2025-08-01

v1.7.7

New Features:

Updated MCP Tools: Our ****MCPTools and MultiMCPTools classes can now be initialized and used without an async context manager, providing a much easier experience. See the updated docs.
Morph Tools: Morph’s Fast Apply model as a Tool, it intelligently merges your original code with update snippets at 98% accuracy and 4500+ tokens/second.

Improvements:

LiteLLM File & Image Understanding: Added support for file and image inputs for LiteLLM.
Upgrade ZepTools to v3 Zep: ZepTools are now compatible with Zep v3.

Bug Fixes:

OpenAIEmbedder() wrong dimensions for text-embedding-3-large: Automatically handles default dimensions length for text-embedding-3-small as 1536 and text-embedding-3-large as 3072.

2025-07-24

v1.7.6

New Features:

Portkey Model Support: Added support for Portkey hosted models.
Bitbucket Tool: Added BitbucketTools with a variety of Bitbucket repository actions.
Jina Embedder: Added JinaEmbedder for using embedding models via Jina.
Row Chunking: Added RowChunking as a CSV chunking strategy.
EVM Toolkit: Added EvmTools to do transactions on EVM compatible blockchains using web3.
LinkUp Toolkit: Added LinkupTools for powerful search.
Background Execution Support for Workflows 2.0: Introduced background execution capabilities for Workflows 2.0, enabling non-blocking workflow execution with polling support. See docs here.

Improvements:

Async Bedrock Support: Added async execution support for the AWS bedrock implementation.
PostgreSQL Tools Updates: Various security and stability overhauls made to the PostgresTools toolkit.
Daytona Toolkit Updates: Added new tools for Daytona agent Toolkit

Bug Fixes:

LiteLLM Metrics: Fixed issue with metrics on streaming responses from LiteLLM.
Team Expected Output: Fixed issue where expected_output of members were overwritten by the team leader agent.
Workflows Async Generators: Fixed how async generator arun functions are treated. It now correctly keeps async generators as async generators and doesn’t convert it to a coroutine.
- Before: Workflows with Async generator arun functions were incorrectly awaited as coroutines, which could cause runtime errors or prevent proper iteration through the yielded asynchronous values.
- After: Async generator workflows are now properly recognised and handled as async generators, allowing for correct iteration over their yielded values using async for. This ensures all yielded results are processed as intended within asynchronous workflows.
LiteLLM Multiple Streaming Tool Calls: When Agno is run through LiteLLM against OpenAI chat models (eg. GPT4.1), multiple streamed tool_calls lost their individual argument streams. This has been resolved.

2025-07-17

v1.7.5

New Features:

SurrealDB: Added SurrealDB support as a vector DB for knowledge bases.

Improvements:

Session Caching: Added cache_session attribute to allow users to switch off session caching, which improves on memory management.
Workflows 2.0 FastAPI Support: Added support for running the new workflows in FastAPIApp .

Bug Fixes:

Nested Tool Hooks: Fixed bug with nested tool hooks.

2025-07-16

v1.7.4

Workflows 2.0

New Features:

Workflows Revamped (beta): Our new Workflows implementation (internally referred to as Workflows 2.0) is a complete redesign of the workflow system in Agno, introducing a step-based architecture that provides better structure, flexibility, and control over multi-stage AI agent processes
- Core Concepts:
  - Flexible Execution: Sequential, parallel, conditional, and loop-based execution
  - Smart Routing: Dynamic step selection based on content analysis or user intent
  - Mixed Components: Combine agents, teams, and functions seamlessly
  - State Management: Share data across steps with session state
- For a comprehensive guide to the new Workflows system, check out the docs.
Pydantic Input for Agents/Teams: Both Agent and Team now accepts structured input (i.e. a pydantic model) on run() and print_response().

2025-07-15

v1.7.3

New Features:

Session State on Run: You can now pass session_state when running an agent or team. See docs for more information.
GCS Support for PDF Knowledge Base: Added GCSPDFKnowledgeBase to support PDFs on Google Cloud Storage.

Bug Fixes:

Workflows Async + Storage: Fixed issues where sessions was not correctly stored with async workflow executions.
Session State Management: Fixed issues around session state management in teams and agents. Session state and session name will now correctly reset and load from storage if sessions are switched.
Metadata Support for Document Knowledge Base: Adds metadata support for DocumentKnowledgeBase
Session Metrics with History: Fixed bug with session metrics on Agent where history is enabled.

2025-07-10

v1.7.2

New Features:

MySQL Storage: Added support for MySQLStorage as an agent/team/workflow session storage backend.
XAi Live Search: Added support for live search on the XAi model provider.
OpenAI Deep Research: Support added for o4-mini-deep-research and o3-deep-research models.
Stagehand MCP example: Added an example of using Stagehand MCP with Agno.
Atla observability example: Added an example of using Atla observability with Agno.
Scrapegraph example: Added an example of using Scrapegraph with Agno.

Improvements:

Memory Growth on Performance Evals: Added memory_growth_tracking as an attribute on PerformanceEval to enable additional debug logs for memory growth.
Agent/Team in Tool Hook: Added agent and team as optional parameters in tool hooks, for more flexibility.

Bug Fixes:

Gemini 2.5 Metrics: Fixed Gemini metrics to correctly include “thinking” tokens.
Claude tool calling: Fixed a bug related to parsing function call responses when using Claude models.
Team Metrics: Fixed a bug with team metrics when teams have history enabled.

2025-06-30

v1.7.1

New Features:

Debug Level: Added debug_level to both Agent and Team. This is an int value of either 1 (default) or 2 (more verbose logging). Currently this only enables more verbose model logs, but will be used more widely in future.

Improvements:

Parser Model on Teams: Added parser_model support for Team. See docs here.
Support for Gemini Thinking: Added thinking_budget and include_thoughts parameters for Gemini model class.
Serper Tools: Made updates to the toolkit to include new tools search_news , search_scholar and scrape_webpage.
Valyu Tools: New Valyu toolkit for Deep Search capabilities of academic sources.
Oxylabs: Added OxylabsTools for adding more web-scraping capabilities to agents.

Bug Fixes:

DuckDB CSV parsing error: For CSV files use the custom read_csv method for improved CSV parsing
Full Team Metrics: Fixed an issue with calculation of the full_team_session_metrics on Teams.

2025-06-26

v1.7.0

New Features:

Agent/Team Add Tool: Added convenience function to Agent and Team → add_tool(tool) to append new tools after inititialisation.
Streaming Structured Output: Implemented structured output during streaming. This means streaming won’t be turned off when response_model is passed. The structured output itself is not streamed, but it is part of the iterator response when running an agent/team with streaming. The response model is set on a single RunResponseContentEvent and on the final RunResponseCompletedEvent.

Improvements:

Linear Teams Tool: Added tool to get the list of teams from Linear.

Bug Fixes:

Uppercase Structured Output: Resolved cases where pydantic model fields contain field names with upper case characters.

Breaking Change:

If you use run(..., stream=True) or arun(..., stream=True) on Agent or Team with a response_model set, the current behaviour would switch off streaming and respond with a single RunResponse object. After the changes mentioned above, you will get the Iterator[RunResponseEvent] / AsyncIterator[RunResponseEvent] response instead.

2025-06-23

v1.6.4

New Features:

Brightdata Toolkit: Added multiple web-based tools via Brightdata.
OpenCV Video/Image Toolkit: Added tools for capturing image/video via your webcam.
DiscordClient App: Added a DiscordClient app for connecting your agent or team with Discord in the form of a discord bot.

Improvements:

FileTools File Search: Added search to FileTools.

Bug Fixes:

Fix User Control Flow with History: Fixed issues where user control flow (HITL) flows failed with message history.
Fix lance db upsert method for supporting knowledge filters: Fixed function upsert not using the parameter filters
Update mongo db hybrid search for filters: Mongodb now correctly uses filters for hybrid search
Fixed team.rename_session(...) that raises an AttributeError: Fixed function rename_session not using the parameter session_id

2025-06-18

v1.6.3

New Features:

User Control Flows on Playground: The Agno Platform now support user control flows on the playground.
Team & Agent Events on RunResponse: Added store_events parameter to optionally add all events that happened during the course of an agent/team run on the RunResponse/TeamRunResponse.
Team Member Responses on Playground: The Agno Platform now shows member responses during team runs.
Behind-the-scenes on Playground: The Agno Platform now shows what is happening during runs, for both agents and teams.
Metadata filtering support for csv and csv_url knowledge bases: Add knowledge filters support for these knowledge base types.

