Changelog

New Features:

• SeltzTools: A new toolkit featuring support for Seltz Search
• add_search_knowledge_instructions: Adds add_search_knowledge_instructions parameter to Agent and Team class to control whether knowledge search instructions are added to the system prompt.`

What's Changed

• fix: knowledge cookbook bugs and improve consistency. Reverted by @Mustafa-Esoofally in #6052
• [feat] Add Seltz tools integration by @WilliamEspegren in #6133
• chore: update search knowledge base context by @ashpreetbedi in #6181
• chore: release 2.4.5 by @ysolanky in #6185

Full Changelog: v2.4.4...v2.4.5

Changelog

New Features:

• Unsplash image search tools: Added UnsplashTools toolkit for searching and retrieving high-quality, royalty-free images from Unsplash API.
• Moonshot model provider: Added moonshot.ai model provider
• external_execution_silent param for tool decorator: Added a new param to make external tool execution silent by not populating the run response content with placeholder strings

Improvements:

• dependencies in Workflows: When using JWTMiddleware with dependencies_claims to inject user data from JWT tokens, the dependencies were correctly available to agent tools when calling the agent directly via /agents/{agent_id}/runs, but not when the same agent ran as part of a workflow via /workflows/{workflow_id}/runs.

Bug Fixes:

• Markdown Chunker - split_on_headings now respects chunk_size parameter.
• LanceDB empty vectors - Added a fix to reduce empty vectors in lancedb due to embedding failures
• SurrealDB - created_at timestamps are now preserved on updates and get_session correctly filters by session type
• Firestore - Fixed a bug causing rename session operation to lead to unintended overwrites

What's Changed

• ci: consolidate Claude Code workflows into single file by @Mustafa-Esoofally in #6152
• ci: use single job with conditional steps for cleaner CI display by @Mustafa-Esoofally in #6153
• Fix broken link by @jss367 in #6154
• [feat] Add Unsplash image search tools by @dvy246 in #6105
• [feat] add moonshot model provider by @Tuyohai in #6142
• fix: Markdownchunker respects chunk size for split_on_headings by @willemcdejongh in #6169
• fix: LanceDB vector field None after Knowledge.insert() by @harshsinha03 in #6149
• fix: pass dependencies to RunContext in Workflows by @kausmeows in #6172
• fix: SurrealDB created_at timestamp preservation and session type filtering by @harshsinha03 in #6176
• fix: add silent option to suppress verbose messages for HITL tools by @rodrigocoliveira in #5842
• [fix]: Store Firestore session data as nested object instead of JSON by @Byunk in #5866
• chore: Release 2.4.4 by @kausmeows in #6171

New Contributors

• @jss367 made their first contribution in #6154
• @dvy246 made their first contribution in #6105
• @Tuyohai made their first contribution in #6142
• @rodrigocoliveira made their first contribution in #5842
• @Byunk made their first contribution in #5866

Full Changelog: v2.4.3...v2.4.4

Changelog

New Features:

• Excel Reader: Added new reader class ExcelReader for better .xls and .xlsx support

Bug Fixes:

• Reasoning: handle unclosed JSON code blocks in reasoning steps
• CustomEvent :
  • Fix incorrect handling of a confirmed tool’s CustomEvent in acontinue_run for HITL
  • Fix added to preserve subclass properties after session reload
• OpenAIEmbedder: Fixed OpenAIEmbedder to pass the dimensions parameter for third-party OpenAI-compatible embedding APIs

What's Changed

• [fix] Update Discord link in README.md by @kyleaton in #6138
• fix: respect explicit stream=False parameter in workflow run methods by @Akaban in #6132
• fix: handle unclosed JSON code blocks in reasoning steps by @Goodnight77 in #5917
• chore: update CLI method names to cli_app by @Ayush0054 in #6144
• fix: yield CustomEvent from tools in continue_run/acontinue_run by @kausmeows in #6146
• feat: add registry into AgentOS demo. by @anuragts in #6147
• fix: Fixed OpenAIEmbedder to pass the dimensions parameter for third-party OpenAI-compatible embedding APIs. by @SyloYamtao in #6143
• fix: preserve CustomEvent subclass properties during deserialization by @harshsinha03 in #6148
• feat: Knowledge Excel Reader by @Mustafa-Esoofally in #6129
• chore: Release 2.4.3 by @kausmeows in #6140

New Contributors

• @Akaban made their first contribution in #6132
• @Goodnight77 made their first contribution in #5917
• @SyloYamtao made their first contribution in #6143

Full Changelog: v2.4.2...v2.4.3

Changelog

New Features:

• Azure Blob Storage: Knowledge now supports private Azure Blob Storage for file loading, similarly to Sharepoint and Github.
• OpenAI Responses API: Add support for the OpenAI Responses API specification for providers that implement it (Ollama v0.13.3+, OpenRouter beta).

