v2.0.1

What's Changed

• fix: knowledge cookbooks by @willemcdejongh in #4491
• fix: agno infra by @ysolanky in #4492
• feat: remove MCP dependency from AgentOS by @manuhortet in #4493
• chore: better check for MCPTools in OS app by @manuhortet in #4497
• chore: Update requirements by @Mustafa-Esoofally in #4496
• chore: Release 2.0.1 by @dirkbrnd in #4494

Full Changelog: v2.0.0...v2.0.1

v2.0.0

What's Changed

See the full changelog

Github Changelog: v1.8.2...v2.0.0

v1.8.2

Changelog

Improvements:

• Structured Output for AG-UI: Added response_model support for AG-UI.
• Structured Output for Discord: Added response_model support for the DiscordClient.

Bug Fixes

• Resolved double thinking chunks in Claude
• Various JSON Schema fixes

What's Changed

• fix: workflow session state error by @kausmeows in #4357
• fix: Double thinking chunks by @dirkbrnd in #4360
• fix: handle non-numeric values in usage.prompt_tokens_details by @simpx in #4359
• [chore] firecrawl dependency upgrade by @harshithvh in #4369
• fix: use async context in arun by @harshithvh in #4386
• docs: add url to azure ai foundry portal by @kimtth in #4380
• feat: pass tracer provider as argument in openinference cookbooks by @georgeneokq in #4404
• [cookbook] Multi Agent Teams using Local Agentic RAG by @lucifertrj in #4238
• feat: Add updated_at to AgentSessionResponse, TeamSessionResponse by @Zulko in #3848
• feat: Support extended JSON-Schema constraints for Gemini by @reikubonaga in #3718
• feat: Add example to build Agentic RAG without APIs by @lucifertrj in #2397
• feat: fix linear create_issue graphql mutation set projectId and assigneeId optional by @danpilch in #3688
• fix: Preserve complete Pydantic schema structure in Claude tool formatting by @aviscode in #4400
• feat: Handle structured outputs for AGUI and Discord apps by @Mustafa-Esoofally in #4413
• [fix] Preserve complete Pydantic schema structure in AWS Bedrock tool formatting by @aviscode in #4457
• chore: Release 1.8.2 by @dirkbrnd in #4475

New Contributors

• @simpx made their first contribution in #4359
• @kimtth made their first contribution in #4380
• @georgeneokq made their first contribution in #4404
• @Zulko made their first contribution in #3848
• @reikubonaga made their first contribution in #3718
• @aviscode made their first contribution in #4400

Full Changelog: v1.8.1...v1.8.2

v1.8.1

Changelog

New Features:

• Reasoning summaries for OpenAI: reasoning summaries are now available with the OpenAIResponses class
• Neo4j Tools: Added Neo4jTools to explore and manipulate graphs in Neo4j.

Improvements:

• Update E2B tool to use Sandbox.create: Update the E2BTools class to use Sandbox.create() from the latest e2b_code_interpreter SDK, replacing the deprecated Sandbox() initialization. No other functionality changed.
• Teams Timezone Identifier: Added timezone_identifier to Team to match Agent functionality.

Bug Fixes:

• Media artifacts validation error: from_artifact() of Image was missing content in its return along with url, so if the generated image is not a url but a encoded content this would fail, or vice-versa.
• Wrong metadata and filters: Fixed issues where filters was not correctly applied on all documents when inserting into the knowledge base.

