: fix small typos in template.en.mdx by @aplio in nitpick: fix small typos in template.en.mdx #13156
• add azure openai api version 2024-12-01-preview by @kei-yamazaki in add azure openai api version 2024-12-01-preview #13135
• fix: not install libmagic raise error by @hjlarry in fix: not install libmagic raise error #13146
• fix: set indexing technique from dataset during update-by-text by @aplio in fix: set indexing technique from dataset during update-by-text #13155
• Solve circular dependency issue between workflow/constants.ts file and default.ts file by @obadakhalili in Solve circular dependency issue between workflow/constants.ts file and default.ts file #13165
• chore(pyproject): Add type stubs into pyproject.toml by @laipz8200 in chore(pyproject): Add type stubs into pyproject.toml #13145
• Install apt dependencies using bookworm source, consistent with base image. Remove unnecessary, error-prone pins by @deephbz in Install apt dependencies using bookworm source, consistent with base image. Remove unnecessary, error-prone pins #13176
• fix: Dollar Sign Handling in Markdown by @Riddhimaan-Senapati in fix: Dollar Sign Handling in Markdown #13178
• chore: add tests for build docker image when dockerfile changed by @bowenliang123 in chore: add tests for build docker image when dockerfile changed #10732
• Modify params for bedrock retrieve generate by @warren830 in Modify params for bedrock retrieve generate #13182
• feat: bedrock_endpoint_url by @2284730142 in feat: bedrock_endpoint_url #12838
• Feat: add linkedin to readme by @Riddhimaan-Senapati in Feat: add linkedin to readme #13203
• fix: SSRF proxy file descriptor leak in concurrent requests by @ZuzooVn in fix: SSRF proxy file descriptor leak in concurrent requests #13108
• Use DefaultAzureCredential for managed identity in azure blob extention by @te-chan2 in Use DefaultAzureCredential for managed identity in azure blob extention #11559
• fix: mypy checks violation in AzureBlobStorage by @bowenliang123 in fix: mypy checks violation in AzureBlobStorage #13215
• chore(model_runtime): Move deepseek ahead in the providers list. by @laipz8200 in chore(model_runtime): Move deepseek ahead in the providers list. #13197
• Fix new variables in the conversation opener would override prompt_variables by @HQidea in Fix new variables in the conversation opener would override prompt_variables #13191
• chore: squash system dependencies installation steps by @bowenliang123 in chore: squash system dependencies installation steps #13206
• feat: Add o3-mini and o3-mini-2025-01-31 model variants by @Joshbly in feat: Add o3-mini and o3-mini-2025-01-31 model variants #13129
• feat(model_providers): Add Groq DeepSeek-R1-Distill-Llama-70b by @laipz8200 in feat(model_providers): Add Groq DeepSeek-R1-Distill-Llama-70b #13229
• feat: added deepseek r1 and v3 to siliconflow by @Riddhimaan-Senapati in feat: added deepseek r1 and v3 to siliconflow #13238
• fix: updated _position.yaml to include the latest model already integ… by @Riddhimaan-Senapati in fix: updated _position.yaml to include the latest model already integ… #13245
• feat: add think display for volcengine and generic openapi by @xhebox in feat: add think display for volcengine and generic openapi  #13234
• feat: add parameter "reasoning_effort" and Openai o3-mini by @k-zaku in feat: add parameter "reasoning_effort" and Openai o3-mini  #13243
• feat: added models :gemini 2.0 flash 001 and gemini 2.0 pro exp 02-05 by @Riddhimaan-Senapati in feat: added models :gemini 2.0 flash 001 and gemini 2.0 pro exp 02-05 #13247
• feat(model): add gemini-2.0-flash-thinking-exp-01-21 by @wangiii in feat(model): add gemini-2.0-flash-thinking-exp-01-21 #13230
• fix: correct linewrap think display in generic openai api by @xhebox in fix: correct linewrap think display in generic openai api #13260
• feat(model): add gemini-2.0 model by @wangiii in feat(model): add gemini-2.0 model #13266
• feat(model_providers): Support deepseek for Azure AI Foundry by @laipz8200 in feat(model_providers): Support deepseek for Azure AI Foundry #13267
• Fix bug large data no render by @weiwenyan-dev in Fix bug large data no render #12683
• feat(model_providers): Support deepseek-r1 for Nvidia Catalog by @laipz8200 in feat(model_providers): Support deepseek-r1 for Nvidia Catalog #13269
• Upgrade oracle models by @engchina in Upgrade oracle models #13174
• chore(deps): bump katex from 0.16.10 to 0.16.21 in /web by @dependabot in chore(deps): bump katex from 0.16.10 to 0.16.21 in /web #13270
• fix: add dataset edit permissions by @WTW0313 in fix: add dataset edit permissions #13223
• feat:add deepseek r1 think display for ollama provider by @leslie2046 in feat:add deepseek r1 think display for ollama provider #13272
• feat: add deepseek models for volcengine provider by @xhdd123321 in feat: add deepseek models for volcengine provider #13283
• fix: incorrect transferMethod assignment for remote file by @le0zh in fix: incorrect transferMethod assignment for remote file #13286
• feat:add think tag display for xinference deepseek r1 by @leslie2046 in feat:add think tag display for xinference deepseek r1 #13291
• feat: added a couple of models not defined in vertex ai, that were already … by @Riddhimaan-Senapati in feat: added a couple of models not defined in vertex ai, that were already … #13296
• Update README_TR.md by @kemalmeler in Update README_TR.md #13294
• feat(credits): Allow to configure model-credit mapping by @laipz8200 in feat(credits): Allow to configure model-credit mapping #13274
• feat(model): add deepseek-r1 for openrouter by @sinomoe in feat(model): add deepseek-r1 for openrouter #13312
• fix(credits): require model name equals by @laipz8200 in fix(credits): require model name equals #13314
• chore: refactor the OpenAICompatible and improve thinking display by @hjlarry in chore: refactor the OpenAICompatible and improve thinking display #13299
• Fix the problem of Workflow terminates after parallel tasks execution, merge node not triggered by @lazyFrogLOL in Fix the problem of Workflow terminates after parallel tasks execution, merge node not triggered #12498
• chore(*): Update style of thinking by @laipz8200 in chore(*): Update style of thinking #13319
• chore:improve thinking display for llm from xinference and ollama pro… by @leslie2046 in chore:improve thinking display for llm from xinference and ollama pro… #13318
• fix: Azure AI Foundry model cannot be used in the workflow by @laipz8200 in fix: Azure AI Foundry model cannot be used in the workflow #13323
• feat: add completion mode and context size options for LLM configuration by @laipz8200 in feat: add completion mode and context size options for LLM configuration #13325
• chore(*): Bump version to 0.15.3 by @laipz8200 in chore(*): Bump version to 0.15.3 #13308