Bug Fixes:

• LightRAG search: Fix added to preserve references in search response.
• MilvusDB: Specify output_fields in document lookup queries to fix Milvus Lite error
• Async Database Drivers: Fix table caching in AsyncSQLiteDb, AsyncPostgresDb, AsyncMySQLDb, and FirestoreDb where None values were incorrectly cached, preventing table creation on subsequent calls.

What's Changed

• fix: preserve LightRAG references in search response [SDK-404] by @harshsinha03 in #6119
• fix: Milvus Lite query bug and upsert hybrid search handling [SDK-294] by @harshsinha03 in #6118
• feat: Azure blob storage by @willemcdejongh in #6120
• [feat] add OpenAI Responses API support for Ollama and OpenRouter by @ysolanky in #6061
• fix: Remove deprecated Pydantic Config and use model_config by @benmezger in #6123
• fix: remove hasattr caching in async db drivers preventing table creation by @kausmeows in #6128
• chore: Release 2.4.2 by @kausmeows in #6130

Full Changelog: v2.4.1...v2.4.2

Changelog

New Features:

• N1N Model Provider: Added n1n.ai as an OpenAI-compatible provider.
• Knowledge: Added first-class Excel ingestion (.xlsx/.xls) by routing spreadsheets through the existing CSV reader, parsing workbooks per sheet into separate documents with sheet metadata.
• Github/Sharepoint support: Added functionality in SDK and on API to support files in private github and sharepoint repos to be added to Knowledge.

Improvements:

• Streaming metrics: Add collect_metrics_on_completion flag to only collect metrics from the final chunk.
• Interfaces: Prevent errors on interfaces from ending up to the user facing message.

Bug Fixes:

• CustomEvent: Add tool_call_id to trace events back to their originating tool calls.
• Knowledge metadata: Fixed to properly propogate metdata filters while using knowledge
• Cerebras Model Provider: Fixed a 422 validation error when using Cerebras models with structured outputs (JSON schema format).
• Team task delegation: Fixed a Python closure bug in Team.adelegate_task_to_members() where loop variables were not properly captured in async function definitions, causing all concurrent member agent runs to use the last member's values instead of their respective values.
• Async generator tools: Handle async generator iteration exceptions gracefully by capturing the error on the tool call instead of re-raising, so the agent can continue and the model can decide how to proceed.

What's Changed

• fix: knowledge metadata propogation by @kausmeows in #6078
• fix: handle empty str password at knowledge level by @kausmeows in #6081
• fix: add additionalProperties: false to Cerebras JSON schemas by @deksprime in #6014
• fix: Python Closure Bug in team.py function adelegate_task_to_members by @d8rt8v in #6067
• ci: Run Claude Code automatically on PRs by @Mustafa-Esoofally in #6088
• fix: draft config retrieval and nested team member loading in components by @kausmeows in #6083
• fix: interfaces error handling by @VirusDumb in #6011
• fix: handle async generator exceptions gracefully like sync generators by @uzaxirr in #6108
• feat: add n1n provider by @n1n-api in #6056
• fix: Perplexity streaming token metrics accumulation by @celobusana in #6058
• feat: add tool_call_id into CustomEvent that yielded from tools in agent by @RayneSun in #6098
• chore: update ag infra templates by @ysolanky in #6079
• chore: improve CLAUDE.md documentation by @ysolanky in #6103
• chore: updated windows dev_setup script by @VirusDumb in #6024
• feat: Knowledge remote support by @willemcdejongh in #6106
• [feat] Support Excel files in Knowledge by @liqiongyu in #5862
• chore: Release 2.4.1 by @kausmeows in #6109

New Contributors

• @deksprime made their first contribution in #6014
• @n1n-api made their first contribution in #6056
• @RayneSun made their first contribution in #6098

Full Changelog: v2.4.0...v2.4.1

Changelog

New Features:

• Knowledge Protocol: Introduced KnowledgeProtocol The main Knowledge implementation now satisfies this interface. Any other implementation may be used with Agent/Team. Note - Only the main Agno Knowledge implementation of this protocol currently supports AgentOS Knowledge management.
• Agent Builder: Users can now persist and manage Agent, Team, and Workflow configurations directly in the database. New AgentOS endpoints enable programmatic creation, retrieval, and updates of component definitions.
• New Events: Added new events for Model, compression and memory.
  • ModelRequestStarted and ModelRequestCompleted
  • CompressionStarted and CompressionCompleted
  • Update MemoryUpdateCompleted to contain the memory content.
• Gemini file inputs: Add direct GCS and external URL support for Gemini file inputs
• TavilyTools: Add api_base_url parameter support to TavilyTools for custom hosted usecases.

