v0.2.31

Highlight

• Enhanced support of Anthropic client
• Enhanced documentation on LLM observability support
  Thanks to @marklysze, @areibman, @qingyun-wu and all the contributors!

What's Changed

• Fixed alternating message role bug in Anthropic client by @marklysze in #2992
• Anthropic Client - Handling multiple consecutive function calls by @marklysze in #2997
• LLM Observability documentation fixes: Broken links, grammar, and spelling by @areibman in #2995
• bump version by @qingyun-wu in #2999

New Contributors

• @areibman made their first contribution in #2995

Full Changelog: v0.2.30...v0.2.31

v0.2.30

Highlights

• 🔥 Enhanced Non-OpenAI Model Support:
  • Anthropic Client: claude-3-5-sonnet-20240620 is seamlessly supported! Check a group chat example between OpenAI GPT models and Claude 3.5 Sonnet in this Anthropic Client notebook
  • Mistral Client
  • Together.AI Client

Thanks to @marklysze, @Hk669, @yiranwu0, @garnermccloud, @qingyun-wu, @thinkall, @olaoluwasalami, @eltociear, @LittleLittleCloud, @ShobhitVishnoi30, @NanthagopalEswaran and all the other contributors.

What's Changed

• Client class utilities by @marklysze in #2949
• change specified api-version by @olaoluwasalami in #2955
• Update agentchat_function_call_currency_calculator.ipynb by @eltociear in #2952
• Bump ws from 7.5.9 to 7.5.10 in /website by @dependabot in #2964
• Function to indicate whether to hide tools for client_utils by @marklysze in #2966
• Anthropic Client by @Hk669 in #2931
• a_initaite_chats update by @ShobhitVishnoi30 in #2958
• Fix #2845 - LocalCommandLineCodeExecutor is not working with virtual environments by @NanthagopalEswaran in #2926
• type fix for ChatResult by @Hk669 in #2973
• Fix #2960 by checking if the values are a list of lists. by @thinkall in #2971
• [.Net] fix #2859 by @LittleLittleCloud in #2974
• [.Net] add ReAct sample by @LittleLittleCloud in #2977
• Mistral Client by @marklysze in #2892
• Set qdrant version to <1.9.2 to fix test errors by @thinkall in #2984
• Anthropic client fixes by @garnermccloud in #2981
• Together AI Client by @marklysze in #2919
• Uniform Interface for calling different LLMs by @yiranwu0 in #2916
• fix: created in ChatCompletion in non-oai clients by @Hk669 in #2988
• Bump version to 0.2.30 by @qingyun-wu in #2990
• update notebook wording and format by @qingyun-wu in #2991

New Contributors

• @NanthagopalEswaran made their first contribution in #2926
• @garnermccloud made their first contribution in #2981

Full Changelog: v0.2.29...v0.2.30

v0.2.29

Highlights

• 🔥 Agent Integration: Llamaindex agent integration
  • Llamaindex agent in a group chat
• 🔥 Observability: AgentOps Runtime Logging Integration
  • AgentChat with AgentOps Code Example
• 🔥 AutoGen.Net: AutoGe.Net new 0.0.15 release which add Gemini support.
• Gemini support improvements: Latest Example of Using Gemini in AutoGen with other LLMs
• Azure client improvements to support AAD auth.

Thanks to @colombod, @krishnashed, @sonichi, @thinkall, @luxzoli, @LittleLittleCloud, @afourney, @WaelKarkoub, @aswny, @bboynton97, @victordibia, @DavidLuong98, @Knucklessg1, @Noir97, @davorrunje, @ken-gravilon, @yiranwu0, @TheTechOddBug, @whichxjy, @LeoLjl, @qingyun-wu, and all the other contributors!