New Contributors

• @Jhvcc made their first contribution in fix: tools translate search #12950
• @ftonato made their first contribution in refactor: Update Firecrawl to use v1 API #12574
• @y10ab1 made their first contribution in Change typo in translation #13004
• @zhangx1n made their first contribution in feat: add inner API to create workspace without requiring email #13021
• @yjc980121 made their first contribution in fix:Improve MIME type detection for remote URL uploads using python-magic #12693
• @hieheihei made their first contribution in fix qwen-vl blocking mode #13052
• @osaimi made their first contribution in Set default LOG_LEVEL to INFO for celery workers and beat #13066
• @aplio made their first contribution in feature. add feat to modify metadata via dataset api #13116
• @obadakhalili made their first contribution in pass model and cost so that langfuse can show cost #13117
• @deephbz made their first contribution in Install apt dependencies using bookworm source, consistent with base image. Remove unnecessary, error-prone pins #13176
• @Riddhimaan-Senapati made their first contribution in fix: Dollar Sign Handling in Markdown #13178
• @2284730142 made their first contribution in feat: bedrock_endpoint_url #12838
• @te-chan2 made their first contribution in Use DefaultAzureCredential for managed identity in azure blob extention #11559
• @HQidea made their first contribution in Fix new variables in the conversation opener would override prompt_variables #13191
• @Joshbly made their first contribution in feat: Add o3-mini and o3-mini-2025-01-31 model variants #13129
• @xhebox made their first contribution in feat: add think display for volcengine and generic openapi  #13234
• @weiwenyan-dev made their first contribution in Fix bug large data no render #12683
• @engchina made their first contribution in Upgrade oracle models #13174
• @kemalmeler made their first contribution in Update README_TR.md #13294
• @lazyFrogLOL made their first contribution in Fix the problem of Workflow terminates after parallel tasks execution, merge node not triggered #12498

Full Changelog: 0.15.2...0.15.3

---

This discussion was created from the release v0.15.3.

[william0312ai]

ollama deepseek R1 ,Thinking Tag don't work.

[leslie2046]

@william0312ai can you test like this and show me your log
curl http://localhost:11434/v1/chat/completions
-H "Content-Type: application/json"
-d '{
"model": "deepseek-r1:32b",
"stream": true,
"messages": [
{
"role": "system",
"content": "You are a helpful assistant."
},
{
"role": "user",
"content": "Hello!"
}
]
}'

[leslie2046]

btw,you can check if select false(false is right)

[william0312ai]

it's work.thanks

[leslie2046]

you are welcome

[daisukeyokoi]

When using the Tool in Dify, I can authenticate with Firecrawl, but I cannot authenticate when using it from Knowledge. This is in my local environment.

[daisukeyokoi]

[cusclo]

这种情况是main分支镜像1.0不存在？我是不是应该手动修改一下

[laipz8200]

这种情况是main分支镜像1.0不存在？我是不是应该手动修改一下

Yes, you can run git checkout 0.15.3 to resolve this problem.

[TrueAlpha-spiral]

Howdy partners

[lachinhan]

How to fix this ???

[laipz8200]

Google has changed their API, and since we have no plan to maintain the old version, you can upgrade to version 1.0.0 to resolve this issue.

[lachinhan]

Dear @laipz8200 1.0.0 is even worse, I can't add any AI API Key from Open AI, Gemini, DeepSeek.... all of them can't be added. Dify Cloud is fine but running on local host shows this error

[laipz8200]

I'm sorry to hear that. But we have a helpful guide on our documentation site that might just do the trick for you!

https://docs.dify.ai/development/migration/migrate-to-v1

[lachinhan]

Thanks @laipz8200 . But I tried before, and it would validate everything, but it wouldn't run the model or run the model and give an error. So I tried deleting everything and reinstalling, but it wouldn't add the AI ​​Api Keys.