Improvements:

• Default DB on AgentOS: Add db to the AgentOS class. If populated, it propagates to all
  agents, teams and workflows without db. It is also used as tracing db.
• Deprecated enable_user_memories : Introduced update_memory_on_run for the same behavior.
• Knowledge config endpoint: Improved checking for available readers without needing instantiation of the reader classes. Reduced unnecessary warnings and improved response times.
• Unify dates format on OS API layer: Unify the format of all dates we send via the OS API. Everything is now a UTC datetime
• AgentOS Trailing Slash Handling: Add a TrailingSlashMiddleware that strips trailing slashes from request paths, all endpoints returns identical responses without 307 redirects.
• Pass instructions as is without XML tags:
  • Agent & Team: Instructions are now added directly to the system message without <instructions> tags by default
  • Added add_instruction_tags=True option to restore the previous behavior for users who prefer structured prompts

Bug Fixes:

• LanceDB: Fixed a bug for duplicate embeddings during async insert and upsert operations.
• Qdrant Async Search: Fixed async search methods which were preventing event loop blocking during concurrent async operations.
• CSV Reader: CSVReader and FieldLabeledCSVReader ignored encoding parameter for BytesIO inputs - always used hardcoded "utf-8" instead of self.encoding or "utf-8".
• PDF Reader: Fixed bug where empty string passwords ("") were incorrectly treated as None due to Python's falsy evaluation
• Openrouter with Gemini: Adds support for preserving Gemini's reasoning_details field when using OpenRouter, fixing multi-turn conversation issues.
• MongoDB Async Search: Fixed async_search() to use async_get_embedding() instead of blocking get_embedding() calls, preventing event loop blocking during concurrent async operations.
• Gemini vertex ai file upload: Fixed a bug where on trying to upload files to Gemini via Vertex AI, the API returned a 400 error.
• Skills System Prompt: Fixed an issue where agents attempted to call skill names directly as functions, resulting in "Function not found" errors.

Breaking changes:

• Remove all deprecated fields:
  • session_state, dependencies, user_id etc in tool functions and hooks where RunContext has replaced it
  • stream_intermediate_steps → stream_events
  • yield_run_response → yield_run_output
  • delegate_task_to_all_members removed from Team class
• Add WebSearchTools: Replacing DDG websearch tool with a Generic web search tool which will be the default on for any websearch related cookbooks & docs.
• Knowledge add_content() renamed to insert(): The add_content() method and its variants have been renamed to insert() insert_many Old method names are still supported, but all examples and docs will soon only reference the new methods.
• AgentOS tracing_db : tracing_db has been deprecated in favor of db parameter on the AgentOS.

What's Changed

• fix: gemini vertex ai file upload error by @Mustafa-Esoofally in #5857
• chore: Demo script for windows by @VirusDumb in #5965
• [fix] Add missing attributes to RemoteAgent for Team compatibility by @AhmedA-afk in #5987
• fix: use async embedder in async search methods for qdrant by @uzaxirr in #6002
• fix: use async embedder in MongoDB async_search to prevent event loop blocking by @uzaxirr in #6003
• fix: prevent duplicate embedding in LanceDB insert() by @pengzy-cosmo in #5919
• feat: Add direct GCS and external URL support for Gemini file inputs by @Mustafa-Esoofally in #5997
• [feat] Add api_base_url parameter support to TavilyTools by @ormeilu in #5962
• feat(cookbook): Revamp 01_demo with new impressive agents by @ashpreetbedi in #6010
• fix: add TrailingSlashMiddleware to normalize URL paths by @uzaxirr in #5985
• fix: Allow empty string passwords for PDF decryption by @Mustafa-Esoofally in #6053
• fix: clarify that skill names are not callable functions by @ysolanky in #6062
• feat: v2.4.0 by @kausmeows in #5818
• fix: Update failing integration test prompt by @willemcdejongh in #6074
• feat: merge v2.4.0 with cookbook refactoring by @ashpreetbedi in #6042
• fix: preserve Router client across copies for background tasks by @ysolanky in #6064
• fix: correct openai-agents install instruction by @harshsinha03 in #6076
• chore: Release 2.4.0 by @kausmeows in #6066

New Contributors

• @AhmedA-afk made their first contribution in #5987
• @pengzy-cosmo made their first contribution in #5919
• @ormeilu made their first contribution in #5962

Full Changelog: v2.3.26...v2.4.0

Changelog

Improvements:

• Per-Request Isolation: Improved request-level isolation for agents/teams/workflows in shared FastAPI processes.