What's Changed

• Add llamaindex agent integration by @colombod in #2831
• Broken links fix by @krishnashed in #2843
• update guide about roadmap issues by @sonichi in #2846
• Fix chromadb get_collection ignores custom embedding_function by @thinkall in #2854
• Use Gemini without API key by @luxzoli in #2805
• Refactor hook registration and processing methods by @colombod in #2853
• [.Net] Add AOT compatible check for AutoGen.Core by @LittleLittleCloud in #2858
• Updated the azure client to support AAD auth. by @afourney in #2879
• add github icon to AutoGen.Net website by @LittleLittleCloud in #2878
• [Refactor] Transforms Utils by @WaelKarkoub in #2863
• allow function to remove termination string in groupchat by @aswny in #2804
• AgentOps Runtime Logging Implementation by @bboynton97 in #2682
• Autogenstudio docs by @victordibia in #2890
• [.Net] Add Goolge gemini by @LittleLittleCloud in #2868
• [.Net] Support image input for Anthropic Models by @DavidLuong98 in #2849
• version update by @sonichi in #2908
• Bugfix: PGVector/RAG - Calculate the Vector Size based on Model Dimensions by @Knucklessg1 in #2865
• Update notebook agentchat_microsoft_fabric by @thinkall in #2886
• Change chunk size of vectordb from max_tokens to chunk_token_size by @Noir97 in #2896
• Chore: pre-commit version update and a few spelling fixes by @davorrunje in #2913
• Chore: CRLF changed to LF by @davorrunje in #2915
• Improve update context condition checking rule by @thinkall in #2883
• Docs typo cli-code-executor.ipynb by @ken-gravilon in #2909
• Fixes human_input_mode annotations by @WaelKarkoub in #2864
• [.Net] Add Gemini samples to AutoGen.Net website + configure Gemini package to be ready for release by @LittleLittleCloud in #2917
• Fix CRLF file format by @davorrunje in #2935
• Bump braces from 3.0.2 to 3.0.3 in /website by @dependabot in #2934
• update release log for AutoGen.Net 0.0.15 by @LittleLittleCloud in #2937
• Allow passing in custom pricing in config_list by @yiranwu0 in #2902
• Fix line numbers within instructions in comments. by @TheTechOddBug in #2867
• Fix typo: double comma by @whichxjy in #2940
• [.Net] update oai tests by using new OpenAI resources by @LittleLittleCloud in #2939
• [Autobuild] improve robustness and reduce cost by @LeoLjl in #2907
• Filter models with tags instead of model name by @qingyun-wu in #2912
• Fix missing messages in Gemini history by @luxzoli in #2906

New Contributors

• @colombod made their first contribution in #2831
• @luxzoli made their first contribution in #2805
• @aswny made their first contribution in #2804
• @bboynton97 made their first contribution in #2682
• @Noir97 made their first contribution in #2896
• @ken-gravilon made their first contribution in #2909
• @TheTechOddBug made their first contribution in #2867
• @whichxjy made their first contribution in #2940

Full Changelog: v0.2.28...v0.2.29

v0.2.28

Highlights

• Guide for GPTAssistantAgent and function calling example.
• New feature: resumable group chat.
• New transformation capability: text compression using LLMLingua.
• Experimental integration of AgentEval.
• New notebook examples:
  • Azure Cognitive Search and Azure Identity.
  • Mitigating Prompt hacking with JSON Mode in Autogen.
  • DBRX (Databricks LLM)
• New gallery example: AutoGen Virtual Focus Group. A virtual consumer focus group with multiple custom personas, product details, and final analysis created with AutoGen, Ollama/Llama3, and Streamlit.
• Improvement in code execution, RAG, logging, studio, reflection, CAP, nested chat, group chat, function call.

Thanks to @beyonddream @ginward @gbrvalerio @LittleLittleCloud @thinkall @asandez1 @DavidLuong98 @jtrugman @IANTHEREAL @ekzhu @skzhang1 @erezak @WaelKarkoub @zbram101 @r4881t @eltociear @robraux @thongonary @moresearch @shippy @marklysze @ACHultman @Gr3atWh173 @victordibia @MarianoMolina @jluey1 @msamylea @Hk669 @ruiwang @rajan-chari @michaelhaggerty @BeibinLi @krishnashed @jtoy @NikolayTV @pk673 @Aretai-Leah @Knucklessg1 @tj-cycyota @tosolveit @MarkWard0110 @Mai0313 and all the other contributors!