What's Changed

• [fix] SSL verification error and Space/Page Recognition Issues by @harshithvh in #4318
• Fix: issue 4313 by @nebulaanish in #4314
• [fix] Update DashScope thinking parameters to keep thinking false by … by @Mustafa-Esoofally in #4308
• feat: support reasoning summaries in OpenAIResponses by @AIDAN1025 in #4326
• Fix NoneType AttributeError in Gemini Vertex AI Search citations by @10Ala10 in #4336
• [feat] allow OpenAIChat to accept both httpx.Client and httpx.AsyncClient by @chiruu12 in #4330
• fix: media artifact validation error by @kausmeows in #4342
• [fix] Knowledge: fix wrong metadata and filters by @guyrosin in #4284
• [refactor] Update E2B tool to use Sandbox.create by @pranzalkhadka in #4346
• feat: Add connection cleanup to Qdrant implementation by @harshithvh in #4340
• Feat add neo4j agent by @swayamyadav05 in #3889
• [fix] Add missing time zone feature to Team and update default API for get_location by @Riobener in #4325
• feat: Release 1.8.1 by @kausmeows in #4354

New Contributors

• @AIDAN1025 made their first contribution in #4326
• @10Ala10 made their first contribution in #4336
• @swayamyadav05 made their first contribution in #3889
• @Riobener made their first contribution in #4325

Full Changelog: v1.8.0...v1.8.1

v1.8.0

Changelog

New Features:

• Memori ToolKit: Added MemoriTools, a ToolKit for Agents and Teams to use GibsonAI’s Memori.

Improvements:

• Workflows File Input: Now files=[] can be provided to Workflow.run or Workflow.arun.
• Scrapegraph Tools Agentic Crawling: Added agentic_crawler to ScrapeGraphTools.
• VertexAI Search: Added Vertex AI Search to the Gemini model.

Updates:

• DuckDuckGoTools: Updated DuckDuckGoTools to work with the ddgs package. This requires users to install ddgs instead of duckduckgo-search.

What's Changed

• feat: add memori tool for agno by @Boburmirzo in #4292
• feat: Update scrapegraph.py by @VinciGit00 in #4281
• [feat] Add Vertex AI Search support and Gemini search cookbooks by @harshithvh in #4294
• [fix] PgVector upsert respects Document.id and unifies record building by @joeportela in #4269
• update DuckDuckGo integration to support latest ddgs API by @dgamee in #4287
• Fix: FastAPI Binary Image output by @VirusDumb in #4311
• [fix] GitHub tool search_code pagination IndexError with empty results by @harshsinha03 in #4243
• feat: add file input to workflows by @kausmeows in #4317
• feat: Release 1.8.0 by @dirkbrnd in #4316

New Contributors

• @joeportela made their first contribution in #4269
• @dgamee made their first contribution in #4287
• @harshsinha03 made their first contribution in #4243

Full Changelog: v1.7.12...v1.8.0

v1.7.12

Changelog

Improvements:

• Teams Collaborate Streaming: Added support for streaming or teams on collaborate mode.
• OpenAI Verbosity: Added verbosity parameter to OpenAIChat and OpenAIResponses.
• Gemini URL Context Tool: Adds support fior Gemini URL context tool.

Bug Fixes:

• GCSKnowledgeBase Load: Fixed cases where .load() functioned incorrectly.
• Youtube Transcripts: Removed deprecated implementation in YoutubeTools.

What's Changed

• [feat] Allow a user to specify DynamoDB provisioned throughput when initializing DynamoDbStorage class. by @oakhan3 in #4112
• [feat] Add Vertex AI support for GeminiEmbedder by @smilewilson1999 in #4223
• feat: Team streaming events on Collaborate mode by @dirkbrnd in #4220
• Fix: Preserve context, session_state, and extra_data when creating default reasoning agents by @hussain-mohammed in #4081
• fix(discord): Ensure Discord ID integrity by using string type by @MatheusCally in #4240
• [fix] Make Gemini aware of pre-tool messages by @guyrosin in #4252
• fix: GCS KnowledgeBase; Brave Search Tool by @willemcdejongh in #4260
• feat: Gemini URL context tool by @Mustafa-Esoofally in #4270
• feat: New OpenAI params by @Mustafa-Esoofally in #4265
• fix: transcript function from YouTube API by @harshithvh in #4280
• chore: Release 1.7.12 by @dirkbrnd in #4274

New Contributors

• @smilewilson1999 made their first contribution in #4223
• @hussain-mohammed made their first contribution in #4081
• @MatheusCally made their first contribution in #4240

Full Changelog: v1.7.11...v1.7.12

v1.7.11

Changelog

New Features:

• In Memory Storage: Added InMemoryStorage for a simpler storage option useful in rapid prototyping.
• Trafilatura Tools: Added TrafilaturaTools SDK for web scraping and text extraction using the Trafilatura library, with 15 comprehensive cookbook examples.
• Qwen and Dashscope Native Models: Use Qwen models via DashScope integration.
• Brandfetch Tools: Added tools allowing agents to fetch brand information and assets via Brandfetch.

