v0.8.3

Highlights

• FIXED: LLMConfig bug that associated an agent's tools with other agents using the same LLM Configuration when using the context manager
• 🌐 BrowserUseTool can now return the URLs used
• 🚀 Anthropic client class now supported by MultimodalConversableAgent for images (great for OCR)
• 🔍 ReasoningAgent improved alignment through prompting and code execution config fix
• 🛠️📖 Fixes and documentation improvements

♥️ Thanks to all the contributors and collaborators that helped make the release happen!

New Contributors

• @jk10001 made their first contribution in #1273

What's Changed

• bug fix for max_consecutive_auto_reply by @marufaytekin in #1374
• [Docs] Generate notebooks section by @harishmohanraj in #1378
• Fix CI by @harishmohanraj in #1398
• [Docs] Add broken link checker for mkdocs site by @harishmohanraj in #1399
• Use LLMConfig in function signatures properly by @kumaranvpl in #1400
• Align ReasoningAgent through prompting by @giorgossideris in #1345
• ReasoningAgent fix code_execution_config by @giorgossideris in #1362
• Use LLMConfig in notebooks by @kumaranvpl in #1391
• Use LLMConfig instead of dict in blogs by @kumaranvpl in #1393
• [Docs] Fix edit links by @harishmohanraj in #1407
• Update BrowserUseResult by @rjambrecic in #1414
• Fix bug in filter_config while using string instead of list as filter… by @kumaranvpl in #1417
• BrowserUseTool quick fix for multiple tool calls by @rjambrecic in #1421
• Anthropic image format fix by @jk10001 in #1273
• Updated OllamaLLMConfigEntry defaults by @marklysze in #1360
• Fix: AttributeError: 'HttpUrl' object has no attribute 'partition' in Ollama Client by @marufaytekin in #1426
• Add voice as parameter and use SecretStr for relevant fields by @kumaranvpl in #1425
• Copy llmconfig by @kumaranvpl in #1428
• Fix usage of LLMConfig in docs guides by @kumaranvpl in #1415
• minor update link Getting-Started.mdx by @sky-coderay in #1387
• Updated model provider compatibility in Structured Output notebook by @marklysze in #1433
• Bump 0.8.3 by @marklysze in #1432

Full Changelog: v0.8.2...v0.8.3

v0.8.2

Highlights

• 🔍 Add real-time web searches to your agents using the new Google Search Tool!
• 📝 LLM configurations can now use a new type-safe LLMConfig object
• 📔⚡ DocAgent can now add citations! See how…
• 🦙🔍 DocAgent can now use any LlamaIndex vector store for embedding and querying its ingested documents! See how...
• 🐝 9 x Swarm patterns with full code examples
• Termination messages added to indicate why a workflow ended
• 📖 Many improvements to documentation
• 🛠️ Fixes, fixes and more fixes

♥️ Thanks to all the contributors and collaborators that helped make the release happen!

What's Changed

• Updated support for OpenAI package 1.66.2 by @marklysze in #1320
• [Docs] Add blogs to mkdocs by @harishmohanraj in #1312
• Remove pysqlite3-binary dependency by @rjambrecic in #1301
• Fix failing tests on Windows by @harishmohanraj in #1329
• Add row to MAINTAINERS.md by @giorgossideris in #1317
• Integrate fast-depends into the repo by @davorrunje in #1327
• Add guidance on using GOOGLE_GEMINI_API_KEY by @marklysze in #1318
• Termination messages by @marklysze in #1336
• Correct code execution document text by @marklysze in #1335
• Suppress warnings by @marklysze in #1334
• [Docs]: Swarm Patterns documentation by @marklysze in #1255
• Add package version to import utils by @davorrunje in #1331
• Implement GoogleSearchTool by @rjambrecic in #1315
• Release 0.8.2rc0 by @davorrunje in #1346
• [DocAgent] Add Text Citation by @AgentGenie in #1338
• Ollama response_format fix by @marklysze in #1347
• Fix get load param if needed function by @davorrunje in #1344
• Add LLMConfig class by @kumaranvpl in #1268
• Handle query_engine response while constructing AnswerWithCitations by @kumaranvpl in #1353
• Fix pydantic import issue for version <2.10.2 by @davorrunje in #1356
• Add UUID to terminating messages by @marklysze in #1348
• Accept config with and without config_list in LLMConfig by @kumaranvpl in #1355
• Removed jsonref package as optional dependancy of gemini by @davorrunje in #1358
• Create LlamaIndex query engine by @Eric-Shang in #1311
• Fix import problem for llm_config by @davorrunje in #1359
• Type fixes for LlamaIndexQueryEngine by @marklysze in #1364
• [Docs] Use LLMConfig instead of config dict by @kumaranvpl in #1354
• Add Mac OS Instructions in pip install by @CAROLZXYZXY in #1372
• Update packages by @davorrunje in #1375
• Add necessary utilities functions to llm config by @kumaranvpl in #1376
• Import optional dependencies inside with optional_import_block by @kumaranvpl in #1380
• Update get_key in openai_utils to fix serialisation issue by @marklysze in #1385
• Version bump 0.8.2 by @marklysze in #1388

