langgenius/dify 0.12.0 on GitHub

✨ What’s New in v0.12.0? ✨

Important

Heads up! Starting with this version, we'll no longer support Python 3.10. Make sure to upgrade to Python 3.11 or 3.12 to stay current with our updates.

Hey everyone! We’ve been busy fine-tuning the platform and packing in some new features in version 0.12.0. Let’s see what’s new and improved:

🚀 Enhanced File Support for Workflow/Chatflow LLM Node and Chat/Agent

We’re excited to introduce the new file support capabilities for LLM Node and Basic Chat/Agent! This update significantly enhances multi-modal capabilities and smart document processing, providing more flexibility and power for your application development. Here are the key highlights:

File Variable Support in LLM Node

LLM Node now supports directly referencing File and Array File variables in Prompts, unlocking greater possibilities for developers.
For example, you can upload multiple file variables (such as PDFs and images) and have the model collaboratively handle complex tasks, like summarizing report content or extracting key data. Multi-file support makes workflow design more efficient and seamless!

Tip

File Support Scope: The file types supported by the model may be more limited than the types allowed for upload. Be sure to read your model provider's official documentation to ensure that uploaded files can be processed correctly.
Document Handling in Unsupported Models: If the model does not natively support processing document files, you will still need to use the Document Extractor for extracting content.
Filtering Unsupported Files: To handle unsupported file types, consider using List Operator or IF-ELSE to screen them, ensuring smooth workflows.
Relation to Vision Features: File support operates independently of the Vision toggle. As long as the model supports Vision capabilities, you can use files as input.

New Document-Type Support for Basic Chat and Agent

Basic Chat and Agent now include a Document toggle. If the model supports processing document-type files, you can enable this feature.
For instance, you can upload a PDF report and let the Agent automatically analyze market data or assess content value, making such workflows simpler and smarter.

Enhanced Multi-Modal Capabilities

The new version extends support for multi-modal tasks. For example, you can upload audio files and have the model perform emotional analysis. This combination of text, audio, and file processing capabilities allows you to build more diverse and rich application scenarios.

⚙️ Improvements

OpenAI o1 Streaming Support: We now support streaming output for OpenAI o1-preview and o1-mini, thanks to @laipz8200 in #10890.
Performance Optimization: The WeightRerankRunner logic has been optimized to achieve constant time complexity, thanks to @yihong0618 in #10849.
UI Enhancements: We’ve made the Table of Contents panel in the API document page collapsible, giving you greater control over your workspace, thanks to @hjlarry in #10875.
JSON Schema for Gemini Models: Added support for JSON schema integration, ensuring smoother alignment with Gemini models, thanks to @hjlarry in #10835.
Redis Clusters: Our ext_redis.py now includes support for Redis clusters, expanding flexibility for data management, implemented by @liuhaoran1212 in #9789.
Enhanced Custom Notes: You can now further enhance custom notes for increased expressiveness, courtesy of @hjlarry in #8885.

That’s the scoop for v0.12.0! Until next time, keep exploring and innovating! 🌟

Upgrade Guide

Docker compose deployments

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 release branch:
```
git checkout 0.12.0
```
Update Python dependencies:
```
cd api
poetry install
```
Then, let's run the migration script:
```
poetry run flask db upgrade
```
Finally, run API server, Worker and Web frontend Server again.