Updates

Removed a prefix from async function names: asearch_knowledge_base, etc will now be the same as their sync counterparts when sent to the model. The names of functions are important for accurate function calling.

Bug Fixes:

AG-UI Fix: Fixed issue related to missing messages when using the Agno AG-UI app.
Chat History Fix: Fixed issue related to history not available when agent_id not set.
MongoDB ObjectId serialization issue when using with agent: Fixed issue while ****using mongodb vectordb with ObjectId present in the metadata it throws Object of type ObjectId is not JSON serializable

2025-06-13

v1.6.1

New Features:

Nebius Embeddings: Added support for embedding models on Nebius.
Firestore Memory and Storage: Added support for Firestore both as memory and storage provider for your agents.

Improvements:

Improved Event Payloads: Added agent_name to agent events, and team_name to team events. Also added team_session_id to team-member events to indicate that it belongs to the top-level team session.
Team Run Events: Added stream_member_events to teams to optionally disable streaming of member events.
DocumentKnowledgeBase Async: Added async support on DocumentKnowledgeBase
Enums on Custom Tools: Added support for enum parameters in custom tools.

Bug Fixes:

Team Events: Fixed issues related to team and member events not being part of the same session. Going forward a team and its members will all have the same session ID.

2025-06-10

v1.6.0

Improvements:

New Streaming Events: We have improved our streaming events system. See the details in “breaking changes” section at the bottom.
Member Events in Teams: The above change includes streaming of events from team members with the top-level team events.

Bug Fixes:

Apify Tools: Fixed the ApifyTools initialize to correctly register functions.

Breaking Changes:

Updates to Run Without Streaming:
- RunResponse now does not have an event attribute. It still represents the responses of the entire run.
- An additional attribute RunResponse.status now indicates whether the run response is RUNNING, PAUSED, or CANCELLED.
Updates to Run Streaming:
- In the case of streaming you now get reformulated run events. These events are streamed if you do agent.run(..., stream=True) or agent.arun(..., stream=True) .
- Agents have the following event types:
  - RunResponseContent
  - RunError
  - RunCancelled
  - ToolCallStarted
  - ToolCallCompleted
  - with stream_intermediate_steps=True:
    - RunStarted
    - RunCompleted
    - ReasoningStarted
    - ReasoningStep
    - ReasoningCompleted
    - MemoryUpdateStarted
    - MemoryUpdateCompleted
- See detailed documentation here.
Updates to Teams:
- Teams have the following event types:
  - TeamRunResponseContent
  - TeamRunError
  - TeamRunCancelled
  - TeamToolCallStarted
  - TeamToolCallCompleted
  - with stream_intermediate_steps=True:
    - TeamRunStarted
    - TeamRunCompleted
    - TeamReasoningStarted
    - TeamReasoningStep
    - TeamReasoningCompleted
    - TeamMemoryUpdateStarted
    - TeamMemoryUpdateCompleted
- Teams will also yield events from team members as they are executed.
- See detailed documentation here.
Updates to Workflows:
- You should now yield WorkflowRunResponseStartedEvent and WorkflowRunResponseCompletedEvent events.

2025-06-06

v1.5.10

New Features:

Playground File Upload: We now support file upload via the Agno Playground. This will send PDF, CSV, Docx, etc files directly to the agents/teams for interpretation by the downstream LLMs. If you have a knowledge base attached to the agent/team, it will upload the file to the knowledge base instead.
Async Evals: Support for async and evaluations. See examples here.

Improvements:

Exa Research: Added research tool on ExaTools. See more on their docs about how research works.
Whatsapp Type Indicator: Add type indicator to Whatsapp responses.

Bug Fixes:

State in Messages Fixes: Fixed issues around nested json inside messages and adding state into messages.

2025-06-05

v1.5.9

New Features:

AG-UI App: Expose your Agno Agents and Teams with an AG-UI compatible FastAPI APP.
vLLM Support: Added support for running vLLM models via Agno.
Serper Toolkit: Added SerperTools toolkit to search Google
LangDB Support: Added LangDB AI Gateway support into Agno.
LightRAG server support: Added LightRAG support which provides a fast, graph-based RAG system that enhances document retrieval and knowledge querying capabilities.
Parser Model: Added ability to use an external model to apply a structured output to a model response
Pdf Bytes Knowledge: Introduced a new knowledge base class: PDFBytesKnowledgeBase, which allows the ingestion of in-memory PDF content via bytes or IO streams instead of file paths.
Qdrant Mcp Server: Added MCP support for Qdrant
Daytona integration: Added DaytonaTools toolkit to let Agents execute code remotely on Daytona sandboxes
Expand URL addition in Crawl4ai: Introduced a new URL expansion feature directly into the Crawl4ai toolkit. Our agents frequently encounter shortened URLs that crawl4ai cannot scrape effectively, leading to incomplete data retrieval. With this update, shortened URLs are expanded to their final destinations before being processed by the crawler.
AWS SES Tools: Added AWSSESTools to send emails via AWS SES.
Location Aware Agents: Added add_location_to_instructions to automatically detect the current location where the agent is running and add that to the system message.

Improvements:

FastAPIApp Update: FastAPIApp was updated and agent was replaced with agents, team with teams and workflows was added. This also now requires you to specify which agent/team/workflow to run.
- E.g. http://localhost:8001/runs?agent_id=my-agent
ZepTools Updates: Updated ZepTools to remove deprecated features.
GmailTools Attachments: GmailTools now support attachments.
Improve code reusability by using fetch with retry and async_fetch_with_retry: updated fetch_with_retry and async_fetc_with_retry to be reused inurl_reader.py
Add name to evaluation examples: Included name param in evaluation examples
XTools Search: Added search_posts for XTools.

Bug Fixes:

Claude Prompt Tokens: Fixed a bug with prompt tokens not propagating in Claude model class
Add type in base App class for registry: Can have different types of app like- slack, whatsapp, etc
Accept empty array of pdf urls: Fixed an issue where empty PDF URL arrays were not accepted, preventing knowledge base queries without adding new documents
Anthropic cache metrics propagation: Fixed a bug where Anthropic’s prompt caching metrics were not propagating to Agent responses, despite the raw Anthropic API working correctly. This minimal fix ensures cache performance metrics are properly captured and reported.
Handle non serializable objects on RunResponse dict parsing:
- Updated RunResponse.to_dict() to handle non-serializable fields, as Python enums

2025-06-03

v1.5.8

New Features:

Slack App: Introducing the SlackApp to allow you to create agents for Slack! The app allows agents to respond to individual messages or on group chats, and it creates threads to respond to messages.
Visualization Tools: Added VisualizationTools that uses matplotlib to give agents the ability to make graphs.
Brave Search Tools: Introduced a new toolkit for integrating BraveSearch that allows agent to search the web using brave search api.

Improvements:

Pass filters for traditional RAG: Properly pass down knowledge_filters even if self.add_references=True which is a case for traditional RAG (diff from Agentic RAG)
Add infer param to Mem0: Added infer as a param to Mem0Tools

Bug Fixes:

Searxng tool initialization: Fixed Searxng tool initialization error
```
AttributeError: 'Searxng' object has no attribute 'include_tools'
```
and added comprehensive unit tests.
Fix for enum as a response model for Gemini: With 1.5.6, a bug was introduced now allowing enum as a data type for Gemini response model.
OpenAI parsing structured output: With the changes in the OpenAI library we don’t need to parse the structured output separately.
Fix accuracy evals monitoring: Added logic to handle monitoring when evaluating Teams in the run function of AccuracyEval

Updates

Updates to Apps:
- FastAPIApp does not have a default prefix anymore and /run → /runs (i.e. the created run endpoint is now <your_domain>/runs)
- serve_fastapi_app is now replaced with .serve() on the instance of FastAPIApp.
- serve_whatsapp_app is now replaced with .serve() on the instance of WhatsappAPI.

2025-05-30

v1.5.6

New Features

Team Evals: All types of Evaluations now support Teams!