Full Changelog: v0.8.1...v0.8.2

v0.8.2rc0

Highlights

• 🤖 Updated support for OpenAI package 1.66.2
• ⛓️‍💥 Removed dependencies: pysqlite3-binary and fast-depends
• 🔍 Added GoogleSearchTool
• 📖 Many improvements to documentation and the API reference
• 🛠️ Fixes, fixes and more fixes

What's Changed

• Updated support for OpenAI package 1.66.2 by @marklysze in #1320
• [Docs] Add blogs to mkdocs by @harishmohanraj in #1312
• Remove pysqlite3-binary dependency by @rjambrecic in #1301
• Fix failing tests on Windows by @harishmohanraj in #1329
• Add row to MAINTAINERS.md by @giorgossideris in #1317
• Integrate fast-depends into the repo by @davorrunje in #1327
• Add guidance on using GOOGLE_GEMINI_API_KEY by @marklysze in #1318
• Termination messages by @marklysze in #1336
• Correct code execution document text by @marklysze in #1335
• Suppress warnings by @marklysze in #1334
• [Docs]: Swarm Patterns documentation by @marklysze in #1255
• Add package version to import utils by @davorrunje in #1331
• Implement GoogleSearchTool by @rjambrecic in #1315
• Release 0.8.2rc0 by @davorrunje in #1346

Full Changelog: v0.8.1...v0.8.2rc0

v0.8.1

Highlights

• 🧠 Google GenAI's latest package is now supported
• 📔 DocAgent now utilises OnContextCondition for a faster and even more reliable workflow
• 🐝 Swarm function registration fixes and notebook improvements
• 📖 Many improvements to documentation and the API reference
• 🛠️ Fixes, fixes and more fixes

♥️ Thanks to all the contributors and collaborators that helped make the release happen!

New Contributors

• @annwag made their first contribution in #1261
• @hexcow made their first contribution in #1270
• @sky-coderay made their first contribution in #1272

What's Changed

• Communication Agent and Tools documentation by @marklysze in #1249
• Configure ruff pre-commit rule to mandate the documentation for every function parameter. by @harishmohanraj in #1244
• Remove obsolete completion by @kumaranvpl in #1223
• [Docs] Fix docstring related mkdocs warnings by @harishmohanraj in #1257
• Add user guide for GroupChat tool calling by @rjambrecic in #1262
• fix: fix script execution check for direct run by @annwag in #1261
• Mongodb query engine by @AgentGenie in #1263
• fix function generate_api_references.py by @futreall in #1239
• Fix formating issue by @davorrunje in #1267
• Fix issubclass checking error in dependency_injection by @davorrunje in #1266
• DocAgent: Internal task manager agent utilise OnContextConditions by @marklysze in #1253
• chore: fix default value for max_round in group_chat_config section by @hexcow in #1270
• fix issues by @sky-coderay in #1272
• Fix failing type checks by @kumaranvpl in #1279
• Docs Polishing by @harishmohanraj in #1281
• [Docs] Add realtime tag in metadata by @harishmohanraj in #1283
• Minor documentation updates by @rjambrecic in #1294
• [Docs] Fix broken Mintlify CI workflow by @harishmohanraj in #1297
• Upgrade google genai version by @sternakt in #1298
• Documentation small fixes by @sternakt in #1304
• Run captain agent tests with agent builder by @kumaranvpl in #1303
• Update Swarm Enhanced Demo notebook by @marklysze in #1309
• Fix swarm context variable function injection by @marklysze in #1264
• Version bump to 0.8.1 by @marklysze in #1308

Full Changelog: v0.8.0...v0.8.1

v0.8.0

Highlights for 0.8

❕ Breaking Change

The openai package is no longer installed by default.

• Install AG2 with the appropriate extra to use your preferred LLMs, e.g. pip install ag2[openai] for OpenAI or pip install ag2[gemini] for Google's Gemini.
• See our Model Providers documentation for details on installing AG2 with different model providers.

