langgenius/dify 0.9.0 on GitHub

✨ What’s New in v0.9.0? ✨

Get ready for a feature-packed update! Version 0.9.0 brings major enhancements, new integrations, and crucial fixes. Let’s dive into the details:

🚀 New Features

🔄 Retrieval-Augmented Generation (RAG)

External Knowledge API: Now you can connect to third-party knowledge bases seamlessly (introduced by @JohnJyong in #8913).

Simply provide the API endpoint, API key, and external knowledge ID to link with the external knowledge base.
Supports retrieval and recall of external knowledge based on user queries.
Enables integration of external knowledge bases within Dify applications.

Website Crawling: Now supports Jina Reader as additional alternative for website crawling, converting web page URL to LLM-friendly input format. Read Sync Data from Website for details. (@mapleeit in #8761).

🧠 Models

Voyage AI: A new model provider added (thanks to @hwzhuhao in #8747).
Llama3.2: Now available with Fireworks and Groq providers for improved performance and variety (@yaoice in #8809 and @Howe829 in #8831).
Introduced InternLM 2.5-20b and Qwen 2.5-coder-7b (@hwzhuhao in #8862).
Jamba: Another new model to expand our capabilities (@longzhihun in #8878).
New Nomic embedding model provider introduced (@yaoice in #8640).
Support for o1 series models for OpenRouter (@sinomoe in #8358).
DeepSeek-v2.5 added as a model provider for SiliconFlow (@hwzhuhao in #8639).
Stream for Gemini now available (@AAEE86 in #8678).
Vertex AI Gemini 1.5 002 and experimental models are now supported (@hellof20 in #8767).

🛠️ Other Updates

Xinference TTS Stream Mode: Now live for a smoother text-to-speech experience (@leslie2046 in #8616).
Regenerate in Chat, Agent, and Chatflow applications (@xuzuodong in #7661).

⚙️ Enhancements

Improved SVG output UX for a more seamless design experience (@xuzuodong in #8765).
Extended Jina Embeddings V3 API parameters (@DresAaron in #8657).
Deprecated older Mistral models for SiliconFlow to streamline performance (@hwzhuhao in #8828).
Added Min-connections and Max-connections settings for pgvector (@hwzhuhao in #8841).
Error Handling Enhancements: Improved iteration filters and error logging for optimal performance (@hwzhuhao in #8461).
Enhanced operational flexibility with fully configurable WORKFLOW_ environment variables* (@fujita-h in #8644).
Upgraded ElasticSearch from 8.14.x to 8.15.x to keep our infrastructure up-to-date (@bowenliang123 in #8197).

🛠️ Bug Fixes

Fixed "Max submit count reached" error during workflow iterations (@takatost in #8595).
Corrected a linking error in help documents (@hjlarry in #8508).
iOS TTS Stream: Now fully functional on Safari 17.1+ (@likenamehaojie in #8645).
Resolved JSONDecodeError in llm_generator.py (@ZuzooVn in #8504).
Fixed errors when sending messages within chatflow and dialogs (@xuzuodong in #8627 and #8682).
Corrected Spark's LLM token calculation error for accurate token usage (@cherryhuahua in #8755).
Resolved embedded chat issues on iOS (@zxhlyh in #8718).

📚 Documentation & Misc

Enhanced documentation to clarify the importance of Celery Service (@fchange in #8558).
Updated configurations in docker-compose files and related docs (@saifeiLee and others in #8516 and #8691).
Added English versions for customizable model scale documentation to broaden accessibility (@hwzhuhao in #8871).

With this release, we’re truly showcasing the power of Dify! We appreciate your continued support—explore the new features and share your feedback with us!

Upgrade Guide

Docker compose deployments

Warning

The docker-compose.yaml has been refactored. If you've made any changes to the file, make sure to check out the "Upgrade to new docker compose deployment" section above for usage and migration tips.

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

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

Get the latest code from the main branch
```
git checkout main
git pull origin main
```
Stop the service，Command, please execute in the docker directory
```
docker compose down
```

Back up data

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

Upgrade services
```
docker compose up -d
```

Source Code deployments

Stop API server, Worker and Web frontend Server.
Get the latest code from the main branch:
```
git checkout 0.9.0
git pull origin 0.9.0
```
Update Python dependencies:
```
cd api
poetry install
```
Then, let's run the migration script:
```
poetry shell
flask db upgrade
```
Finally, run API server, Worker and Web frontend Server again.