Improvements:

Async Workflows: Added arun support for Workflows, so they can now be used with async Python.
Parallel Memory Updates: Made speed improvements when user memories and session summaries are generated.
Reimplement tool_call_limit: Revamp of tool_call_limit to make it work across a whole agent run.
Gemini / OpenAI Structure Response: Improved Gemini and OpenAI Structured Response support. Dict types can now be used when defining structured responses.

Bug Fixes:

Mistral Structured Outputs with Tools: Fixed an issue preventing Mistral model use with structured output and tools.
Images In Run Without Prompt: Fixed issues related to images being ignored if there wasn’t a prompt provided on run.
Pgvector Upsert Fix: Fixed Pgvector upsert not copying metadata properly.
Handle AgnoInstrumentor failing with OpenAIResponses: PR merged in Arize’s openinference repo: https://github.com/Arize-ai/openinference/pull/1701.
Pinecone Filters: Enabled filters for pinecone vector db.
Combined KB Async: Add missing async method to Combined KB.
Team Session State Fix: team_session_state is now correctly propagated and shared between all members and sub-teams of a team.
Gemini type fix for integers:
- Pydantic models with Dict[str, int] fields (and other Dict types) were failing when used as response_schema for both OpenAI and Gemini models due to schema format incompatibilities.
Session Name: session_name is now available after a run.
Handle UUIDs while serialization in RedisStorage: Fixed error object of type UUID is not JSON serializable.

Updates:

For managing team_session_state, you now have to set team_session_state on the Team instead of session_state.

2025-05-27

v1.5.5

New Features:

Claude File Upload: We can now upload a file to Anthropic directly and then use it as an input to an agent.
Claude 4 Code Execution Tool: Updated Claude to execute Python code in a secure, sandboxed environment.
**Prompt caching with Anthropic Models: ** Allowed resuming from specific prefixes in your prompts. This approach significantly reduces processing time and costs for repetitive tasks or prompts with consistent elements.
Vercel v0 Model: Added support for new Vercel v0 models and cookbook examples.
Qdrant Hybrid Search support
Markdown Knowledge Base: Added native support for Markdown-based knowledge bases.
AI/ML API platform integration: Introduced integration with AI/ML API, a platform providing AI/ML models. AI/ML API provides 300+ AI models including Deepseek, Gemini, ChatGPT. The models run at enterprise-grade rate limits and uptimes.
Update Pydantic and dataclass in function handling: Added support for Pydantic and dataclass objects as input to a function. See here for an example.

Improvements:

Timeout handling for API calls in ExaTools class:
- Timeout functionality to Exa API calls to prevent indefinite hanging of search operations. The implementation uses Python’s concurrent.futures module to enforce timeouts on all Exa API operations (search, get contents, find similar, and answer generation).
- This change addresses the issue where Exa search functions would hang indefinitely, causing potential service disruptions and resource leaks.
Fetch messages from last N sessions:
- A tool for the agent, something like get_previous_session_messages(number_of_sessions: int) that returns a list of messages that the agent can analyse
- Switch on with search_previous_sessions_history
Redis Expiration: Added expire key to set TTL on Redis keys.
Add Anthropic Cache Write to Agent Session Metrics: Added cache_creation_input_tokens to agent session metrics, to allow for tracking Anthropic cache write statistics

Bug Fixes:

Huggingface Embedder Updates:
- Huggingface has changed some things on their API and they’ve deprecated .post on their InferenceClient()- https://discuss.huggingface.co/t/getting-error-attributeerror-inferenceclient-object-has-no-attribute-post/156682
- We can also no longer use id: str = "jinaai/jina-embeddings-v2-base-code" as default, because these models are no longer provided by the HF Inference API. Changed the default to id: str = "intfloat/multilingual-e5-large"
Add role_map for OpenAIChat: This allows certain models that don’t adhere to OpenAI’s role mapping to be used vir OpenAILike.
Use Content Hash as ID in Upsert in Pgvector: Use reproducible content_hash in upsert as ID.
Insert in Vector DB passes only last chunk meta_data: Insert in vector db passes only last chunk meta_data. issue link- https://discord.com/channels/965734768803192842/1219054452221153463/1376631140047130649
Remove Argument Sanitization: Replaced with a safer way to do this that won’t break arguments that shouldn’t be sanitized
Handle async tools when running async agents on playground: Fixed a regression where using Agents with async tools (e.g. MCP tools) was breaking in the Playground.

2025-05-23

v1.5.4

New Features:

User Control Flows: This is the beta release of Agno’s Human-in-the-loop flows and tools.
- We now allow agent runs to be paused awaiting completion of certain user requirements before the agent can continue.
- This also adds the agent.continue_run and agent.acontinue_run functions.
- The control flows that are available:
  - User confirmation flow → Decorate a function with @tool(requires_confirmation=True) and the agent will expect user confirmation before executing the tool.
  - User input required → Decorate a function with @tool(requires_user_input=True) to have the agent stop and ask for user input before continuing.
  - External tool execution → Decorate a function with @tool(external_execution=True) to indicate that you will execute this function outside of the agent context.
  - Dynamic user input → Add UserControlFlowTools() to an agent to give the agent the ability to dynamically stop the flow and ask for user input where required.
- See a host of examples here.
Mem0 Toolkit: Added a toolkit for managing memories in Mem0.
Firecrawl Search: Added support for Firecrawl web search in FirecrawlTools.

Bug Fixes:

Firecrawl Tools and Reader: Fixed parameter parsing for the Firecrawl reader and tools.
Include/Exclude on all Tools: Ensure all toolkits support include_tools and exclude_tools.

2025-05-21

v1.5.3

Improvements:

Improved Accuracy Evals: Updated the way accuracy evals works for more accurate agent-based evaluation.

Bug Fixes:

MCP Client Timeout: Client timeouts now work correctly and use the timeout set on parameters.

2025-05-20

v1.5.2

New Features:

Agno Apps (Beta): Introducing Agno Apps, convenience functions to assist with building production-ready applications. The first supported apps are:
- FastAPIApp → A really simple FastAPI server that provides access to your agent or team.
- WhatsappAPIApp → An app that implements the Whatsapp protocol allowing you to have an Agno agent running on Whatsapp. Supports image / audio / video input and can generate images as responses. Supports reasoning.
Couchbase Vector DB Support: Added support for Couchbase as a vector DB for knowledge bases.
Knowledge Filters Update for Teams: Filters (manual + agentic) can now be used with Teams.
Azure Cosmos DB for MongoDB (vCore) vector db support: In the MongoDB vector db class add support for cosmosdb mongo vcore support by enabling cosmos_compatibility=True
Google Big Query Tools: Added Toolkit for Google BigQuery support.
Async Support for s3 Readers: Add async support for pdf and text s3 readers.
stop_after_tool_call and show_result for Toolkits: Now the base Toolkit class has stop_after_tool_call_tools and show_result_tools similar to the @tool decorator.

2025-05-16

v1.5.1

New Features:

Nebius Model Provider: Added Nebius as a model provider.
Extended Filters Support on Vector DBs: Added filtering support for other vector DBs.
- pgvector
- Milvus
- Weaviate
- Chroma

Improvements:

Redis SSL: Added the ssl parameter to Redis storage.

2025-05-13

v1.5.0

New Features:

Azure OpenAI Tools: Added image generation via Dall-E via Azure AI Foundry.
OpenTelemetry Instrumentation: We have contributed to the OpenInference project and added an auto-instrumentor for Agno agents. This adds tracing instrumentation for Agno Agents for any OpenTelemetry-compatible observability provider. These include Arize, Langfuse and Langsmith. Examples added to illustrate how to use each one (here).
Evals Updates: Added logic to run accuracy evaluations with pre-generated answers and minor improvements for all evals classes.
Hybrid Search and Reranker for Milvus Vector DB: Added support for hybrid_search on Milvus.
MCP with Streamable-HTTP: Now supporting the streamable-HTTP transport for MCP servers.

Improvements:

Knowledge Filters Cookbook: Instead of storing the sample data locally, we now pull it from s3 at runtime to keep the forking of the repo as light as possible.

Bug Fixes:

Team Model State: Fixed issues related to state being shared between models on teams.
Concurrent Agent Runs: Fixed certain race-conditions related to running agents concurrently.