0.7.6 to 0.8 Highlights

• 📔 DocAgent
  • Reliability improvements, better URL handling, internal prompt refinements, check out the video
  • In-memory Query Engine, no vector store required, load web pages and documents directly into your system message using an LLM to query
  • Performance analysis, can it do what you need? If not, let us know!
• 🔍 ReasoningAgent now has an internal prompt rewriter for better performance!
• 🐝 Swarm - Addition of OnContextCondition and ContextExpression for context variable-based handoffs that don't require an LLM, creating faster and more predictable flows
• Lots of documentation improvements, including Reference Tools section

0.7 to 0.8 Highlights

🧠 Agents:

• We welcomed our Reference Agents - DocAgent, DeepResearchAgent, DiscordAgent, SlackAgent, TelegramAgent, and WebSurferAgent
• RealtimeAgent was refactored to support both OpenAI and Google Gemini
• Improvements to ReasoningAgent and CaptainAgent
• New run method for chatting directly with an agent

💭 LLM Model Providers:

• Streamlined packages making all of them optional
• Structured Output support for OpenAI, Google Gemini, Anthropic, and Ollama
• Support for new Google and Cohere libraries
• Support for OpenAI's o1 models

🐝 Swarm:

• More robust workflows with context-based handoffs using OnContextCondition and ContextExpression
• More transfer options with AfterWorkOption support in SwarmResults
• Introduction of a Swarm Manager for organic LLM-based transitions

General:

• 🎈 Significantly lighter default installation package
• 🔒 Better data security with the addition of Dependency Injection
• 💬 Streaming of workflow messages with Structured Messages
• 📚 Documentation overhaul - restructured and rewritten
• 🔧 Lots of behind-the-scenes testing improvements, improving robustness

♥️ Thanks to all the contributors and collaborators that helped make release 0.8!

New Contributors

• @SITADRITA1 made their first contribution in #1180
• @Danyylka made their first contribution in #1183
• @pendrue made their first contribution in #1161

What's Changed

• [Docs] Fix broken links by @harishmohanraj in #1170
• In Memory query engine for DocAgent by @marklysze in #1137
• Update docs with openai optional dependency by @rjambrecic in #1176
• Update docs with openai optional dependency by @rjambrecic in #1177
• Add experimental tools user guides by @rjambrecic in #1175
• Fix Terminology and Grammar in Documentation by @SITADRITA1 in #1180
• spelling errors by @Danyylka in #1183
• [@giorgossideris] Add prompt rewriting for multiple messages in ReasoningAgent by @BabyCNM in #1140
• [Docs] Fix broken links by @harishmohanraj in #1191
• [Docs] MkDocs refactoring by @harishmohanraj in #1189
• Move Reasoning agent to the "experimental" folder. by @BabyCNM in #1187
• Update RetrieveChat notebook example and minor internal naming updates by @marklysze in #1186
• Version 0.8.0b1 Version bump, documentation, and warning by @marklysze in #1182
• Fix rag query engine protocol by @AgentGenie in #1197
• spelling issue by @pendrue in #1161
• Tools user guides restructuring by @rjambrecic in #1194
• Use -U option in notebook and blog posts pip install commands by @kumaranvpl in #1207
• Remove whitespace from the agent name by @rjambrecic in #1208
• Use kwargs to create agent_configs in AgentBuilder by @kumaranvpl in #1209
• Tests fixes due to optional openai dependancy by @davorrunje in #1178
• Update tools user guides by @rjambrecic in #1217
• Chromadb query engine by @Eric-Shang in #1195
• Install openai optional dependency to fix tests being skipped in CI by @kumaranvpl in #1211
• Update dependency versions by @kumaranvpl in #1230
• Add deep research agent user guide by @rjambrecic in #1233
• Fix gemini failing and skipped tests by @sternakt in #1229
• DocAgent: RAGQueryEngine protocol alignment, URL analysis, better format handling, performance report, internal agent updates by @marklysze in #1184
• Convert input_docs to list before passing to SimpleDirectoryReader by @kumaranvpl in #1238
• Fix swarm doc link by @davorrunje in #1240
• Use pydantic's model_dump_json instead of json.dumps when available by @kumaranvpl in #1242
• [Swarm] ContextExpression and OnContextCondition by @marklysze in #1221
• Add realtime agent user guides by @rjambrecic in #1245
• Downgrade google-genai to working version by @sternakt in #1247
• ContextExpression support for len for collections by @marklysze in #1251
• Fix swarm test requiring OpenAI package import by @marklysze in #1254
• Bump version to 0.8.0 by @marklysze in #1252

Full Changelog: v0.7.6...v0.8.0

v0.8.0b1

❕ Breaking Change

The openai package is no longer installed by default.

• Install AG2 with the appropriate extra to use your preferred LLMs, e.g. pip install ag2[openai] for OpenAI or pip install ag2[gemini] for Google's Gemini.
• See our Model Providers documentation for details on installing AG2 with different model providers.