What's Changed

• feat: isolation by @ashpreetbedi in #5990
• chore: v2.3.26 by @kausmeows in #5993

Full Changelog: v2.3.25...v2.3.26

Changelog

New Features:

• LearningMachine: A unified learning system that enables agents to learn from every interaction. It coordinates multiple learning types, each with its own storage backend and retrieval pattern.
• Code Chunker: A new chunker that leverages ASTs to split code into contextually relevant segments

Improvements:

• BrowserbaseTools: Fixed get_page_content() tool failing on large pages by parsing HTML to extract only visible text content (strips scripts, styles, comments)
• Models: Skip retries for non-retryable errors from LLM providers (auth failures, context limits)
• AgentOS: Fixed inconsistent URL handling where trailing slashes caused 404 errors or unnecessary redirects.

Bug Fixes:

• GoogleBigQueryTools: Fixed SQL statement parsing/cleanup when queries include comments.
• DocumentChunking: Fixed to properly respect chunk_size Parameter

What's Changed

• fix: GoogleBigQueryTools SQL statements parsing by @manuhortet in #5957
• chore: new demo env script by @ashpreetbedi in #5961
• Add projects folder to gitignore for local symlink setup by @ashpreetbedi in #5964
• feat: Add Claude Code GitHub Workflow by @Mustafa-Esoofally in #5976
• fix: parse html content from browserbase tools by @kausmeows in #5967
• Fix: DocumentChunking Now Properly Respects chunk_size Parameter by @bhatt-neel-dev in #5916
• fix: add retry logic to skip non-retryable model provider errors. by @uzaxirr in #5984
• feat: Learning Machine by @ashpreetbedi in #5897
• feat: Add CodeChunking strategy for AST-based code chunking [SDK-418] by @harshsinha03 in #5981
• chore: v2.3.25 by @kausmeows in #5986

Full Changelog: v2.3.24...v2.3.25

Changelog

Improvements:

• MongoDB connection handshake now includes Agno version metadata. This enables better connection identification and log analysis for users hosting MongoDB clusters with multiple applications.
• Crawl4aiTools: Add the proxy_config parameter, allowing further configuration of the Toolkit.
• Forbid tools to operate out of base directory: Update PythonTools and MLXTranscribeTools to automatically disallow operating outside of the contextual base directory. Add the restrict_to_base_dir to opt out of this feature and allow operations outside of the base directory.
• MarkdownChunker: Now support split_on_headings parameter to finely control how chunks are separated.
• Knowledge ContentsDB requirement: Improved warning and log messages when contentsDB is not used with Knowledge . Improved filter handling for agentic search.

Bug Fixes:

• Team session names: Fix an issue when using get_session_name directly from Team instances.
• HITL consecutive user input: Fix an issue on certain HITL flows when getting user input in multiple consecutive rounds.
• AgentOS Knowledge router: Fix an issue where AsyncDB was not correctly routed to for content update operations

What's Changed

• fix: generating session name for Team by @kepler in #5933
• Fix: active_requirements not populated for subsequent HITL round by @mishramonalisha76 in #5926
• [chore] Include Agno version details in MongoDB handshake by @NoahStapp in #5922
• fix: introduce proxy_config to the Crawl4aiTools toolkit by @aayush598 in #5859
• fix: replace eval() with type mapping [SDK-417] by @harshsinha03 in #5912
• feat: add restrict_to_base_dir param to PythonTools and MLXTranscribeTools by @manuhortet in #5940
• fix: Knowledge apatch content support by @willemcdejongh in #5942
• fix: AttributeError in get_session_name when run_input is a string by @pbassut in #5869
• chore: Knowledge logging and filtering logic updates by @willemcdejongh in #5927
• chore: v2.3.24 by @manuhortet in #5949

New Contributors

• @NoahStapp made their first contribution in #5922
• @pbassut made their first contribution in #5869

Full Changelog: v2.3.23...v2.3.24

Changelog

Improvements:

• Async Tool Support: Toolkit now supports async tool functions, which are automatically selected when the agent runs in an async context.

Bug Fixes:

• header_provider : Fix for check on MCPTools.
• Team Run Cancellation: Fixed issue on async team runs where cancellation was not handled correctly.

What's Changed

• fix: handle foreign key ref for custom traces table name by @kausmeows in #5888
• [fix] Add response_usage propagation to model response by @tyook in #5872
• Fix: team run cancellation by @dirkbrnd in #5932
• chore: async tool usage by @kausmeows in #5889
• chore: Release 2.3.23 by @dirkbrnd in #5934

New Contributors

• @tyook made their first contribution in #5872

Full Changelog: v2.3.22...v2.3.23