Breaking changes:

Evals Refactoring:
- Our performance evaluation class has been renamed from PerfEval to PerformanceEval
- Our accuracy evaluation class has new required fields: agent, prompt and expected_answer
Concurrent Agent Runs: We removed duplicate information from some events during streaming (stream=True). Individual events will have more relevant data now.

2025-05-10

v1.4.6

New Features:

Cerebras Model Provider: Added Cerebras as a model provider.
Claude Web Search: Added support for Claude’s new web search tool.

Knowledge Base Metadata Filtering (Beta): Added support for filtering documents by metadata

Two Ways to Apply Filters:

Explicit Filtering: Pass filters directly to Agent or during run/query

# Option 1: Filters on Agent initialization
agent = Agent(
					knowledge=knowledge_base, 
					knowledge_filters={"filter_1": "abc"}
				)
     
# Option 2: Filters on run execution
agent.run("Tell me about...", knowledge_filters={"filter_1": "abc"})

See docs here

Agentic Filtering: Agent automatically detects and applies filters from user queries

# Enable automatic filter detection
agent = Agent(
					knowledge=knowledge_base, 
					enable_agentic_knowledge_filters=True
				)
     
# Agent extracts filters from query
agent.run("Tell me about John Doe's experience...")

See docs here

Two approaches for adding metadata to documents:

During Knowledge Base Initialization:

knowledge_base = PDFKnowledgeBase(path=[
     {
		     "path": "file_1.pdf", 
		     "metadata": {
				     "user_id": "abc"
				  }
		 },
		 {
		     "path": "file_2.pdf", 
		     "metadata": {
				     "user_id": "xyz"
				  }
		 }
])

During Individual Document Loading:

knowledge_base.load_document(
     path="file.pdf",
     metadata={"user_id": "abc"}
)

Compatibility
- Knowledge Base Types: PDF, Text, DOCX, JSON, and PDF_URL
- Vector Databases: Qdrant, LanceDB, and MongoDB

Improvements:

User and Session ID in Tools: Added current_user_id and current_session_id as default variables in session_data for Agent and Team.

Bug Fixes:

Knowledge Base ID Clashes: Knowledge files with overlapping names (e.g., abc.-.xyz.pdf and abc.-.def.pdf) were being incorrectly identified due to the readers using formatted names as unique id which were getting uniqueness conflict. Introduced a unique ID for each document in all the readers using uuidv4() to ensure strict identification and prevent conflicts.

2025-05-06

v1.4.5

New Features:

Embedder Support via AWS Bedrock: AwsBedrockEmbedder has been added with a default embedding model of cohere.embed-multilingual-v3.
Gemini Video Generation Tool: Added video generation capabilities to GeminiTools.

Improvements:

Apify Revamp: Complete revamp of ApifyTools to make it completely compatible with Apify actors.

Bug Fixes:

Tools with Optional Parameters on Llama API: Fixed edge cases with functions.

2025-05-03

v1.4.4

New Features:

OpenAI File Support: Added support for File attached to prompts for agents with OpenAIChat models.

Improvements:

Llama API: Various improvements for Llama and LlamaOpenAI model classes including structured output and image input support
Async Custom Retriever: The retriever parameter can now be an async function to be used with agent.arun and agent.aprint_response.
Gemini Video URL Input: Added support for Video(url=...) for Gemini.

Bug Fixes:

OpenAI Responses o3 / o4 Tools: Fixed broken tool use for advanced reasoning models on OpenAIResponses.
MCP on CLI Support: Fixed support for MCPTools usage while calling agent.acli_app.

2025-04-30

v1.4.3

New Features:

Llama API: Added native SDK and OpenAI-like model classes.

Improvements:

Claude: Added support for AWS Session token for Claude.
DynamoDB: Added support for AWS profile-based authentication.

Bug Fixes:

Session Metrics: Fix for session metrics showing up as 0.
HF Embedder fix: Fixed Hugging Face Embedder.

2025-04-25

v1.4.2

New Features:

MCP SSE Support: Added support for connecting to SSE MCP Servers.
Tool Hooks: You can now have a hook that is wrapped around all tool calls. This works for Toolkits and custom tools. See this example.
Team Session State: You can now manage a single state dictionary across a team leader and team members inside tools given to the team leader/members. See this example.
Cartesia Tool: Added support for Cartesia for text-to-speech capabilities.
Gemini Image Tools: Added a tool that uses Gemini models to generate images.
Groq Audio Tools: Added a tool that uses Groq models to translate, transcribe and generate audio.

Improvements:

PubmedTools Expanded Results: Added expanded result sets for PubmedTools .
Variety in Tool Results: Custom tools can now have any return type and it would be handled before being provided to the model.

Bug Fixes:

Teams Shared Model Bug: Fixed issues where a single model is used across team members. This should reduce tool call failures in team execution.

2025-04-23

v1.4.0

New Features:

Memory Generally Available: We have made improvements and adjustments to how Agentic user memory management works. This is now out of beta and generally available. See these examples and these docs for more info.
OpenAI Tools: Added OpenAITools to enable text-to-speech and image generation through OpenAI’s APIs.
Zep Tools: Added ZepTools and AsyncZepTools to manage memories for your Agent using zep-cloud

Improvements:

Azure AI Foundry Reasoning: Added support for reasoning models via Azure AI Foundry. E.g. Deepseek-R1.
Include/Exclude Tools: Added include_tools and exclude_tools for all toolkits. This allows for selective enabling / disabling of tools inside toolkits, which is especially useful for larger toolkits.

Bug Fixes:

Gemini with Memory: Fixed issue with deepcopy when Gemini is used with Memory.

Breaking Changes:

Memory: Agents will now by default use an improved Memory instead of the now deprecated AgentMemory. - agent.memory.messages → run.messages for run in agent.memory.runs (or agent.get_messages_for_session()) - create_user_memories → enable_user_memories and is now set on the Agent/Team directly. - create_session_summary → enable_session_summaries and is now set on the Agent/Team directly.

2025-04-21

v1.3.5

Improvements:

Further Async Vector DB Support: Support added for:
- Clickhouse
- ChromaDB
- Cassandra
- PineconeDB
- Pgvector
Reasoning on Agno Platform:
- Added extensive support for reasoning on the Agno Platform. Go see your favourite reasoning agents in action!
- Changes from SDK
  - send proper events in different types of reasoning and populate the reasoning_content on RunResponse for stream/non-stream, async/non-async
  - unified json structure for all types of reasoning in Reasoning events
Google Caching Support: Added support for caching files and sending the cached content to Gemini.

Bug Fixes

Firecrawl Scrape: Fixed issues with non-serializable types for during Firecrawl execution. https://github.com/agno-agi/agno/issues/2883

2025-04-18

v1.3.4

New Features:

Web Browser Tool: Introduced a webbrowser tool for agents to interact with the web.
Proxy Support: Added proxy parameter support to both URL and PDF tools for network customization.

Improvements:

Session State: Added examples for managing session state in agents.
AzureOpenAIEmbedder: Now considers client_params passed in the client_params argument for more flexible configuration.
LiteLLM: Now uses built-in environment validation to simplify setup.
Team Class: Added a mode attribute to team data serialization for enhanced team configuration.
Insert/Upsert/Log Optimization: insert/upsert/log_info operations now trigger only when documents are present in the reader.
Database Preference: Session state now prefers database-backed storage if available.
Memory Management: Internal memory system updated for better session handling and resource efficiency.
Module Exports: Init files that only import now explicitly export symbols using __all__.

Bug Fixes:

DynamoDB Storage: Fixed an issue with storage handling in DynamoDB-based setups.
DeepSeek: Fixed a bug with API key validation logic.

2025-04-17

v1.3.3

Improvements:

Gemini File Upload: Enabled direct use of uploaded files with Gemini.
Metrics Update: Added audio, reasoning and cached token counts to metrics where available on models.
Reasoning Updates: We now natively support Ollama and AzureOpenAI reasoning models.

Bug Fixes:

PPrint Util Async: Added apprint_run_response to support async.
Mistral Reasoning: Fixed issues with using a Mistral model for chain-of-thought reasoning.

2025-04-16

v1.3.2

New Features:

Redis Memory DB: Added Redis as a storage provider for Memory. See here.