New Contributors

• @SITADRITA1 made their first contribution in #1180
• @Danyylka made their first contribution in #1183

What's Changed

• [Docs] Fix broken links by @harishmohanraj in #1170
• In Memory query engine for DocAgent by @marklysze in #1137
• Update docs with openai optional dependency by @rjambrecic in #1176
• Update docs with openai optional dependency by @rjambrecic in #1177
• Add experimental tools user guides by @rjambrecic in #1175
• Fix Terminology and Grammar in Documentation by @SITADRITA1 in #1180
• spelling errors by @Danyylka in #1183
• [@giorgossideris] Add prompt rewriting for multiple messages in ReasoningAgent by @BabyCNM in #1140
• [Docs] Fix broken links by @harishmohanraj in #1191
• [Docs] MkDocs refactoring by @harishmohanraj in #1189
• Move Reasoning agent to the "experimental" folder. by @BabyCNM in #1187
• Update RetrieveChat notebook example and minor internal naming updates by @marklysze in #1186
• Version 0.8.0b1 Version bump, documentation, and warning by @marklysze in #1182

Full Changelog: v0.7.6...v0.8.0b1

v0.7.6

Note: OpenAI's openai package is no longer installed by default. See release notes for v0.8.0b1 for further information.

Highlights

• 🚀 LLM provider streamlining and updates:
  • OpenAI package now optional (pip install ag2[openai])
  • Cohere updated to support their Chat V2 API
  • Gemini support for system_instruction parameter and async
  • Mistral AI fixes for use with LM Studio
  • Anthropic improved support for tool calling
• 📔 DocAgent - DocumentAgent is now DocAgent and has reliability refinements (with more to come), check out the video
• 🔍 ReasoningAgent is now able to do code execution!
• 📚🔧 Want to build your own agents or tools for AG2? Get under the hood with new documentation that dives deep into AG2:
  • How AG2 Works
  • Creating an agent
  • Creating a tool
• Fixes, fixes, and more fixes!

Thanks to all the contributors on 0.7.6!

New Contributors

• @NeoByteXx made their first contribution in #1072
• @comfsrt made their first contribution in #1116
• @marufaytekin made their first contribution in #1101
• @Fallengirl made their first contribution in #1136
• @liminma made their first contribution in #1090
• @futreall made their first contribution in #1150
• @ghoto made their first contribution in #1155

What's Changed

• Manually notify codecov to make PR comment with coverge percentage by @kumaranvpl in #1063
• Support for loading response format as a pydantic model from configuration files by @sternakt in #1023
• Update PR comments automatically after 30 reports by @kumaranvpl in #1066
• Fix LLM parameter propagation and refactor BrowserUseTool by @rjambrecic in #1033
• Fix typing for RAG and Graph RAG by @davorrunje in #1065
• Refactor query engine interface by @AgentGenie in #1058
• Remove after_n_builds param by @kumaranvpl in #1071
• Crawl4ai tool refactoring by @rjambrecic in #1070
• Grammar and Clarity Fixes by @NeoByteXx in #1072
• Remove accidentally committed RFC by @kumaranvpl in #1079
• [Docs] Add google analytics to the gh-pages domain by @harishmohanraj in #1082
• Require permit-tests environment approval for non llm tests by @kumaranvpl in #1084
• Reduce LFS usage in CI by @davorrunje in #1081
• Update a_generate_reply function call with correct params in a_receive by @kumaranvpl in #1076
• Fix Mistral running in LM Studio by @rjambrecic in #1086
• Remove merge_group check from unnecessary workflows by @kumaranvpl in #1085
• [Docs] Fix broken links by @harishmohanraj in #1046