What's Changed

fix: script rendering in message by @zxhlyh in #10807
feat: AnalyticDB vector store supports invocation via SQL. by @lpdink in #10802
chore: update .gitignore to include mise.toml by @BenjaminX in #10778
fix: correct typo in ETL type comment in .env.example by @simonmysun in #10822
refactor: improve handling of leading punctuation removal by @zandko in #10761
doc: add clarification for length limit of init password by @simonmysun in #10824
feat: add TOC to app develop doc by @hjlarry in #10799
feat: Knowledge-base-api-get-post-method-text-error-#10836 by @dajianguo in #10837
fix: close child modal on log drawer close by @kimjion in #10839
fix: upload file component support multiple by @zxhlyh in #10817
Fix: crash of workflow file upload by @JzoNgKVO in #10831
style: refactor fetch and context by @AkaraChen in #10795
Support Video Proxy and TED Embedding by @taowang1993 in #10819
feat: support json schema for gemini models by @hjlarry in #10835
fix: better WeightRerankRunner run logic use O(1) and delete unused code by @yihong0618 in #10849
fix httpx doesn't support stream parameter by @JohnJyong in #10859
fix tongyi embedding endpoint return None output by @JohnJyong in #10857
Feat/add langsmith dotted order by @ZhouhaoJiang in #10856
Encode invitee email in the invitation link by @nht1206 in #10842
fix: download some remote files raise error by @hjlarry in #10781
Fix the situation where output_tokens/input_tokens may be None in response.usage by @AAlexDing in #10728
feat: add a minimal separator between pinned apps and unpinned apps in the explore page by @kurokobo in #10871
feat: make toc panel can collapse by @hjlarry in #10875
feat: support function call for ollama block chat api by @GeorgeCaoJ in #10784
feat: enhance the custom note by @hjlarry in #8885
feat: admin can also change member role by @cuckootan in #10651
Fix : Add a process to fetch the mime type from the file name for signed url in remote_url by @kitotakumi in #10872
fix: update email validation regex to allow periods in local part by @Montaserz in #10868
fix: tool's file input display string by @hjlarry in #10887
fix(http_request): allow content type application/x-javascript by @laipz8200 in #10862
feat: support custom tool upload file by @hjlarry in #10796
ext_redis.py support redis clusters --- Fixes #9538 by @liuhaoran1212 in #9789
fix: aws presign url is not workable remote url by @ybalbert001 in #10884
Fix/input-value-type-in-moderation by @laipz8200 in #10893
Support streaming output for OpenAI o1-preview and o1-mini by @laipz8200 in #10890
fix: explicitly use new token when retrying ssePost after refresh by @llinvokerl in #10864
Gitee AI embedding tool by @fchange in #10903
fix: handle redis authentication for healthcheck command by @shisaru292 in #10907
Fix : Add a process to fetch the mime type from the file name for signed url in remote_url #10872 version2 by @kitotakumi in #10908
doc: fix better doc for api develop, droping dead hint by @yihong0618 in #10906
fix: refactor all 'or []' and 'or {}' logic to make code more clear by @yihong0618 in #10883
Add googlenews tools from rapidapi by @Tuyohai in #10877
fix image files not deleted on indexing_estimate #9541 by @wy96f in #10798
feat : Add Japanese translations for API documentation: chat, advanced-chat, completion, and workflow by @Kota-Yamaguchi in #10927
fix: chat history might be empty in log detail view by @xuzuodong in #10905
Update parse.py to handle empty list result by @PedroGomes02 in #10915
fix: default max_chunks set to 1 as other providers by @yihong0618 in #10937
fix(gpt-4o-audio-preview): Remove the vision feature by @laipz8200 in #10932
fix weight rerank of knowledge retrieval by @AkisAya in #10931
fix: external dataset missing score_threshold_enabled by @hanqingwu in #10943
Resolve 8475 support rerank model from infinity by @LastHopeOfGPNU in #10939
Feat: support json output for bing-search by @xu-song in #10904
fix: fix bugs of frontend-workflow panel operator by @marvin-season in #10945
feat: add cURL import for http request node by @hjlarry in #8656
fix: tiny lora bug found by mypy by @yihong0618 in #10959
fix: unstructured io credential environment variables missing by @fdb02983rhy in #10953
feat: add gpt-4o-2024-11-20 by @agungbesti in #10951
feat: slidespeak slides generation by @fdb02983rhy in #10955
refactor: split manual from auto generation css variables by @WTW0313 in #10961
chore: Added the new gemini exp-1121 and learnlm-1.5 models by @CXwudi in #10963