Improvements:

Memory Updates: Various performance improvements made and convenience functions added:
- agent.get_session_summary() → Use to get the previous session summary from the agent.
- agent.get_user_memories() → Use to get the current user’s memories.
- You can also add additional instructions to the MemoryManager or SessionSummarizer.
Confluence Bypass SSL Verification: If required, you can now skip SSL verification for Confluence connections.
More Flexibility On Team Prompts: Added add_member_tools_to_system_message to remove the member tool names from the system message given to the team leader, which allows flexibility to make teams transfer functions work in more cases.

Bug Fixes:

LiteLLM Streaming Tool Calls: Fixed issues with tool call streaming in LiteLLM.
E2B Casing Issue: Fixed issues with parsed Python code that would make some values lowercase.
Team Member IDs: Fixed edge-cases with team member IDs causing teams to break.

2025-04-12

v1.3.0

New Features:

Memory Revamp (Beta): This is a beta release of a complete revamp of Agno Memory. This includes a new Memory class that supports adding, updating and deleting user memories, as well as doing semantic search with a model. This also adds additional abilities to the agent to manage memories on your behalf. See the docs here.
User ID and Session ID on Run: You can now pass user_id and session_id on agent.run(). This will ensure the agent is set up for the session belonging to the session_id and that only the memories of the current user is accessible to the agent. This allows you to build multi-user and multi-session applications with a single agent configuration.
Redis Storage: Support added for Redis as a session storage provider.

2025-04-11

v1.2.16

Improvements:

Teams Improvements: Multiple improvements to teams to make task forwarding to member agents more reliable and to make the team leader more conversational. Also added various examples of reasoning with teams.
Knowledge on Teams: Added knowledge to Team to better align with the functionality on Agent. This comes with retriever to set a custom retriever and search_knowledge to enable Agentic RAG.

Bug Fixes:

Gemini Grounding Chunks: Fixed error when Gemini Grounding was used in streaming.
OpenAI Defaults in Structured Outputs: OpenAI does not allow defaults in structured outputs. To make our structured outputs as compatible as possible without adverse effects, we made updates to OpenAIResponses and OpenAIChat.

2025-04-08

v1.2.14

Improvements:

Improved Github Tools: Added many more capabilities to GithubTools.
Windows Scripts Support: Converted all the utility scripts to be Windows compatible.
MongoDB VectorDB Async Support: MongoDB can now be used in async knowledge bases.

Bug Fixes:

Gemini Tool Formatting: Fixed various cases where functions would not be parsed correctly when used with Gemini.
ChromaDB Version Compatibility: Fix to ensure that ChromaDB and Agno are compatible with newer versions of ChromaDB.
Team-Member Interactions: Fixed issue where if members respond with empty content the team would halt. This is now be resolved.
Claude Empty Response: Fixed a case when the response did not include any content with tool calls resulting in an error from the Anthropic API

2025-04-07

v1.2.12

New Features:

Timezone Identifier: Added a new timezone_identifier parameter in the Agent class to include the timezone alongside the current date in the instructions.
Google Cloud JSON Storage: Added support for JSON-based session storage on Google Cloud.
Reasoning Tools: Added ReasoningTools for an advanced reasoning scratchpad for agents.

Improvements:

Async Vector DB and Knowledge Base Improvements: More knowledge bases have been updated for async-await support: - URLKnowledgeBase → Find some examples here. - FireCrawlKnowledgeBase → Find some examples here. - DocxKnowledgeBase → Find some examples here.

2025-04-07

v1.2.11

Bug Fixes:

Fix for structured outputs: Fixed cases of structured outputs for reasoning.

2025-04-07

v1.2.10

1.2.10 New Features:

Knowledge Tools: Added KnowledgeTools for thinking, searching and analysing documents in a knowledge base.

2025-04-05

v1.2.9

1.2.9 Improvements:

Simpler MCP Interface: Added MultiMCPTools to support multiple server connections and simplified the interface to allow command to be passed. See these examples of how to use it.

2025-04-04

v1.2.8

1.2.8

Changelog

New Features:

Toolkit Instructions: Extended Toolkit with instructions and add_instructions to enable you to specify additional instructions related to how a tool should be used. These instructions are then added to the model’s “system message” if add_instructions=True .

Bug Fixes:

Teams transfer functions: Some tool definitions of teams failed for certain models. This has been fixed.

2025-04-02

v1.2.7

1.2.7 New Features:

Gemini Image Generation: Added support for generating images straight from Gemini using the gemini-2.0-flash-exp-image-generation model.

Improvements:

Vertex AI: Improved use of Vertex AI with Gemini Model class to closely follow the official Google specification
Function Result Caching Improvement: We now have result caching on all Agno Toolkits and any custom functions using the @tool decorator. See the docs here.
Async Vector DB and Knowledge Base Improvements: Various knowledge bases, readers and vector DBs now have async-await support, so it will be used in agent.arun and agent.aprint_response. This also means that knowledge_base.aload() is possible which should greatly increase loading speed in some cases. The following have been converted:
- Vector DBs:
  - LanceDb → Here is a cookbook to illustrate how to use it.
  - Milvus → Here is a cookbook to illustrate how to use it.
  - Weaviate → Here is a cookbook to illustrate how to use it.
- Knowledge Bases:
  - JSONKnowledgeBase → Here is a cookbook to illustrate how to use it.
  - PDFKnowledgeBase → Here is a cookbook to illustrate how to use it.
  - PDFUrlKnowledgeBase → Here is a cookbook to illustrate how to use it.
  - CSVKnowledgeBase → Here is a cookbook to illustrate how to use it.
  - CSVUrlKnowledgeBase → Here is a cookbook to illustrate how to use it.
  - ArxivKnowledgeBase → Here is a cookbook to illustrate how to use it.
  - WebsiteKnowledgeBase → Here is a cookbook to illustrate how to use it.
  - YoutubeKnowledgeBase → Here is a cookbook to illustrate how to use it.
  - TextKnowledgeBase → Here is a cookbook to illustrate how to use it.

Bug Fixes:

Recursive Chunking Infinite Loop: Fixes an issue with RecursiveChunking getting stuck in an infinite loop for large documents.

2025-03-28

v1.2.6

1.2.6 Bug Fixes:

Gemini Function call result fix: Fixed a bug with function call results failing formatting and added proper role mapping .
Reasoning fix: Fixed an issue with default reasoning and improved logging for reasoning models .

2025-03-27

v1.2.5

1.2.5 New Features:

E2B Tools: Added E2B Tools to run code in E2B Sandbox

Improvements:

Teams Tools: Add tools and tool_call_limit to Team. This means the team leader itself can also have tools provided by the user, so it can act as an agent.
Teams Instructions: Improved instructions around attached images, audio, videos, and files. This should increase success when attaching artifacts to prompts meant for member agents.
MCP Include/Exclude Tools: Expanded MCPTools to allow you to specify tools to specifically include or exclude from all the available tools on an MCP server. This is very useful for limiting which tools the model has access to.
Tool Decorator Async Support: The @tool() decorator now supports async functions, including async pre and post-hooks.

Bug Fixes:

Default Chain-of-Thought Reasoning: Fixed issue where reasoning would not default to manual CoT if the provided reasoning model was not capable of reasoning.
Teams non-markdown responses: Fixed issue with non-markdown responses in teams.
Ollama tool choice: Removed tool_choice from Ollama usage as it is not supported.
Worklow session retrieval from storage: Fixed entity_id mappings.

2025-03-25

v1.2.4

1.2.4 Improvements:

Tool Choice on Teams: Made tool_choice configurable.

Bug Fixes:

Sessions not created: Made issue where sessions would not be created in existing tables without a migration be more visible. Please read the docs on storage schema migrations.
Todoist fixes: Fixed update_task on TodoistTools.

2025-03-24

v1.2.3

1.2.3 Improvements:

Teams Error Handling: Improved the flow in cases where the model gets it wrong when forwarding tasks to members.

2025-03-24

v1.2.2

1.2.2 Bug Fixes:

Teams Memory: Fixed issues related to memory not persisting correctly across multiple sessions.

2025-03-24

v1.2.1

1.2.1 Bug Fixes:

Teams Markdown: Fixed issue with markdown in teams responses.