• Preserve chat history when using 'agent.run' by @rjambrecic in #1087
• [Docs] Fix broken image path by @harishmohanraj in #1089
• DocAgent rename and demonstration documentation by @marklysze in #1059
• Further improvements to DocAgent and demo by @marklysze in #1093
• Correct hook processing order on ConversableAgent.a_generate_reply by @marklysze in #1096
• Disable workflow runs for all branches except on main branch on schedule by @kumaranvpl in #1104
• Enhance typing and docs in oai client by @CAROLZXYZXY in #1106
• Fix failing core llm tests because of missing websockets dependency by @kumaranvpl in #1100
• [Docs] Generate initial mkdocs site with static pages by @harishmohanraj in #1108
• Pin browser-use version to specific version by @kumaranvpl in #1109
• [Docs] Fix favicon and mkdocs cleanup by @harishmohanraj in #1112
• Add interoperability user guides by @rjambrecic in #1113
• Update dependency versions by @kumaranvpl in #1111
• Fix typos and improve documentation formatting by @kilavvy in #1117
• [CI/CD] Only check tracked files in git pre-submit by @CAROLZXYZXY in #1107
• Filter out transfer messages in swarm by @yiranwu0 in #1119
• fix minor error transforms.py by @comfsrt in #1116
• Update Cohere client library to V2 with Structured Outputs by @marklysze in #370
• Fix serialization of TextMessage containing PIL image by @rjambrecic in #1126
• Add jsonl files and use it to initiate falkordb by @kumaranvpl in #1125
• Remove permission requirement for non llm tests by @kumaranvpl in #1128
• Fix path comparison for Windows compatibility by @harishmohanraj in #1130
• Fix check_termination_and_human_reply() when called with empty list as messages by @davorrunje in #1132
• Move PIL to optional import block by @kumaranvpl in #1134
• [Bug Fix] Tool signature error for Anthropic #1091 by @marufaytekin in #1101
• fix function errors by @Fallengirl in #1136
• Reasoning Agent can execute code now. by @BabyCNM in #545
• Swarm and Google Gemini client test fixes by @marklysze in #1139
• Fix agents and agents tests checks by @marklysze in #1141
• Pass system message as system_instruction to Gemini #1011. by @liminma in #1090
• rename ysaml to yaml by @futreall in #1150
• Voice hooks fixes by @davorinrusevljan in #1154
• Solves issue with SwarmResult.agent AfterWorkOption -> str by @ghoto in #1155
• Remove detect-secrets from pre-commit by @davorrunje in #1158
• Add LiteLLM user guides by @rjambrecic in #1152
• [Infra] Remove openai dependency from base dependencies by @sternakt in #1149
• [Docs] Add API reference section by @harishmohanraj in #1144
• [Docs] Add docs versioning by @harishmohanraj in #1151
• Set default value for GroupChat messages by @rjambrecic in #1148
• Enable tool removal by name by @marklysze in #1143
• Update Discord/Slack/Telegram Agent inits by @marklysze in #1159
• Bump version to 0.7.6b1 by @davorrunje in #1160
• How AG2 Works documentation for agent and tool builders by @marklysze in #1098
• GPT Assistant tests - cater for openai dependency by @marklysze in #1162
• Further test updates to cater for openai dependency by @marklysze in #1163
• Further changes to accommodate openai as dependency (P3) by @marklysze in #1164
• OpenAI exception handling and httpx dependency by @marklysze in #1165
• Deepseek tests update for openai dependency by @marklysze in #1166
• Update to OnCondition, AfterWork, and UpdateSystemMessage in documentation by @marklysze in #1169
• Bump version to 0.7.6 by @marklysze in #1168