Improvements:

• Async Support for Milvus Hybrid Search
• Bedrock Files Support: Now supporting File for compatible AWS Bedrock models.

Bug Fixes:

• MarkdownKnowledgeBase Custom Chunking Strategy: Fixed issues with setting a custom chunking strategy.
• Teams Routing/Coordination System Message: Added criteria for when the team must respond directly or forward requests to its members.

What's Changed

• fix: Allow setting of customer markdown chunking strategy by @dirkbrnd in #4199
• fix: async fastapi team media processing by @VirusDumb in #4189
• fix: add 'packaging' dependency and regenerate requirements.txt by @jramnai in #4153
• fix: lack of workflow initialization causes storage.mode to not be se… by @dreamcoin1998 in #4147
• [fix] Document ids in Markdown and Text readers by @5739n4 in #4143
• [fix] Using clickhouse as a vector database will result in an error by @Forget-C in #4150
• [feat] Add async hybrid search support for Milvus integration by @nhathoang0110 in #4163
• feat: Add Trafilatura to Web Scraping Tools by @harshithvh in #4188
• fix: Teams route/coordinate system message by @harshithvh in #4194
• feat: Add workers parameter to fastapi app by @VirusDumb in #4202
• fix: Resolve citations deserialization issue in RunResponse and TeamRunResponse models by @0xRichardH in #3186
• [feat] Create a flexible in-memory storage option that can be optionally used with any custom persistence solution (AWS S3, Snowflake, etc.). by @oakhan3 in #4115
• Feat: Normalize tool call identifiers by @RockPaperSmittie in #4205
• [feat] Add support for bedrock file input by @dcumelles in #3673
• feat: Dashscope models by @Mustafa-Esoofally in #4208
• fix: update run_messages in Team.arun() by @seandi in #4216
• chore: dependency update for: cookbook/examples/streamlit_apps/sql_agent by @thierryturpin in #4200
• Toolkit: Add new brandfetch toolkit with async and sync support by @adiberk in #4141
• chore: Release 1.7.11 by @dirkbrnd in #4217

New Contributors

• @jramnai made their first contribution in #4153
• @Forget-C made their first contribution in #4150
• @oakhan3 made their first contribution in #4115
• @dcumelles made their first contribution in #3673
• @thierryturpin made their first contribution in #4200
• @adiberk made their first contribution in #4141

Full Changelog: v1.7.10...v1.7.11

v1.7.10

Changelog

New Features:

• GPT-5 Compatibility: Support for GPT-5 with our OpenAIResponses class

Improvements:

• Knowledge Base PDF with Password Support: Added support for password-protected PDFs on PDFKnowledgeBase.
• Pagination for GitHub Tools: Added pagination with metadata for GitHub Tools.
• Team Role Parameter: Added a role parameter to the Team class for defining the team's purpose and specialization.

Bug fixes:

• Support for all GPT reasoning models with our OpenAIResponses class

What's Changed

• feat: KB - PDF with password by @Mustafa-Esoofally in #4106
• [fix] Implement weaviate knowledge upsert. Improve AgentKnowledge class. by @Siete-F in #3784
• fix: handle receiving a list of Message as input for team arun by @manuhortet in #4167
• cookbook: update mcp library for gibsonai tool by @Boburmirzo in #4173
• feat: pagination with metadata for github issues by @harshithvh in #4166
• [feat] Add role parameter to Team class by @harshithvh in #4172
• feat: support new responses API for GPT-5 and all other GPT reasoning models by @RockPaperSmittie in #4186
• Fix session states merge precedence by @abiari in #4182
• chore: Release 1.7.10 by @manuhortet in #4187