2025-03-24

v1.2.0

1.2.0 New Features:

Financial Datasets Tools: Added tools for https://www.financialdatasets.ai/.
Docker Tools: Added tools to manage local docker environments.

Improvements:

Teams Improvements: Reasoning enabled for the team.
MCP Simplification: Simplified creation of MCPTools for connections to external MCP servers. See the updated docs.

Bug Fixes:

Azure AI Factory: Fix for a broken import in Azure AI Factory.

2025-03-23

v1.1.17

1.1.17 Improvements:

Better Debug Logs: Enhanced debug logs for better readability and clarity.

2025-03-22

v1.1.16

1.1.16 New Features:

Async Qdrant VectorDB: Implemented async support for Qdrant VectorDB, improving performance and efficiency.
Claude Think Tool: Introduced the Claude Think tool, following the specified implementation guide.

2025-03-21

v1.1.15

1.1.15 Improvements:

Tool Result Caching: Added caching of selected searchers and scrapers. This is only intended for testing and should greatly improve iteration speed, prevent rate limits and reduce costs (where applicable) when testing agents. Applies to:
- DuckDuckGoTools
- ExaTools
- FirecrawlTools
- GoogleSearchtools
- HackernewsTools
- NewspaperTools
- Newspaper4kTools
- Websitetools
- YFinanceTools
Show tool calls: Improved how tool calls are displayed when print_response and aprint_response is used. They are now displayed in a separate panel different from response panel. It can also be used in conjunction in response_model.

2025-03-20

v1.1.14

1.1.14 - Teams Revamp

New Features:

Teams Revamp: Announcing a new iteration of Agent teams with the following features:
- Create a Team in one of 3 modes: “Collaborate”, “Coordinate” or “Route”.
- Various improvements have been made that was broken with the previous teams implementation. Including returning structured output from member agents (for “route” mode), passing images, audio and video to member agents, etc.
- It has added features like “agentic shared context” between team members and sharing of individual team member responses with other team members.
- This also comes with a revamp of Agent and Team debug logs. Use debug_mode=True and team.print_response(...) to see it in action.
- Find the docs here. Please look at the example implementations here.
- This is the first release. Please give us feedback. Updates and improvements will follow.
- Support for Agent(team=[]) is still there, but deprecated (see below).
LiteLLM: Added LiteLLM support, both as a native implementation and via the OpenAILike interface.

Improvements:

Change structured_output to response_format: Added use_json_mode: bool = False as a parameter of Agent and Team, which in conjunction with response_model=YourModel, is used to indicate whether the agent/team model should be forced to respond in json instead of (now default) structured output. Previous behaviour defaulted to “json-mode”, but since most models now support native structured output, we are now defaulting to native structured output. It is now also much simpler to work with response models, since now only response_model needs to be set. It is not necessary anymore to set structured_output=True to specifically get structured output from the model.
Website Tools + Combined Knowledgebase: Added functionality for WebsiteTools to also update combined knowledgebases.

Bug Fixes:

AgentMemory: Fixed get_message_pairs() fetching incorrect messages.
UnionType in Functions: Fixed issue with function parsing where pipe-style unions were used in function parameters.
Gemini Array Function Parsing: Fixed issue preventing gemini function parsing to work in some MCP cases.

Deprecations:

Structured Output: Agent.structured_output has been replaced by Agent.use_json_mode. This will be removed in a future major version release.
Agent Team: Agent.team is deprecated with the release of our new Teams implementation here. This will be removed in a future major version release.

2025-03-14

v1.1.13

1.1.13 Improvements:

OpenAIResponses File Search: Added support for the built-in “File Search” function from OpenAI. This automatically uploads File objects attached to the agent prompt.
OpenAIReponses web citations: Added support to extract URL citations after usage of the built-in “Web Search” tool from OpenAI.
Anthropic document citations: Added support to extract document citations from Claude responses when File objects are attached to agent prompts.
Cohere Command A: Support and examples added for Coheres new flagship model

Bug Fixes:

Ollama tools: Fixed issues with tools where parameters are not typed.
Anthropic Structured Output: Fixed issue affecting Anthropic and Anthropic via Azure where structured output wouldn’t work in some cases. This should make the experience of using structured output for models that don’t natively support it better overall. Also now works with enums as types in the Pydantic model.
Google Maps Places: Support from Google for Places API has been changed and this brings it up to date so we can continue to support “search places”.

2025-03-13

v1.1.12

1.1.12 New Features:

Citations: Improved support for capturing, displaying, and storing citations from models, with integration for Gemini and Perplexity.

Improvements:

CalComTools: Improvement to tool Initialization.

Bug Fixes:

MemoryManager: Limit parameter was added fixing a KeyError in MongoMemoryDb.

2025-03-13

v1.1.11

1.1.11 New Features:

OpenAI Responses: Added a new model implementation that supports OpenAI’s Responses API. This includes support for their “websearch” built-in tool.
Openweather API Tool: Added tool to get real-time weather information.

Improvements:

Storage Refactor: Merged agent and workflow storage classes to align storage better for agents, teams and workflows. This change is backwards compatible and should not result in any disruptions.

2025-03-12

v1.1.10

1.1.10 New Features:

File Prompts: Introduced a new File type that can be added to prompts and will be sent to the model providers. Only Gemini and Anthropic Claude supported for now.
LMStudio: Added support for LMStudio as a model provider. See the docs.
AgentQL Tools: Added tools to support AgentQL for connecting agents to websites for scraping, etc. See the docs.
Browserbase Tool: Added Browserbase tool.

Improvements:

Cohere Vision: Added support for image understanding with Cohere models. See this cookbook to try it out.
Embedder defaults logging: Improved logging when using the default OpenAI embedder.

Bug Fixes:

Ollama Embedder: Fix for getting embeddings from Ollama across different versions.

2025-03-06

v1.1.9

1.1.9 New Features:

IBM Watson X: Added support for IBM Watson X as a model provider. Find the docs here.
DeepInfra: Added support for DeepInfra. Find the docs here.
Support for MCP: Introducing MCPTools along with examples for using MCP with Agno agents.

Bug Fixes:

Mistral with reasoning: Fixed cases where Mistral would fail when reasoning models from other providers generated reasoning content.

2025-03-03

v1.1.8

1.1.8 New Features:

Video File Upload on Playground: You can now upload video files and have a model interpret the video. This feature is supported only by select Gemini models with video processing capabilities.

Bug Fixes:

Huggingface: Fixed multiple issues with the Huggingface model integration. Tool calling is now fully supported in non-streaming cases.
Gemini: Resolved an issue with manually setting the assistant role and tool call result metrics.
OllamaEmbedder: Fixed issue where no embeddings were returned.

2025-02-26

v1.1.7

1.1.7 New Features:

Audio File Upload on Playground: You can now upload audio files and have a model interpret the audio, do sentiment analysis, provide an audio transcription, etc.

Bug Fixes:

Claude Thinking Streaming: Fix Claude thinking when streaming is active, as well as for async runs.

2025-02-24

v1.1.6

1.1.6 New Features:

-Claude 3.7 Support: Added support for the latest Claude 3.7 Sonnet model

Bug Fixes:

-Claude Tool Use: Fixed an issue where tools and content could not be used in the same block when interacting with Claude models.

2025-02-24

v1.1.5

1.1.5 New Features:

Audio Responses: Agents can now deliver audio responses (both with streaming and non-streaming).

The audio is in the agent.run_response.response_audio.

This only works with OpenAIChat with the gpt-4o-audio-preview model. See their docs for more on how it works. For example

from agno.agent import Agent
from agno.models.openai import OpenAIChat
from agno.utils.audio import write_audio_to_file

agent = Agent(
    model=OpenAIChat(
        id="gpt-4o-audio-preview",
        modalities=["text", "audio"],  # Both text and audio responses are provided.
        audio={"voice": "alloy", "format": "wav"},
    ),
)
agent.print_response(
    "Tell me a 5 second story"
)
if agent.run_response.response_audio is not None:
    write_audio_to_file(
        audio=agent.run_response.response_audio.base64_audio, filename=str(filename)
    )

See the audio_conversation_agent cookbook to test it out on the Agent Playground.

Image understanding support for Together.ai and XAi: You can now give images to agents using models from XAi and Together.ai.