Full Changelog: v0.7.5...v0.7.6

v0.7.6b1

Highlights

• 🚀 LLM provider streamlining and updates:
  • OpenAI package now optional (pip install ag2[openai])
  • Cohere updated to support their Chat V2 API
  • Gemini support for system_instruction parameter and async
  • Mistral AI fixes for use with LM Studio
  • Anthropic improved support for tool calling
• 📔 DocAgent - DocumentAgent is now DocAgent and has reliability refinements (with more to come)
• 🔍 ReasoningAgent is now able to do code execution!
• 📚🔧 Want to build your own agents or tools for AG2? Get under the hood with new documentation that dives deep into AG2:
  • How AG2 Works
  • Creating an agent
  • Creating a tool
• Fixes, fixes, and more fixes!

Thanks to all the contributors on 0.7.6!

New Contributors

• @NeoByteXx made their first contribution in #1072
• @comfsrt made their first contribution in #1116
• @marufaytekin made their first contribution in #1101
• @Fallengirl made their first contribution in #1136
• @liminma made their first contribution in #1090
• @futreall made their first contribution in #1150
• @ghoto made their first contribution in #1155

What's Changed

• Manually notify codecov to make PR comment with coverge percentage by @kumaranvpl in #1063
• Support for loading response format as a pydantic model from configuration files by @sternakt in #1023
• Update PR comments automatically after 30 reports by @kumaranvpl in #1066
• Fix LLM parameter propagation and refactor BrowserUseTool by @rjambrecic in #1033
• Fix typing for RAG and Graph RAG by @davorrunje in #1065
• Refactor query engine interface by @AgentGenie in #1058
• Remove after_n_builds param by @kumaranvpl in #1071
• Crawl4ai tool refactoring by @rjambrecic in #1070
• Grammar and Clarity Fixes by @NeoByteXx in #1072
• Remove accidentally committed RFC by @kumaranvpl in #1079
• [Docs] Add google analytics to the gh-pages domain by @harishmohanraj in #1082
• Require permit-tests environment approval for non llm tests by @kumaranvpl in #1084
• Reduce LFS usage in CI by @davorrunje in #1081
• Update a_generate_reply function call with correct params in a_receive by @kumaranvpl in #1076
• Fix Mistral running in LM Studio by @rjambrecic in #1086
• Remove merge_group check from unnecessary workflows by @kumaranvpl in #1085
• [Docs] Fix broken links by @harishmohanraj in #1046
• Preserve chat history when using 'agent.run' by @rjambrecic in #1087
• [Docs] Fix broken image path by @harishmohanraj in #1089
• DocAgent rename and demonstration documentation by @marklysze in #1059
• Further improvements to DocAgent and demo by @marklysze in #1093
• Correct hook processing order on ConversableAgent.a_generate_reply by @marklysze in #1096
• Disable workflow runs for all branches except on main branch on schedule by @kumaranvpl in #1104
• Enhance typing and docs in oai client by @CAROLZXYZXY in #1106
• Fix failing core llm tests because of missing websockets dependency by @kumaranvpl in #1100
• [Docs] Generate initial mkdocs site with static pages by @harishmohanraj in #1108
• Pin browser-use version to specific version by @kumaranvpl in #1109
• [Docs] Fix favicon and mkdocs cleanup by @harishmohanraj in #1112
• Add interoperability user guides by @rjambrecic in #1113
• Update dependency versions by @kumaranvpl in #1111
• Fix typos and improve documentation formatting by @kilavvy in #1117
• [CI/CD] Only check tracked files in git pre-submit by @CAROLZXYZXY in #1107
• Filter out transfer messages in swarm by @yiranwu0 in #1119
• fix minor error transforms.py by @comfsrt in #1116
• Update Cohere client library to V2 with Structured Outputs by @marklysze in #370
• Fix serialization of TextMessage containing PIL image by @rjambrecic in #1126
• Add jsonl files and use it to initiate falkordb by @kumaranvpl in #1125
• Remove permission requirement for non llm tests by @kumaranvpl in #1128
• Fix path comparison for Windows compatibility by @harishmohanraj in #1130
• Fix check_termination_and_human_reply() when called with empty list as messages by @davorrunje in #1132
• Move PIL to optional import block by @kumaranvpl in #1134
• [Bug Fix] Tool signature error for Anthropic #1091 by @marufaytekin in #1101
• fix function errors by @Fallengirl in #1136
• Reasoning Agent can execute code now. by @BabyCNM in #545
• Swarm and Google Gemini client test fixes by @marklysze in #1139
• Fix agents and agents tests checks by @marklysze in #1141
• Pass system message as system_instruction to Gemini #1011. by @liminma in #1090
• rename ysaml to yaml by @futreall in #1150
• Voice hooks fixes by @davorinrusevljan in #1154
• Solves issue with SwarmResult.agent AfterWorkOption -> str by @ghoto in #1155
• Remove detect-secrets from pre-commit by @davorrunje in #1158
• Add LiteLLM user guides by @rjambrecic in #1152
• [Infra] Remove openai dependency from base dependencies by @sternakt in #1149
• [Docs] Add API reference section by @harishmohanraj in #1144
• [Docs] Add docs versioning by @harishmohanraj in #1151
• Set default value for GroupChat messages by @rjambrecic in #1148
• Enable tool removal by name by @marklysze in #1143
• Update Discord/Slack/Telegram Agent inits by @marklysze in #1159
• Bump version to 0.7.6b1 by @davorrunje in #1160
• How AG2 Works documentation for agent and tool builders by @marklysze in #1098
• GPT Assistant tests - cater for openai dependency by @marklysze in #1162
• Further test updates to cater for openai dependency by @marklysze in #1163
• Further changes to accommodate openai as dependency (P3) by @marklysze in #1164
• OpenAI exception handling and httpx dependency by @marklysze in #1165
• Deepseek tests update for openai dependency by @marklysze in #1166

Full Changelog: v0.7.5...v0.7.6b1

v0.7.5

Highlights

• 📔 DocumentAgent - A RAG solution built into an agent!
  • Ingest one or many PDFs, images with OCR, web pages, and more...
  • Query your documents with natural language
  • Introduction to DocumentAgent
  • Notebook
• 🎯 Added support for Couchbase Vector database
  • Notebook
• 🧠 Updated OpenAI and Google GenAI package support
• 📖 Many documentation improvements
• 🛠️ Fixes, fixes and more fixes

♥️ Thanks to all the contributors and collaborators that helped make the release happen!