New Contributors

• @abiari made their first contribution in #4182

Full Changelog: v1.7.9...v1.7.10

v1.7.9

Changelog

Improvements:

• Reranker in Hybrid Search: Added support for Reranker in Pgvector hybrid search
• Output Model: Added support for stream_intermediate_steps with output model
• PDF page split: Refactor of PDF Readers and addition of Page Number Handling

Bug Fixes:

• LanceDb Remote Connection: Fixed a bug preventing connection to a remote LanceDb connection.
• Messages: Fixed the order when using both message and messages.

What's Changed

• [feat] Add batch file for running tests on windows by @harshithvh in #4126
• [feat] Process PDF page numbers and add (dont) split pages flag, improving chunking. by @Siete-F in #3774
• feat: fix messaging order for conversation by @Mustafa-Esoofally in #4109
• adding docstring to AWSLambdaTools by @apuertolara in #4130
• fix: add reranker in pgvector hybrid search by @kausmeows in #4139
• fix: output model events by @ysolanky in #4133
• fix: remote lancedb by @ysolanky in #4145
• chore: Release 1.7.9 by @ysolanky in #4144

New Contributors

• @harshithvh made their first contribution in #4126
• @apuertolara made their first contribution in #4130

Full Changelog: v1.7.8...v1.7.9

v1.7.8

Changelog

New Features:

• Output model: Added support for output_model on Agents and Teams to generate the final response from output_model instead of model

Improvements:

• Support for OpenAI Flex Processing: Added service_tier to OpenAIChat and OpenAIResponses.

Bug Fixes:

• Print Response:
  • Fixed show_member_responses not working correctly on Team
  • Fixed printing of MCP responses on streamable HTTP
• Session State on Team: Fixed precedence for session state from sessions DB.
• YouTubeTranscriptApi has no attribute get_transcript :
  • The YoutubeTranscriptApi is updated and now uses .fetch(video_id) for getting transcripts.
• HITL streaming:
  • Added the required attributes- tools_requiring_confirmation, tools_requiring_user_input, tools_awaiting_external_execution on the class BaseAgentRunResponseEvent
  • If you are using streaming, it is recommended to pass the run_id and a list of updated_tools to the continue_run or acontinue_run method.

What's Changed

• Fix: Certain dbs containing null created_at fields by @RockPaperSmittie in #4067
• fix: show tool calls with no args on print and aprint by @manuhortet in #4066
• feat: Adds support for OpenAI Service Tier by @Mustafa-Esoofally in #4065
• fix: incorrect async type checking and pydantic version conflict in function.py by @kausmeows in #4074
• Fix AttributeError in message.py tool_calls handling and confirm parameter validation in decorator.py by @dreamcoin1998 in #4069
• Fibonacci calculations agent by @nile-solutions-info in #3981
• fix: extend command whitelist for MCP server commands by @manuhortet in #4093
• fix: respect show_members_responses flag in streaming mode (#4064) by @ItsRoy69 in #4071
• feat: Allow custom port for the Google Sheets Toolkit Auth by @Mustafa-Esoofally in #4077
• fix: improve tool handling in OpenAIResponses for better robustness by @kepler in #4094
• feat: Enhance Google Calendar toolkit with unified authentication by @Gyani25k in #2426
• fix: youtube get_transcript by @mishramonalisha76 in #4089
• fix: hitl stream breaking by @kausmeows in #4124
• feat: add Gemini thinking to response by @Mustafa-Esoofally in #4063
• feat: output model by @ysolanky in #4105
• chore: Release 1.7.8 by @dirkbrnd in #4092

New Contributors

• @RockPaperSmittie made their first contribution in #4067
• @dreamcoin1998 made their first contribution in #4069
• @nile-solutions-info made their first contribution in #3981
• @Gyani25k made their first contribution in #2426

Full Changelog: v1.7.7...v1.7.8