Improvements:

Automated Tests: Added integration tests for all models. Most of these will be run on each pull request, with a suite of integration tests run before a new release is published.
Grounding and Search with Gemini: Grounding and Search can be used to improve the accuracy and recency of responses from the Gemini models.

Bug Fixes:

Structured output updates: Fixed various cases where native structured output was not used on models.
Ollama tool parsing: Fixed cases for Ollama with tools with optional parameters.
Gemini Memory Summariser: Fixed cases where Gemini models were used as the memory summariser.
Gemini auto tool calling: Enabled automatic tool calling when tools are provided, aligning behavior with other models.
FixedSizeChunking issue with overlap: Fixed issue where chunking would fail if overlap was set.
Claude tools with multiple types: Fixed an issue where Claude tools would break when handling a union of types in parameters.
JSON response parsing: Fixed cases where JSON model responses returned quoted strings within dictionary values.

2025-02-17

v1.1.4

1.1.4 Improvements:

Gmail Tools: Added get_emails_by_thread and send_email_reply methods to GmailTools.

Bug Fixes:

Gemini List Parameters: Fixed an issue with functions using list-type parameters in Gemini.
Gemini Safety Parameters: Fixed an issue with passing safety parameters in Gemini.
ChromaDB Multiple Docs: Fixed an issue with loading multiple documents into ChromaDB.
Agentic Chunking: Fixed an issue where OpenAI was required for chunking even when a model was provided.

2025-02-16

v1.1.3

1.1.3 Bug Fixes:

Gemini Tool-Call History: Fixed an issue where Gemini rejected tool-calls from historic messages.

2025-02-15

v1.1.2

1.1.2 Improvements:

Reasoning with o3 Models: Reasoning support added for OpenAI’s o3 models.

Gemini embedder update: Updated the GeminiEmbedder to use the new Google’s genai SDK. This update introduces a slight change in the interface:

# Before
embeddings = GeminiEmbedder("models/text-embedding-004").get_embedding(
    "The quick brown fox jumps over the lazy dog."
)

# After
embeddings = GeminiEmbedder("text-embedding-004").get_embedding(
    "The quick brown fox jumps over the lazy dog."
)

Bug Fixes:

Singlestore Fix: Fixed an issue where querying SingleStore caused the embeddings column to return in binary format.
MongoDB Vectorstore Fix: Fixed multiple issues in MongoDB, including duplicate creation and deletion of collections during initialization. All known issues have been resolved.
LanceDB Fix: Fixed various errors in LanceDB and added on_bad_vectors as a parameter.

2025-02-14

v1.1.1

1.1.1 Improvements:

File / Image Uploads on Agent UI: Agent UI now supports file and image uploads with prompts.
- Supported file formats: .pdf , .csv , .txt , .docx , .json
- Supported image formats: .png , .jpeg , .jpg , .webp
Firecrawl Custom API URL: Allowed users to set a custom API URL for Firecrawl.

Updated ModelsLabTools Toolkit Constructor: The constructor in /libs/agno/tools/models_labs.py has been updated to accommodate audio generation API calls. This is a breaking change, as the parameters for the ModelsLabTools class have changed. The url and fetch_url parameters have been removed, and API URLs are now decided based on the file_type provided by the user.

MODELS_LAB_URLS = {
    "MP4": "https://modelslab.com/api/v6/video/text2video",
    "MP3": "https://modelslab.com/api/v6/voice/music_gen",
    "GIF": "https://modelslab.com/api/v6/video/text2video",
}

MODELS_LAB_FETCH_URLS = {
    "MP4": "https://modelslab.com/api/v6/video/fetch",
    "MP3": "https://modelslab.com/api/v6/voice/fetch",
    "GIF": "https://modelslab.com/api/v6/video/fetch",
}

The FileType enum now includes MP3 type:

class FileType(str, Enum):
    MP4 = "mp4"
    GIF = "gif"
    MP3 = "mp3"

Bug Fixes:

Gemini functions with no parameters: Addressed an issue where Gemini would reject function declarations with empty properties.
Fix exponential memory growth: Fixed certain cases where the agent memory would grow exponentially.
Chroma DB: Fixed various issues related to metadata on insertion and search.
Gemini Structured Output: Fixed a bug where Gemini would not generate structured output correctly.
MistralEmbedder: Fixed issue with instantiation of MistralEmbedder.
Reasoning: Fixed an issue with setting reasoning models.
Audio Response: Fixed an issue with streaming audio artefacts to the playground.

2025-02-12

v1.1.0

1.1.0 - Models Refactor and Cloud Support

Model Improvements:

Models Refactor: A complete overhaul of our models implementation to improve on performance and to have better feature parity across models.
- This improves metrics and visibility on the Agent UI as well.
- All models now support async-await, with the exception of AwsBedrock.
Azure AI Foundry: We now support all models on Azure AI Foundry. Learn more here..
AWS Bedrock Support: Our redone AWS Bedrock implementation now supports all Bedrock models. It is important to note which models support which features.
Gemini via Google SDK: With the 1.0.0 release of Google’s genai SDK we could improve our previous implementation of Gemini. This will allow for easier integration of Gemini features in future.
Model Failure Retries: We added better error handling of third-party errors (e.g. Rate-Limit errors) and the agent will now optionally retry with exponential backoff if exponential_backoff is set to True.

Other Improvements

Exa Answers Support: Added support for the Exa answers capability.
GoogleSearchTools: Updated the name of GoogleSearch to GoogleSearchTools for consistency.

Deprecation

Our Gemini implementation directly on the Vertex API has been replaced by the Google SDK implementation of Gemini.
Our Gemini implementation via the OpenAI client has been replaced by the Google SDK implementation of Gemini.
Our OllamaHermes has been removed as the implementation of Ollama was improved.

Bug Fixes

Team Members Names: Fixed a bug where teams where team members have non-aphanumeric characters in their names would cause exceptions.

2025-02-07

v1.0.8

1.0.8 New Features:

Perplexity Model: We now support Perplexity as a model provider.
Todoist Toolkit: Added a toolkit for managing tasks on Todoist.
JSON Reader: Added a JSON file reader for use in knowledge bases.

Improvements:

LanceDb: Implemented name_exists function for LanceDb.

Bug Fixes:

Storage growth bug: Fixed a bug with duplication of run_messages.messages for every run in storage.

2025-02-05

v1.0.7

1.0.7 New Features:

Google Sheets Toolkit: Added a basic toolkit for reading, creating and updating Google sheets.
Weviate Vector Store: Added support for Weviate as a vector store.

Improvements:

Mistral Async: Mistral now supports async execution via agent.arun() and agent.aprint_response().
Cohere Async: Cohere now supports async execution via agent.arun() and agent.aprint_response().

Bug Fixes:

Retriever as knowledge source: Added small fix and examples for using the custom retriever parameter with an agent.

2025-02-05

v1.0.6

1.0.6 New Features:

Google Maps Toolkit: Added a rich toolkit for Google Maps that includes business discovery, directions, navigation, geocode locations, nearby places, etc.
URL reader and knowledge base: Added reader and knowledge base that can process any URL and store the text contents in the document store.

Bug Fixes:

Zoom tools fix: Zoom tools updated to include the auth step and other misc fixes.
Github search_repositories pagination: Pagination did not work correctly and this was fixed.

2025-02-03

v1.0.5

1.0.5 New Features:

Gmail Tools: Add tools for Gmail, including mail search, sending mails, etc.

Improvements:

Exa Toolkit Upgrade: Added find_similar to ExaTools
Claude Async: Claude models can now be used with await agent.aprint_response() and await agent.arun().
Mistral Vision: Mistral vision models are now supported. Various examples were added to illustrate example.

2025-02-02

v1.0.4

1.0.4 Bug Fixes:

Claude Tool Invocation: Fixed issue where Claude was not working with tools that have no parameters.

2025-01-31

v1.0.3

1.0.3 Improvements:

OpenAI Reasoning Parameter: Added a reasoning parameter to OpenAI models.

2025-01-31

v1.0.2

1.0.2 Improvements:

Model Client Caching: Made all models cache the client instantiation, improving Agno agent instantiation time
XTools: Renamed TwitterTools to XTools and updated capabilities to be compatible with Twitter API v2.