New Contributors

• @PixelPil0t1 made their first contribution in #921

• @lokesh-couchbase made their first contribution in #174

• @racerole made their first contribution in #1038

• What's Changed

• Add custom field serializer for agent in SwarmResult by @sternakt in #949

• Check integration test merge queue by @kumaranvpl in #954

• Fix gemini client with anyof nullable function parameters by @rjambrecic in #959

• Fix docs workflows by @kumaranvpl in #958

• Remove obsolete deepseek pricing test by @kumaranvpl in #961

• Remove references to the old WebSurferAgent by @rjambrecic in #964

• Update the single agent run examples by @sternakt in #967

• Add DeepResearchAgent blogpost by @rjambrecic in #968

• Update agents_deep_researcher.ipynb by @sonichi in #972

• [Docs] Update docstring formatting by @harishmohanraj in #965

• [Docs] Communication Tools and Agents documentation by @marklysze in #816

• Typographical Corrections in Variable Names by @PixelPil0t1 in #921

• Fix mutable objects as default parameter values in a function by @davorrunje in #946

• Cleanup workflows by @kumaranvpl in #952

• Update google-genai version by @sternakt in #973

• Add Couchbase VectorDB Support for AG2 by @lokesh-couchbase in #174

• Couchbase fixes for documentation and tests by @marklysze in #979

• Bugfix 902 revise readme by @willhama in #913

• [Docs] Add Comms Agents to Reference Agents documentation by @marklysze in #982

• [Docs] Fix warning, incorrect link, and add API links. by @marklysze in #992

• Fix typing by @davorrunje in #986

• Update cache behaviour in function docstring by @kumaranvpl in #994

• Update dependency versions by @kumaranvpl in #1001

• Update Swarm Document by @yiranwu0 in #977

• Add tools property to ConversableAgent by @marklysze in #989

• Set cancel-in-progress to true for all branches by @kumaranvpl in #1009

• [Docs] Fix broken links in home page by @harishmohanraj in #1010

• [Docs] Add links to symbols by @harishmohanraj in #1018

• Remove unwanted character from types by @harishmohanraj in #1020

• Fix Gemini client for function parameters with nested Pydantic schema by @rjambrecic in #1002

• Fix check broken links by @kumaranvpl in #1016

• Add integration-test.sh and refactor CI by @rjambrecic in #1021

• [Docs] Change register_hand_off function calls in swarm demo by @CAROLZXYZXY in #1013

• Replace AutoGen to AG2 in comments and texts by @davorrunje in #1007

• Fix Swarm registration of an agent's tools from their tools property by @marklysze in #990

• [Docs] Improved Conversation Patterns documentation by @marklysze in #1005

• Fix typing on interop submodule by @davorrunje in #1025

• Add codecov configuration and disable codecov comments by @kumaranvpl in #1030

• Notebooks outputs cleaned by @davorrunje in #1032

• [Docs] Update website demo code so that it is runnable by @CAROLZXYZXY in #1029

• Log cache value message only once using lru_cache by @kumaranvpl in #1024

• [Docs] Fix broken link in homepage by @harishmohanraj in #1037

• Cache seed usage log changed from info to debug level by @davorrunje in #1042

• chore: remove redundant word by @racerole in #1038

• [Docs] Fix broken links by @harishmohanraj in #1034

• Document agent phase 1 by @AgentGenie in #978

• Voice hooks in RealtimeAgent by @davorinrusevljan in #1040

• Bump version to 0.7.5 by @marklysze in #1043

• Add web documentation for document agent by @AgentGenie in #1045

Full Changelog: 0.7.4...v0.7.5

0.7.4

Highlights

• 🔍 New DeepResearchAgent was added
• 💬 New Messaging Platform Tools & Agents were added for the following platforms:
  • Blog
  • Notebook
  • DiscordAgent
  • TelegramAgent
  • SlackAgent
• 🌐 Updated WebSurferAgent
  • Added support for alternative LLM-s
  • Added structured output support
• 📞 Update RealTime Agent
  • New JS library for faster client development
• 📖 Revamped documentation
• 🛠️ Fixes, fixes and more fixes