What's Changed

fix(workflow): "Max submit count reached" error occurred when executing workflow as tool in iteration by @takatost in #8595
fix: a helper link error by @hjlarry in #8508
chore: translate i18n files by @github-actions in #8557
Enhance Readme Documentation to Clarify the Importance of Celery Service by @fchange in #8558
docs: Update ssrf_proxy related doc link in docker-compose file by @saifeiLee in #8516
Update the PerfXCloud provider model list，Update PerfXCloudProvider validate_provider_credentials method. by @BingGeX in #8587
add MixtralAI Model by @AAEE86 in #8517
fix: api/core/ops/ops_trace_manager.py by @ZuzooVn in #8501
add Qwen custom add model interface by @AAEE86 in #8565
refactor(ops): Optimize the iteration for filter_none_values and use logging.error to record logs when an exception occurs by @hwzhuhao in #8461
feat: update pyproject.toml by @ZhouhaoJiang in #8368
feat: regenerate in Chat, agent and Chatflow app by @xuzuodong in #7661
docs: Add Japanese documentation for tools by @totsukash in #8469
feat:use xinference tts stream mode by @leslie2046 in #8616
Add Fireworks AI as new model provider by @yaoice in #8428
feat(tools/cogview): Updated cogview tool to support cogview-3 and the latest cogview-3-plus by @ox01024 in #8382
Add model parameter translation by @AAEE86 in #8509
fix: form input add tabIndex by @LLmoskk in #8478
feat: support o1 series models for openrouter by @sinomoe in #8358
chore: add Gemini newest experimental models (close #7121) by @CXwudi in #8621
chore: enhance configuration descriptions by @ZuzooVn in #8624
fix: redundant check for available_document_count by @y-omr in #8491
fix: commands.py by @ZuzooVn in #8483
fix: llm_generator.py JSONDecodeError by @ZuzooVn in #8504
fix: send message error when chatting with opening statement by @xuzuodong in #8627
feat: add qwen2.5 for model provider siliconflow by @hwzhuhao in #8630
docs: fix predefined_model_scale_out.md redirect error by @hwzhuhao in #8633
feat: add deepseek-v2.5 for model provider siliconflow by @hwzhuhao in #8639
Make WORKFLOW_* configurable as environment variables. by @fujita-h in #8644
chore: Enable Japanese descriptions for Tools by @totsukash in #8646
feat: extend api params for Jina Embeddings V3 by @DresAaron in #8657
fix: Assignment exception by @haike-1213 in #8663
chore: skip PLR6201 linter rule by @bowenliang123 in #8666
fix: png avatar upload as jpeg by @hjlarry in #8665
add Qwen model translate by @AAEE86 in #8674
add zhipuai web search by @AAEE86 in #8668
add Spark Max-32K by @AAEE86 in #8676
chore: Optimize I18nObject class for better performance and readability by @ZuzooVn in #8681
fix: send message error when last sent message not succeeded by @xuzuodong in #8682
fixbug tts(stream) not work on ios safari(17.1+) by @likenamehaojie in #8645
fix: document_create_args_validate by @vikeychen in #8569
Add nomic embedding model provider by @yaoice in #8640
fix: prompt for the follow-up suggestions by @crazywoola in #8685
fix: typos and improve naming conventions: by @ZuzooVn in #8687
fix: change the brand name to Jina AI by @Nick17t in #8691
feat: add xinference sd web ui api tool by @themanforfree in #8385
fix llm integration problem: It doesn't work on docker env by @ybalbert001 in #8701
feat: add mixedbread as a new model provider by @hwzhuhao in #8523
fix: Remove unsupported parameters in qwen model by @AAEE86 in #8699
fix: change model provider name issue Ref #8691 by @BenjaminX in #8710
fix: update nomic model provider token calculation by @yaoice in #8705
fix: embedded chat on ios by @zxhlyh in #8718
chore: remove windows platform timezone set by @hjlarry in #8712
fix: remove harm category setting from vertex ai by @totsukash in #8721
Fix/update jina ai products labels and descriptions by @Nick17t in #8730
add embedding input type parameter by @JohnJyong in #8724
Add embedding models in fireworks provider by @yaoice in #8728
Fix Issue: switch LLM of SageMaker endpoint doesn't take effect by @ybalbert001 in #8737
fix: . is missing in file_extension by @crazywoola in #8736
chore: make url display in the middle of http node by @hjlarry in #8741
chore: apply ruff reformat for python-client sdk by @hwzhuhao in #8752
fix:Spark's large language model token calculation error #7911 by @cherryhuahua in #8755
fix: expose the configuration of HTTP request node to Docker by @zhuiyue132 in #8716
Add GitHub Actions Workflow for Web Tests by @xuzuodong in #8753
fix: the translation result may cause a different meaning by @douxc in #8763