What's Changed

• Remove unneeded duplicate check for pydantic v1 since we are already checking that in the else block. by @beyonddream in #2467
• Update token_count_utils.py by @ginward in #2531
• feat: add bind_dir arg to DockerCommandLineExecutor + docs update by @gbrvalerio in #2309
• use conditional check to replace path filter in build and dotnet-ci workflow by @LittleLittleCloud in #2546
• Fix chroma import error by @thinkall in #2557
• Docker multilanguage executor saver with policy by @asandez1 in #2522
• Adding an action to set workflow as success when no change is made in target paths by @LittleLittleCloud in #2553
• Update dotnet-build.yml to add merge_group trigger by @LittleLittleCloud in #2567
• [.Net] Support raw-data in ImageMessage by @DavidLuong98 in #2552
• [.NET] Return ChatCompletions instead of ChatResponseMessage for token usage. by @DavidLuong98 in #2545
• Function Calling with GPTAssistantAgent by @jtrugman in #2375
• Add a guide doc for GPTAssistantAgent by @IANTHEREAL in #2562
• Add note in the lfs check action to help contributors fix Git LFS check failure. by @ekzhu in #2563
• add faq for autogen and openai assistant compatible version by @IANTHEREAL in #2587
• Fix for http client by @AbdurNawaz in #2579
• [.Net] refactor over streaming version api by @LittleLittleCloud in #2461
• Update AgentOptimizer BibTeX by @skzhang1 in #2578
• Correct link to Jupyter Code Executor in code-executors.ipynb by @erezak in #2589
• Text Compression Transform by @WaelKarkoub in #2225
• notebook showing assistant agents connecting azure ai search and azur… by @zbram101 in #2594
• Update to correct pip install for litellm by @r4881t in #2602
• docs: update tutorial.ipynb by @eltociear in #2606
• fix: event logging with nested chats by @robraux in #2600
• [.Net] fix #2609 by @LittleLittleCloud in #2618
• [.Net] Add an example to show how to connect to third party OpenAI API endpoint + upgrade Azure.AI.OpenAI package by @LittleLittleCloud in #2619
• [.Net]: Introduce ChatCompletionAgent to AutoGen.SemanticKernel package by @DavidLuong98 in #2584
• Fix chess example by @thongonary in #2631
• [.Net] Add KernelPluginMiddleware in AutoGen.SemanticKernel by @LittleLittleCloud in #2595
• [.Net] release note for 0.0.13 by @LittleLittleCloud in #2641
• Update graph_utils.py by @moresearch in #2601
• Add instructions for Docker issue with hash mismatch to FAQ by @shippy in #2639
• [.Net] Fix 2652 && 2650 by @LittleLittleCloud in #2655
• Resuming a GroupChat by @marklysze in #2627
• fix notebook doc typo by @ACHultman in #2642
• Feature: Add ability to use a separate python environment in local executor by @Gr3atWh173 in #2615
• Rewrite AutoGen Studio Database Layer to Use SQLModel ORM by @victordibia in #2425
• [.Net] Remove Workflow class && bump version to 0.0.14 by @LittleLittleCloud in #2675
• [.Net] Fix #2660 and add tests for AutoGen.DotnetInteractive by @LittleLittleCloud in #2676
• Add role to reflection with llm by @MarianoMolina in #2527
• Agenteval integration by @jluey1 in #2672
• AutoGen Virtual Focus Group by @msamylea in #2598
• Adding gpt-4o to pricing by @r4881t in #2674
• pricing url fixed by @Hk669 in #2684
• [.Net] feature: Ollama integration by @iddelacruz in #2693
• [.Net] Fix #2687 by adding global:: keyword in generated code by @LittleLittleCloud in #2689
• update news by @sonichi in #2694
• [.Net] Set up Name field in OpenAIMessageConnector by @LittleLittleCloud in #2662
• Custom Runtime Logger <> FileLogger by @Hk669 in #2596
• Update groupchat.py to remove Optional type hint when they are not ch… by @ruiwang in #2703
• Add gpt4o token count to the utils. by @Hk669 in #2717
• [CAP] Improved AutoGen Agents support & Pip Install by @rajan-chari in #2711
• [.Net] fix #2722 by @LittleLittleCloud in #2723
• [.Net] Mark Message as obsolete and add ToolCallAggregateMessage type by @LittleLittleCloud in #2716
• Add nuget package badge to readme by @LittleLittleCloud in #2736
• Update human-in-the-loop.ipynb by @michaelhaggerty in #2724
• [CAP] Refactor: Better Names for classes and methods by @rajan-chari in #2734
• Avoid requests 2.32.0 to fix build by @ekzhu in #2761
• Debug: Gemini client was not logged and causing runtime error by @BeibinLi in #2749
• [Add] Fix invoking Assistant API by @krishnashed in #2751
• Add silent option in nested chats and group chat by @robraux in #2712
• Fix the assistant test case error caused by openai incompatible change by @IANTHEREAL in #2718
• add warning if duplicate function is registered by @jtoy in #2159
• Ability to ignore Select Speaker Prompt for GroupChat by @marklysze in #2726
• added Gemini safety setting and Gemini generation config by @NikolayTV in #2429
• Update Deprecation Warning for CompressibleAgent and TransformChatHistory by @WaelKarkoub in #2685
• Fix for runtime logging not supported with GPTAssistantAgent by @pk673 in #2659
• Ignore Some Messages When Transforming by @WaelKarkoub in #2661
• [.Net] rename Autogen.Ollama to AutoGen.Ollama and add more test cases to AutoGen.Ollama by @LittleLittleCloud in #2772
• [.Net] add AutoGen.SemanticKernel.Sample project by @LittleLittleCloud in #2774
• [.Net] add ollama-sample and adds more tests by @LittleLittleCloud in ht...