What's Changed

• Update agentchat_assistant_agent_standalone notebook with metadata by @marklysze in #708
• WebSurferAgent and Browser Tool notebook tidy by @marklysze in #709
• docs: integrate new contributor guide by @devin-ai-integration in #711
• 604 write gemini realtime client blogpost by @sternakt in #686
• Fix failing tests by @rjambrecic in #726
• [Docs] Fixed minor formatting issues in docs by @harishmohanraj in #720
• Sanitize pytest output by patching the _pytest._io.TerminalWritter by @sternakt in #725
• Every exception name should end with Error suffix by @kumaranvpl in #722
• [Docs] Getting started and Use Cases by @marklysze in #547
• [Docs] Fixing links, converting old pages by @marklysze in #732
• [Docs] More link updates for new documentation by @marklysze in #733
• [Docs] Installation link fixes for new documentation by @marklysze in #734
• Suppress error code 503 for gemini tests by @kumaranvpl in #703
• Fix tools execution by @rjambrecic in #724
• Fix incorrect BLOG_POST_URL in tests by @rjambrecic in #741
• Add websurfing tools blog post by @rjambrecic in #742
• Add WebSurferAgent blog by @rjambrecic in #743
• Upload coverage report to codecov.io using codecov token by @kumaranvpl in #739
• [Docs] Restored Resuming a Group Chat documentation and more fixed links by @marklysze in #745
• [Docs] Rename Home to Key Features, move Quick Start, Meet our Agents > Overview by @marklysze in #748
• Add missing licence headers by @davorrunje in #744
• Fix typos in agent and interoperability modules by @vipocenka in #723
• [Docs] Correct blog link in Browsing Capabilities blog by @marklysze in #750
• [Docs] Fix popular resources text color in dark mode by @harishmohanraj in #737
• [Docs] Fix vulnerability issue by @harishmohanraj in #761
• Fix codecov, dependabot and other workflows by @kumaranvpl in #736
• Add Agent-E to community gallery by @rjambrecic in #779
• [Docs] Fix broken links by @harishmohanraj in #772
• Remove pydantic V1 by @kumaranvpl in #777
• docs: Fix grammar in image addition instructions by @defiberrys in #752
• [Docs] Clean up console output during documentation build by @harishmohanraj in #759
• Improve WebSurferAgent notebook and blog by @rjambrecic in #773
• Hotfix failing isinstance checks for swarmable agent by @sternakt in #804
• Remove sensitive prints by @sternakt in #799
• [Docs] Update X account url by @harishmohanraj in #807
• Merge SwarmAgent into ConversableAgent by @marklysze in #316
• Add read permission to check broken links workflow by @kumaranvpl in #815
• Update installation instructions by @rjambrecic in #778
• Expose question_message param in register swarm by @sternakt in #820
• [Docs] Set up redirects for old URLs by @harishmohanraj in #825
• Support other models for browser use by @rjambrecic in #801
• In realtime examples notebooks, use javascript code from the js package by @davorinrusevljan in #828
• [Docs] Do not deploy non-navigation files and directories by @harishmohanraj in #829
• Support other models for crawl4ai by @rjambrecic in #830
• Update realtime swarm notebooks to new swarm API by @sternakt in #832
• [Docs] Fix export_module decorator by @harishmohanraj in #833
• [Bug fix]: Fix tool registration from a Tool by @marklysze in #835
• Bump actions-gh-pages action version to v4 by @kumaranvpl in #839
• Bump dev and test dependency versions by @kumaranvpl in #841
• [Docs] Fix the export_module decorator to update the module attribute of a symbol only when building the docs by @harishmohanraj in #842
• Fix failing CI tests in forked repo PRs by @kumaranvpl in #818
• Bump pydantic-ai version by @kumaranvpl in #847
• Remove paths filter and run llm tests on main on schedule by @kumaranvpl in #788
• Add num_ctx param to ollama options dictionary by @mingxzhao in #790
• [Docs] Replace run with initiate_chat for docs until code execution default changed by @marklysze in #856
• update loggers by @lazToum in #810
• Bump mypy version by @davorrunje in #858
• Remove default code execution config from run executor agent by @marklysze in #853
• update captain agent tools by @lazToum in #802
• Update captainagent.py by @lazToum in #800
• Fix stuck build check by @kumaranvpl in #866
• update WebRTC blog, to use js package by @davorinrusevljan in #850
• Bump optional dependencies by @kumaranvpl in #862
• [Docs] Remove old files from website directory by @harishmohanraj in #844
• [Docs] Fix broken links by @harishmohanraj in #870
• [Docs] Fix mdx parsing issue in docs by @harishmohanraj in #872
• Add structured output to browser use tool by @rjambrecic in #874
• fix(docs): Correct typos and minor wording issues in documentation by @NikolaiKryshnev in #871
• Messaging Platform Tools - Discord, Telegram, Slack by @marklysze in #754
• Communication Agents (Discord/Slack/Telegram, using tools) by @marklysze in #787
• docs: enhance README with key features and practical examples by @crStiv in #546
• version bump to v0.7.4 by @qingyun-wu in #882
• Update DeepSeek pricing by @harishmohanraj in #887
• Add a way to manually trigger workflow when needed by @kumaranvpl in #891
• [Docs] Fix broken links by @harishmohanraj in #892
• Update dependency versions by @kumaranvpl in #884
• Run GraphRag tests with openai secrets by @kumaranvpl in #889
• [Docs] Docs build refactoring and add edit urls to pages by @harishmohanraj in #869
• Fix failing neo4j graph rag tests by @kumaranvpl in #896
• Css Polishing by @harishmohanraj in #897
• [Docs] Prevent export module decorator from modifying the module of a symbol for docs generation by @harishmohanraj in #898
• Add notebook to demonstrate small LLM usage by @kellyaa in #642
• Add package-lock.json to LFS by @davorrunje in #901
• Fix codespell by @davorrunje in #900
• Added preview rules to ruff by @davorrunje in #903
• add 074 by @qingyun-wu i...