feat: add support for Vertex AI Gemini 1.5 002 and experimental models by @hellof20 in #8767
dep: bump ElasticSearch from 8.14.x to 8.15.x by @bowenliang123 in #8197
chore: fix comfyui tool doc url by @QunBB in #8775
fix: TavilySearch tool get api link by @hjlarry in #8780
chore: add input_type param desc for the _invoke method of text_embedding by @hwzhuhao in #8778
fix: credential *** should be string by @hjlarry in #8785
fix: update usage for Jina Embeddings v3 by @DresAaron in #8771
Add more models for SiliconFlow by @jackalcooper in #8779
fix: the Image-1X tool by @hjlarry in #8787
fix: start node input config modal raise 'variable name is required' by @hjlarry in #8793
fix: Ollama modelfeature set vision, and an exception occurred at the… by @free-cx in #8783
refactor: update delete method as an abstract method by @hwzhuhao in #8794
fix: workflow as tool always outdated by @hjlarry in #8798
fix: change wrong company name by @iamjoel in #8801
feat: add qwen to add custom model parameters by @AAEE86 in #8759
feat: add stream for Gemini by @AAEE86 in #8678
feat: improved SVG output UX by @xuzuodong in #8765
Add llama3.2 model in fireworks provider by @yaoice in #8809
add more detailed doc for models of qwen series by @wenmengzhou in #8799
chore: massive update of the Gemini models based on latest documentation by @CXwudi in #8822
Updated the YouTube channel to Dify's by @shaiss in #8817
feat: add pgvecto-rs and analyticdb in docker/.env.example by @hwzhuhao in #8823
fix: close log status option raise error by @hjlarry in #8826
fix: AnalyticdbVector retrieval scores by @lpdink in #8803
feat: deprecate mistral model for siliconflow by @hwzhuhao in #8828
Add Llama3.2 models in Groq provider by @Howe829 in #8831
fix: delete harm catalog settings for gemini by @CXwudi in #8829
feat(Tools): add feishu tools by @hgnulb in #8800
feat: add min-connection and max-connection for pgvector by @hwzhuhao in #8841
chore: fix wrong VectorType match case by @yaoice in #8857
fix(workflow/nodes/knowledge-retrieval/use-config): Preserve rerankin… by @Kevin9703 in #8842
fix: update qwen2.5-coder-7b model name by @hwzhuhao in #8861
fix: customize model credentials were invalid despite the provider credentials being active by @takatost in #8864
feat: add internlm2.5-20b and qwen2.5-coder-7b model by @hwzhuhao in #8862
feat: deprecate gte-Qwen2-7B-instruct embedding model by @hwzhuhao in #8866
refactor: update Callback to an abstract class by @hwzhuhao in #8868
chore: bump ruff to 0.6.8 for fixing violation in SIM910 by @bowenliang123 in #8869
fix(workflow): update tagging logic in GitHub Actions by @laipz8200 in #8882
Add Jamba and Llama3.2 model support by @longzhihun in #8878
fix: #8843 event: tts_message_end always return in api streaming resp… by @ic-xu in #8846
docs: add english versions for the files customizable_model_scale_out and predefined_model_scale_out by @hwzhuhao in #8871
feat: add voyage ai as a new model provider by @hwzhuhao in #8747
feat(website-crawl): add jina reader as additional alternative for website crawling by @mapleeit in #8761
feat(api): add version comparison logic by @laipz8200 in #8902
refactor(api/variables): VariableError as a ValueError. by @laipz8200 in #8554
fix: fix the data validation consistency issue in keyword content review by @hwzhuhao in #8908
refactor: optimize the calculation of rerank threshold and the logic for forbidden characters in model_uid by @hwzhuhao in #8879
external knowledge api by @JohnJyong in #8913
fix: line in iteration node is not straight by @YIXIAO0 in #8918
fix: Compatible with special characters in pg full-text search. by @ThreeFish-AI in #8921
chore: translate i18n files by @github-actions in #8917
fix: fix the issue with the system model configuration update by @hwzhuhao in #8923
Feat: add debounce for search in logs by @JzoNgKVO in #8924
chore(versioning): bump version to 0.9.0 by @laipz8200 in #8911

New Contributors

@fchange made their first contribution in #8558
@CXwudi made their first contribution in #8621
@fujita-h made their first contribution in #8644
@haike-1213 made their first contribution in #8663
@likenamehaojie made their first contribution in #8645
@cherryhuahua made their first contribution in #8755
@zhuiyue132 made their first contribution in #8716
@jackalcooper made their first contribution in #8779
@free-cx made their first contribution in #8783
@wenmengzhou made their first contribution in #8799
@shaiss made their first contribution in #8817
@mapleeit made their first contribution in #8761

Full Changelog: 0.8.3...0.9.0

langgenius/dify 0.9.0 v0.9.0 on GitHub