feat: Check and compare the DSL version before import an app by @laipz8200 in #10969
chore: translate i18n files by @github-actions in #10970
Feat: update icon and Divider components by @douxc in #10975
fix(model): make sure AppModelConfig.model_dict returns a dict. by @laipz8200 in #10972
feat: Allow using file variables directly in the LLM node and support more file types. by @laipz8200 in #10679
feat: add a meta(mac) ctrl(windows) key by @crazywoola in #10978
chore(app_dsl_service): Downgrade DSL Version by @laipz8200 in #10979
feat: support LLM process document file by @hjlarry in #10966
chore: update base image to Python 3.12 in Dockerfile by @laipz8200 in #10358
fix: rules should not be None for in by @yihong0618 in #10977
chore (dep): bump gevent from v23 to v24 for better support for Python 3.11 and 3.12 by @bowenliang123 in #10387
fix: fish audio wrong validate credentials interface by @yihong0618 in #11019
chore(devcontainer): upgrade Python version to 3.12 in Dockerfile and configuration by @fujita-h in #11017
fix: gitlab file url not correctly encoded by @AkisAya in #10996
Fixing #11005: Incorrect max_tokens in yaml file for AWS Bedrock US Cross Region Inference version of 3.5 Sonnet v2 and 3.5 Haiku by @kazuhisa-wada in #11013
feat: Add support for TEI API key authentication by @kenwoodjw in #11006
doc: Updated Python version requirements to match English version by @fujita-h in #11015
Fix typo by @johnpccd in #11024
fix: uuid not import bug by @yihong0618 in #11014
fix: wrong convert in PromptTemplateConfigManager by @yihong0618 in #11016
chore: bump minimum supported Python version to 3.11 by @bowenliang123 in #10386
fix: json parse err when http node send request by @litterGuy in #11001
fix error with xinference tool calling with qwen2-instruct and add timeout retry setttings for xinference by @cyflhn in #11012
feat(file_factory): Standardize custom file type into known types by @laipz8200 in #11028
fix: query will be None if the query_prompt_template not exists by @laipz8200 in #11031
fix: Japanese typo by @TakakiMoriguchi in #11034
fix: Validate file only when file type is set to custom by @laipz8200 in #11036
fix: wrong param name by @mazyu36 in #11039
make tool parameters parsing compatible with the response of glm4 model in xinference provider when function tool call integerated by @cyflhn in #11049
Fix Deepseek Function/Tool Calling by @taowang1993 in #11023
ISSUE=11042: add tts model in siliconflow by @senseb in #11043
Feat/add admin check by @KMerdan in #11050
fix the wrong LINDORM_PASSWORD variable name in docker-compose.yaml by @AlwaysBluer in #11052
chore: update chromadb version to 0.5.20 by @yihong0618 in #11038
fix: slidespeak text output is not the download link by @fdb02983rhy in #10997
fix: llm node do not pass sys.query in chatflow app init by @iamjoel in #11053
feat(i18n): update Japanese translation for login page by @nomi3 in #10993
fix(workflow_entry): Support receive File and FileList in single step run. by @laipz8200 in #10947
Improvement: update api doc of workflow by @JzoNgKVO in #11054
fix: better path trigger for vdb and fix the version by @yihong0618 in #11057
feat: add cookie management by @iamjoel in #11061
fix: ignore empty outputs in Tool node by @PedroGomes02 in #10988
fix: WorkflowNodeExecution.created_at may be earlier than WorkflowRun.created_at by @laipz8200 in #11070
chore: bump version to 0.12.0 by @laipz8200 in #11056

New Contributors

@simonmysun made their first contribution in #10822
@zandko made their first contribution in #10761
@nht1206 made their first contribution in #10842
@AAlexDing made their first contribution in #10728
@kitotakumi made their first contribution in #10872
@Montaserz made their first contribution in #10868
@Tuyohai made their first contribution in #10877
@wy96f made their first contribution in #10798
@hanqingwu made their first contribution in #10943
@LastHopeOfGPNU made their first contribution in #10939
@xu-song made their first contribution in #10904
@marvin-season made their first contribution in #10945
@agungbesti made their first contribution in #10951
@kazuhisa-wada made their first contribution in #11013
@kenwoodjw made their first contribution in #11006
@litterGuy made their first contribution in #11001
@cyflhn made their first contribution in #11012
@TakakiMoriguchi made their first contribution in #11034
@KMerdan made their first contribution in #11050
@nomi3 made their first contribution in #10993

Full Changelog: 0.11.2...0.12.0

langgenius/dify 0.12.0 v0.12.0 on GitHub