v0.2.27

Highlights

• New language support: AutoGen.NET.
• Support OpenAI assistant v2 API.
• New features: Allow initializing an agent with message history; event logging.
• More robust group chat: Re-query speaker name when multiple speaker names returned during Group Chat speaker selection.
• New language support in code execution: HTML, CSS and Javascript in LocalCommandLineCodeExecutor.
• New caching backend using Azure Cosmos DB.

Thanks to @thinkall @WaelKarkoub @luxuncang @IANTHEREAL @LittleLittleCloud @lalo @asandez1 @themataleao @Hk669 @wmwxwa @Knucklessg1 @giorgossideris @marklysze and all the other contributors!

What's Changed

• Update RetrieveChat extra dependencies by @thinkall in #2449
• Adds Message History to a ConversableAgent by @WaelKarkoub in #2437
• Make the port number optional in JupyterConnectionInfo() by @luxuncang in #2473
• Support openai assistant v2 API by @IANTHEREAL in #2466
• enable lfs support in dotnet workflow by @LittleLittleCloud in #2483
• feat: add event logging api and more tracing by @lalo in #2478
• Add support for HTML, CSS and Javascript in LocalCommandLineCodeExecutor with Mapping executor/saver #2303 by @asandez1 in #2464
• Fix python not supported in macos by @thinkall in #2503
• fix typo by @themataleao in #2497
• fix: autogen.AssistantAgent at https://microsoft.github.io/autogen/blog/2023/10/18/RetrieveChat/ by @Hk669 in #2504
• AutoGen cache using Azure Cosmos DB by @wmwxwa in #2327
• Bring Dotnet AutoGen by @LittleLittleCloud in #924
• 2447 fix pgvector tests and notebook by @Knucklessg1 in #2455
• Use config list in notebook by @sonichi in #2537
• Streamline Testing with pytest-cov and pytest Defaults by @WaelKarkoub in #2490
• tags for config in notebooks by @sonichi in #2539
• Min tokens in token limiter by @giorgossideris in #2400
• [.Net] fix code ql for dotnet build && update trigger for dotnet workflow by @LittleLittleCloud in #2529
• Re-query speaker name when multiple speaker names returned during Group Chat speaker selection by @marklysze in #2304

New Contributors

• @luxuncang made their first contribution in #2473
• @asandez1 made their first contribution in #2464
• @themataleao made their first contribution in #2497

Full Changelog: v0.2.26...v0.2.27

v0.2.26

Highlights

• New contrib feature: customizable vector db for retrieval-augmented chat. PGVector example and notebook.
• New integration example with promptflow.
• Enhance initiate_chats for selective carryover.
• Support openai 'sk-proj-' api key format.

Thanks to @thinkall @ekzhu @Hk669 @lordlinus @ShobhitVishnoi30 @saberuster @olgavrou @Knucklessg1 and all the other contributors!

What's Changed

• Support setting vector_db as a param by @thinkall in #2313
• Fix function call notebook to make sure title is displayed. by @ekzhu in #2403
• Updated type hints for Agent Capabilities by @Hk669 in #2408
• add promptflow example by @lordlinus in #2369
• Add link to PromptFlow Sample App by @ekzhu in #2418
• improve test speed by @sonichi in #2406
• initate chats enhancement by @ShobhitVishnoi30 in #2404
• Update the auto_feedback_from_code_execution notebook to use code executor by @ekzhu in #2366
• support openai 'sk-proj-' apikey format by @saberuster in #2426
• speed up notebook in CI and fix CI failures by @sonichi in #2427
• Add Git LFS check action, fix broken lfs files by @ekzhu in #2436
• Use latest APIs for tool use, code execution and cache in topic pages by @ekzhu in #2071
• conversable agent: actually print the response in the warning by @olgavrou in #2438
• [Re-Opened] Support for PGVector Database in Autogen by @Knucklessg1 in #2439

New Contributors

• @saberuster made their first contribution in #2426
• @Knucklessg1 made their first contribution in #2439

Full Changelog: v0.2.25...v0.2.26

v0.2.25

Highlights

• New model support: Gemini.
• New notebook: Writing a software application using function calls.
• New gallery example from the AutoGPT Hackathon winner team: AutoTx: Your AI-Powered Transaction Companion powered by AutoGen.
• Update price info for new gpt-4-turbo model.

Thanks to @BeibinLi @GregorD1A1 @rihp @Mai0313 @DustinX @skzhang1 and all the other contributors!

What's Changed

• Update chat-termination.ipynb by @DustinX in #2393
• Update code-executors.ipynb by @DustinX in #2394
• Update code-executors.ipynb by @DustinX in #2396
• add AgentOptimizer test in CI by @skzhang1 in #2380
• [Feat] browser should support custom base url for bing search by @Mai0313 in #2321
• Update code-executors.ipynb by @DustinX in #2395
• Code writing with a function call notebook created by @GregorD1A1 in #2121
• DOCS: Add AutoTX to gallery by @rihp in #2390
• Merge "Gemini" feature into the main branch by @BeibinLi in #2360
• update cost info by @sonichi in #2401

New Contributors

• @DustinX made their first contribution in #2393
• @Mai0313 made their first contribution in #2321
• @rihp made their first contribution in #2390

Full Changelog: v0.2.24...v0.2.25

v0.2.24

Highlights

• Generalized and Improved RAG: customizable vector database; adding source to answer.
• Function call support for Anthropic Claude.
• Usage and cost info are made easier to read.
• Improvement to long context handling, stateflow, agent optimizer, code execution.

Thanks to @thinkall @kevin666aa @WaelKarkoub @skzhang1 @PyroGenesis @wmwxwa @levscaut @JoshTrim @giorgossideris @Hk669 and all the other contributors!

What's Changed

• Add source to the answer for default prompt by @thinkall in #2289
• Add a vectordb module by @thinkall in #2263
• Update StateFlow to use LocalCommandLineCodeExecutor by @kevin666aa in #2335
• [Fix] Improves Token Limiter by @WaelKarkoub in #2350
• support llm_config in AgentOptimizer by @skzhang1 in #2299
• fix markdown for long context user guide by @PyroGenesis in #2351
• Add extra_require for cosmosdb in setup.py by @wmwxwa in #2355
• [Fix] Handle Missing Contents in Messages in MessageTokenLimiter by @WaelKarkoub in #2364
• add support for Anthropic Claude function call by @levscaut in #2311
• Add "py" as lang in conversable agent (#1062) by @JoshTrim in #2144
• Standardize printing of MessageTransforms by @giorgossideris in #2308
• Made the cost info easier to read by @Hk669 in #2356
• fix local test failures by @sonichi in #2386

New Contributors

• @wmwxwa made their first contribution in #2355
• @JoshTrim made their first contribution in #2144
• @giorgossideris made their first contribution in #2308
• @Hk669 made their first contribution in #2356

Full Changelog: v0.2.23...v0.2.24

v0.2.23

This is a minor release with cost update for OpenAI models.

Highlights

• New sample app using websockets with FastAPI and AutoGen
• New user guide for long context handling

Thanks to @liujun121533 @davorrunje @martinzrrl @thinkall @WaelKarkoub @sethupavan12 and all the other contributors!

What's Changed

• Use assistant tag to filter config by @sonichi in #2316
• Fix typo in conversation-patterns.ipynb by @liujun121533 in #2322
• Add Websockets example by @davorrunje in #2291
• Improve readability of notebooks by @martinzrrl in #2200
• cost update for gpt-3.5-turbo-0125 by @sonichi in #2332
• Add back to top and year by @thinkall in #2324
• Document Migration to TransformMessages by @WaelKarkoub in #2247
• Fix docs related to Azure OpenAI's "model" by @sethupavan12 in #2315

New Contributors

• @liujun121533 made their first contribution in #2322
• @martinzrrl made their first contribution in #2200
• @sethupavan12 made their first contribution in #2315

Full Changelog: v0.2.22...v0.2.23

v0.2.22

Highlights

• New capability TransformMessage: a generalization of previous long context handling capabilities. The previous capabilities will be deprecated.
• New research AutoDefense: Defend against jailbreak attacks with AutoGen.
• New support for non-OpenAI models:
  • Anthropic Claude
  • Mistral for Group chat.
  • Tips for non-OpenAI models.
• New user guide for custom speaker selection in group chat
• Update of StateFlow blogpost and paper.
• New notebook on web scraping with tool use.
• New example in gallery: Create Issues from Code Commits.

Thanks to @WaelKarkoub @ekzhu @skzhang1 @davorrunje @afourney @Wannabeasmartguy @jackgerrits @rajan-chari @XHMY @jtoy @marklysze @Andrew8xx8 @thinkall @BeibinLi @benstein @sharsha315 @levscaut @Karthikeya-Meesala @r-b-g-b @cheng-tan @kevin666aa and all the other contributors!

What's Changed

• Transform Messages Capability by @WaelKarkoub in #1923
• Bump express from 4.18.2 to 4.19.2 in /website by @dependabot in #2157
• Add website analytics by @ekzhu in #2201
• Render AgentOptimzier notebook in web by @skzhang1 in #2197
• Replace iostream warning message with info by @davorrunje in #2207
• AutoGenBench: Handle Ctrl-C more gracefully. by @afourney in #2174
• fix the problem config_list_from_dotenv() can't correctly get other p… by @Wannabeasmartguy in #2172
• Fix generate_init_message for Multimodal Messages by @WaelKarkoub in #2124
• Support string based UDFs by @jackgerrits in #2195
• Ignore protobuf generated files in CAP by @rajan-chari in #2215
• AutoDefense Blog by @XHMY in #1982
• 2147 - Fix notebook bug to make a copy of configs for each agent by @jtoy in #2176
• Improve Image Generation Notebook by @WaelKarkoub in #2213
• Added ability to specify 'role' field for select speaker messages for Group Chats (Replaces PR #2167) by @marklysze in #2199
• Fixed types related to function calling by @davorrunje in #2230
• Replace unofficial with official pre-commit hook for ruff by @davorrunje in #2229
• Re-commit of #2182: Customisation of speaker select message and prompt in GroupChat by @marklysze in #2233
• Fix websockets demo notebook by @davorrunje in #2227
• Fixed spelling in agentchat_transform_messages.ipynb by @Andrew8xx8 in #2234
• Improving website navigation and help click-through analytics collection by @ekzhu in #2205
• Dont include api ref tree in docs sidebar by @jackgerrits in #2244
• Fix unstructured deps installation error by @thinkall in #2248
• Add PNG and JPEG files to LFS automatically by @davorrunje in #2226
• Allow custom name for functions module by @jackgerrits in #2241
• Reduce sleep() in CAP library code by @rajan-chari in #2189
• Notebook on web scraping with tool use by @ekzhu in #2192
• Break up contributor guide for better navigability by @ekzhu in #2250
• Update vision capability test: create dummy image before testing by @BeibinLi in #2258
• Auto collapse side bar entries when selecting other sections by @jackgerrits in #2257
• Fix Vision Capability Test for MacOS by @BeibinLi in #2260
• Update listdir() in sample code docs to take correct args by @benstein in #2261
• Multiline docstrings fix by @sharsha315 in #2130
• Add in memory cache class by @jackgerrits in #2266
• Add support for Anthropic Claude models by @levscaut in #2245
• Composio: add new example by @Karthikeya-Meesala in #2237
• Re-commit of #2222: Expanded speaker name matching during speaker selection by @marklysze in #2267
• RetrieveUserProxyAgent, use context_max_tokens from retrieve_config if provided by @r-b-g-b in #2259
• fix logging cost flaky test by @cheng-tan in #2277
• Remove subdirectory .gitattributes files and add relevant files to lfs by @ekzhu in #2270
• Update RetrieveChat blog to fix some outdated APIs by @thinkall in #2273
• Add support for system message in Claude Notebook by @levscaut in #2280
• add composio to ecosystem page by @Karthikeya-Meesala in #2272
• Add isort by @thinkall in #2265
• Remove info log for IOStream using default. by @ekzhu in #2281
• Fix LMM test in CI by installing lfs by @ekzhu in #2282
• Fix FileNotFoundError english-words.txt in RetrieveChatTest by @thinkall in #2284
• Add html parser for RAG and some improvements by @thinkall in #2271
• Composio: update text - "Composio Logo" -> "Composio Example" by @Karthikeya-Meesala in #2288
• bump version to 0.2.22 by @sonichi in #2256
• Copy llm_config to avoid external changes by @jackgerrits in #2290
• Fix typo in websockets notebook by @davorrunje in #2287
• Fix types in tool tests by @davorrunje in #2285
• [Bug Fix] [CAP] Sleep removal introduced an Actor resolution bug. by @rajan-chari in #2296
• Add Custom GroupChat Speaker Selection to tutorial by @kevin666aa in #2219
• Update StateFlow blog by @kevin666aa in #2262
• Fix cost calculation when completion_token is None by @kevin666aa in #2292

New Contributors

• @Wannabeasmartguy made their first contribution in #2172
• @XHMY made their first contribution in #1982
• @jtoy made their first contribution in #2176
• @Andrew8xx8 made their first contribution in #2234
• @benstein made their first contribution in #2261
• @sharsha315 made their first contribution in #2130
• @levscaut made their first contribution in #2245
• @Karthikeya-Meesala made their first contribution in #2237
• @r-b-g-b made their first contribution in #2259

Full Changelog: v0.2.21...v0.2.22