Bug Fixes:

Agent Dataclass Compatibility: Removed slots=True from the agent dataclass decorator, which was not compatible with Python < 3.10.
AzureOpenAIEmbedder: Made AzureOpenAIEmbedder a dataclass to match other embedders.

2025-01-31

v1.0.1

1.0.1 Improvement:

Mistral Model Caching: Enabled caching for Mistral models.

2025-01-30

v1.0.0

1.0.0 - Agno

This is the major refactor from phidata to agno, released with the official launch of Agno AI.See the migration guide for additional guidance.

Interface Changes:

phi.model.x → agno.models.x
phi.knowledge_base.x → agno.knowledge.x (applies to all knowledge bases)
phi.document.reader.xxx → agno.document.reader.xxx_reader (applies to all document readers)
All Agno toolkits are now suffixed with Tools. E.g. DuckDuckGo → DuckDuckGoTools

Multi-modal interface updates:

agent.run(images=[]) and agent.print_response(images=[]) is now of type Image

class Image(BaseModel):
    url: Optional[str] = None  # Remote location for image
    filepath: Optional[Union[Path, str]] = None  # Absolute local location for image
    content: Optional[Any] = None  # Actual image bytes content
    detail: Optional[str] = None # low, medium, high or auto (per OpenAI spec https://platform.openai.com/docs/guides/vision?lang=node#low-or-high-fidelity-image-understanding)
    id: Optional[str] = None

agent.run(audio=[]) and agent.print_response(audio=[]) is now of type Audio

class Audio(BaseModel):
    filepath: Optional[Union[Path, str]] = None  # Absolute local location for audio
    content: Optional[Any] = None  # Actual audio bytes content
    format: Optional[str] = None

agent.run(video=[]) and agent.print_response(video=[]) is now of type Video

class Video(BaseModel):
    filepath: Optional[Union[Path, str]] = None  # Absolute local location for video
    content: Optional[Any] = None  # Actual video bytes content

RunResponse.images is now a list of type ImageArtifact

class ImageArtifact(Media):
    id: str
    url: str  # Remote location for file
    alt_text: Optional[str] = None

RunResponse.audio is now a list of type AudioArtifact

class AudioArtifact(Media):
    id: str
    url: Optional[str] = None  # Remote location for file
    base64_audio: Optional[str] = None  # Base64-encoded audio data
    length: Optional[str] = None
    mime_type: Optional[str] = None

RunResponse.videos is now a list of type VideoArtifact

class VideoArtifact(Media):
    id: str
    url: str  # Remote location for file
    eta: Optional[str] = None
    length: Optional[str] = None

RunResponse.response_audio is now of type AudioOutput

class AudioOutput(BaseModel):
    id: str
    content: str  # Base64 encoded
    expires_at: int
    transcript: str

Models:
- Hermes → OllamaHermes
- AzureOpenAIChat → AzureOpenAI
- CohereChat → Cohere
- DeepSeekChat → DeepSeek
- GeminiOpenAIChat → GeminiOpenAI
- HuggingFaceChat → HuggingFace

Embedders now all take id instead of model as a parameter. For example

db_url = "postgresql+psycopg://ai:ai@localhost:5532/ai"

knowledge_base = PDFUrlKnowledgeBase(
    urls=["https://agno-public.s3.amazonaws.com/recipes/ThaiRecipes.pdf"],
    vector_db=PgVector(
        table_name="recipes",
        db_url=db_url,
        embedder=OllamaEmbedder(id="llama3.2", dimensions=3072),
    ),
)
knowledge_base.load(recreate=True)

Agent Storage class
- PgAgentStorage → PostgresDbAgentStorage
- SqlAgentStorage → SqliteDbAgentStorage
- MongoAgentStorage → MongoDbAgentStorage
- S2AgentStorage → SingleStoreDbAgentStorage
Workflow Storage class
- SqlWorkflowStorage → SqliteDbWorkflowStorage
- PgWorkflowStorage → PostgresDbWorkflowStorage
- MongoWorkflowStorage → MongoDbWorkflowStorage
Knowledge Base
- phi.knowledge.pdf.PDFUrlKnowledgeBase → agno.knowledge.pdf_url.PDFUrlKnowledgeBase
- phi.knowledge.csv.CSVUrlKnowledgeBase → agno.knowledge.csv_url.CSVUrlKnowledgeBase
Readers
- phi.document.reader.arxiv → agno.document.reader.arxiv_reader
- phi.document.reader.docx → agno.document.reader.docx_reader
- phi.document.reader.json → agno.document.reader.json_reader
- phi.document.reader.pdf → agno.document.reader.pdf_reader
- phi.document.reader.s3.pdf → agno.document.reader.s3.pdf_reader
- phi.document.reader.s3.text → agno.document.reader.s3.text_reader
- phi.document.reader.text → agno.document.reader.text_reader
- phi.document.reader.website → agno.document.reader.website_reader

Improvements:

Dataclasses: Changed various instances of Pydantic models to dataclasses to improve the speed.
Moved Embedder class from pydantic to data class

Removals

Removed all references to Assistant
Removed all references to llm
Removed the PhiTools tool
On the Agent class, guidelines, prevent_hallucinations, prevent_prompt_leakage, limit_tool_access, and task has been removed. They can be incorporated into the instructions parameter as you see fit.

Bug Fixes:

Semantic Chunking: Fixed semantic chunking by replacing similarity_threshold param with threshold param.

New Features:

Evals for Agents: Introducing Evals to measure the performance, accuracy, and reliability of your Agents.

Changelog

​New Features:

​Improvements:

​Bug Fixes:

​New Features:

​Improvements:

​Bug Fixes:

​New Features:

​Improvements:

​Bug Fixes:

​Workflows 2.0

​New Features:

​New Features:

​Bug Fixes:

​New Features:

​Improvements:

​Bug Fixes:

​New Features:

​Improvements:

​Bug Fixes:

​New Features:

​Improvements:

​Bug Fixes:

​Breaking Change:

​New Features:

​Improvements:

​Bug Fixes:

​New Features:

​Updates

​Bug Fixes:

​New Features:

​Improvements:

​Bug Fixes:

​Improvements:

​Bug Fixes:

​Breaking Changes:

​New Features:

​Improvements:

​Bug Fixes:

​New Features:

​Improvements:

​Bug Fixes:

​New Features:

​Improvements:

​Bug Fixes:

​Updates

​New Features

​Improvements:

​Bug Fixes:

​Updates:

​New Features:

​Improvements:

​Bug Fixes:

​New Features:

​Bug Fixes:

​Improvements:

​Bug Fixes:

​New Features:

​New Features:

​Improvements:

​New Features:

​Improvements:

​Bug Fixes:

​Breaking changes:

​New Features:

​Improvements:

​Bug Fixes:

​New Features:

​Improvements:

​Bug Fixes:

​New Features:

​Improvements:

​Bug Fixes:

​New Features:

​Improvements:

​Bug Fixes:

​New Features:

​Improvements:

​Bug Fixes:

​New Features:

New Features:

Improvements:

Bug Fixes:

New Features:

Improvements:

Bug Fixes:

New Features:

Improvements:

Bug Fixes:

Workflows 2.0

New Features:

New Features:

Bug Fixes:

New Features:

Improvements:

Bug Fixes:

New Features:

Improvements:

Bug Fixes:

New Features:

Improvements:

Bug Fixes:

Breaking Change:

New Features:

Improvements:

Bug Fixes:

New Features:

Updates

Bug Fixes:

New Features:

Improvements:

Bug Fixes:

Improvements:

Bug Fixes:

Breaking Changes:

New Features:

Improvements:

Bug Fixes:

New Features:

Improvements:

Bug Fixes:

New Features:

Improvements:

Bug Fixes:

Updates

New Features

Improvements:

Bug Fixes:

Updates:

New Features:

Improvements:

Bug Fixes:

New Features:

Bug Fixes:

Improvements:

Bug Fixes:

New Features:

New Features:

Improvements:

New Features:

Improvements:

Bug Fixes:

Breaking changes:

New Features:

Improvements:

Bug Fixes:

New Features:

Improvements:

Bug Fixes:

New Features:

Improvements:

Bug Fixes:

New Features:

Improvements:

Bug Fixes:

New Features:

Improvements:

Bug Fixes:

New Features:

Improvements: