NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
PyPI, #2885 most downloaded on PyPI
The programming language for agentic software.
Latest stable
3.2.0
Last release 9 days ago
12 Nov 2026
Ships on a steady schedule
a new release about every 9 days
Nearly every release is documented
notes for 59 of the last 60 stable releases
Nothing withdrawn
no release was ever pulled
2 years old
278 releases · first in 2024
Drag across the ruler to compare, or tap a release to open it.
Compression Manager Token Counting: Added support for token count based compression across providers. Read more here
RunRequirement to ensure it is properly converted into JSON before
persisting the related run, to avoid DB errors or missing data.retry_with_guidance_limit parameter to adjust how many times these
retries should happen.Full Changelog: https://github.com/agno-agi/agno/compare/v2.3.11...v2.3.12
Your coding agent can read these notes before it upgrades. Set up the MCP server →
Fixed an issue with custom readers not always displaying correctly on AgentOS.
RunOutput and RunCompletedEvent classes, to make them available both in streaming and non-streaming cases.AsyncPostgresDb or AsyncSqliteDb classes.Full Changelog: https://github.com/agno-agi/agno/compare/v2.3.10...v2.3.11
Shopify Toolkit: Get analytics and insights for your shopify store backend. Analyse sales, customers, etc. with ShopifyTools.
ShopifyTools.stream=False parameter to all agent runs within eval code.yield events and updates are made to session_state .Full Changelog: https://github.com/agno-agi/agno/compare/v2.3.9...v2.3.10
Async MySQL: Added support for AsyncMySQLDb with compatibility for the asyncmy driver. See the docs.
AsyncMySQLDb with compatibility for the asyncmy driver. See the docs.create_schema=False when initializing your database class (e.g. PostgresDb) to skip automatic schema creation. This is useful for environments where the schema needs to be managed externally.get_step_output for nested steps in Parallel groups: Add recursive search to get_step_output() to find steps nested inside Parallel, Condition, Router, Loop, and Steps.reasoning_content to the message for thinking mode.run_id param in run methods: Added a run_id parameter to the run and arun methods in the Agent, Team and Workflow classes, to set the id of the run instead of automatically generating one.add_content_ functions instead of wrapping with asyncioInputCheckError was raised with retries=True.Full Changelog: https://github.com/agno-agi/agno/compare/v2.3.8...v2.3.9
Model-Level Retries: Improved retry logic added to the model execution layer. That means setting retries=n on your model, would retry directly at the
retries=n on your model, would retry directly at the model execution layer, which is useful for dealing with ratelimits imposed by the model provider. Agent-level retries still exist and now apply to any exceptions raised in the agent execution loop. See the docs.contextvars (used by tracing/instrumentation libraries like OpenTelemetry) are not propagated when steps run in parallel using ThreadPoolExecutor.dependencies instead of run_context on aget_relevant_docs_from_knowledgeMemoriTools has been removed by the Memori team, since it is now obsolete given how the Memori framework will work going forward. See the new docs.Full Changelog: https://github.com/agno-agi/agno/compare/v2.3.7...v2.3.8
Amazon Redshift Tools: Added toolkit for exploring Amazon Redshift databases and running queries. See the docs.
RunRequirement class to handle HITL requirements. Now during HITL flows, you will find these objects in the Agent response or RunPaused event (if streaming). You can handle and resolve the HITL requirements via these objects. You can read more and see examples in the docs.continue_run: You can now pass yield_run_response to the continue run methods when streaming, to yield a RunOutput object at the end.run_context into get_relevant_documents_from_knowledge so custom retrievers has access to dependenciesAsyncPostgresDb class.Full Changelog: https://github.com/agno-agi/agno/compare/v2.3.6...v2.3.7
Spotify Tools: Added toolkit for managing your Spotify library. See the docs.
session_state in certain scenarios when using Worfklows.or {} with explici… by @opleonnn in https://github.com/agno-agi/agno/pull/5589Full Changelog: https://github.com/agno-agi/agno/compare/v2.3.5...v2.3.6
Hooks as Background Tasks: Agent/Team pre- and post-hooks can now be run as Background Tasks on the AgentOS, enabling completely non-blocking, concurr
cancel_run endpoint failures by ensuring register_run() is called earlier in the execution flow, eliminating the gap where run cancellation requests could fail.RunEvent serialization: Fixed a bug serializing RunEvent instances when working with pre-hooks and inputs containing media (e.g images).created_at field of the user Memory table when using Dynamo as database.delete_user_memory function of the AsyncPostgresDB class.Full Changelog: https://github.com/agno-agi/agno/compare/v2.3.4...v2.3.5
AsyncPostgresDB: Fixed a signature mismatch with delete_user_memories function.
delete_user_memories function.Full Changelog: https://github.com/agno-agi/agno/compare/v2.3.3...v2.3.4
Context Compression: Introducing compression of tool call results in a running context. Context Compression allows you to manage your agent context wh
optimize_memories and aoptimize_memories to the MemoryManager as a way to summarize memories outside of agent runs. Only one strategy is currently supported, but this will be extended in future. Thie is still a beta feature. See the docs.cookbook/models/google/gemini/file_search_basic.pycookbook/models/google/gemini/file_search_advanced.pycookbook/models/google/gemini/file_search_rag_pipeline.pyoutput_schema on run/arun: override output_schema at runtime via the run() and arun() methods for both Agent and Team, as well as AgentOS API endpoints.api_key support for AWS Bedrock. See the documentation.thinking_level on Geminiget_user_message functions on Agent and Teamstream_member_events=True but stream_events=False on Team.StopAgentRun exception from not correctly ending the agent run.tool_name_prefix: Made the tool_name_prefix parameter None by default to prevent it from adding _ by default as a prefix.tool_call_id fixedFull Changelog: https://github.com/agno-agi/agno/compare/v2.3.2...v2.3.3
chore: relax deprecation logs by @manuhortet in https://github.com/agno-agi/agno/pull/5480
Full Changelog: https://github.com/agno-agi/agno/compare/v2.3.1...v2.3.2
Nano Banana Support: You can now generate images using Google’s Nano Banana with the NanoBananaTools toolkit.
NanoBananaTools toolkit.MCPTools or MultiMCPTools classes inside the AgentOS.Full Changelog: https://github.com/agno-agi/agno/compare/v2.3.0...v2.3.1
Removed AgentOS Parameters: The following parameters have been removed from the AgentOS class after being deprecated in our previous v2.2.0 version:
ModelLabsTools.MigrationManager class together with our first migrations for the sessions and memories tables.
get_messages and get_chat_history in the Session classes.get_session_messages and get_chat_history in the Agent, Team and Workflow classes.get_chat_history also on the Workflow Step class.get_messages and get_session_messages functions now have all filtering capabilities. See the docs.AsyncMongoDb class in contexts where multiple event loops exist, like some advance use cases with the AgentOS.httpx client for much better resource allocation and instantiation speed. Also adds http2 for better http multiplexing.show_member_responses: You can now set show_member_responses on print_response, aprint_response for Team.RedisCluster when configuring a RedisDb instance for agent database.knowledge now requires the use of contents_db.knowledge.get_filters() method to always return valid filter keys instead of dummy keys.get_previous_session_messages() to return the most recent N sessions instead of the oldest N sessions.created_at added as column for memories tablefeedback added as column for memories table where it was missingPostgresDb, all JSON columns are converted to JSONB.MigrationManager (docs) to ensure these migrations are applied correctly. Alternatively if you are still in development, you can delete your current tables and start with a clean table which would have the updated schema already.api.tokenfactory.nebius.com endpoint.
base_url, but by default the class uses the Token Factory URL as per the Nebius product release.delegate_task_to_all_members parameter in Team to delegate_to_all_members.GoogleSearchTools Toolkit has been removed, as scrapping Google results in the way it was done here is not longer possible. We recommend using DuckDuckGoTools instead for your web searching needs.stream_events parameter has been removed from the print_response, aprint_response, and CLI methods, for Agent, Team and Workflow. Event handling is now always done correctly when streaming a response with these methods.knowledge(...) now requires setting the contents_db if you want to use knowledge_filters . See the docs on filtering.get_messages_for_session from Agent and Team.get_messages_from_last_n_runs from the Session, Agent, and Team classes.get_messages and get_chat_history in the Session classes.get_session_messages and get_chat_history in the Agent, Team and Workflow classes.get_chat_history also on the Step class.get_messages and get_session_messages functions now have all filtering capabilities. See the docs.v2.2.0 version:
os_id: You should now use idfastapi_app: You should now use base_appenable_mcp: You should now use enable_mcp_serverreplace_routes: You should now use on_route_conflictknowledge.get_filters() to return valid filter keys by @kausmeows in https://github.com/agno-agi/agno/pull/5445get_previous_session_messages function to return most recent session msgs by @kausmeows in https://github.com/agno-agi/agno/pull/5472Full Changelog: https://github.com/agno-agi/agno/compare/v2.2.13...v2.3.0
Workflow Step Events: Fixed event streaming for Agents and Teams in custom executors steps
Full Changelog: https://github.com/agno-agi/agno/compare/v2.2.12...v2.2.13
Knowledge Search Filter Expressions: Added a new metadata-based Filter DSL (EQ, IN, GT/LT, NOT, AND, OR) enabling targeted Knowledge searches across.
reply_to_mentions_only parameter was also added to adjust the behavior.any as a type was not parsed correctly for Gemini. This affected Agentic Knowledge Filtering, which is also fixed.Full Changelog: https://github.com/agno-agi/agno/compare/v2.2.11...v2.2.12
Parallel Tools: Added ParallelTools toolkit for AI-optimized web search and content extraction. Includes both direct API integration and MCP server su
duration to the top-level Workflow metrics, showing the total duration of the complete Workflow run.resync.Full Changelog: https://github.com/agno-agi/agno/compare/v2.2.10...v2.2.11
Adds a new yield_run_output flag on Agent/Team run/arun functions to be used in favor of yield_run_response . yield_run_response to be deprecated
run_context parameter is now the recommended way to read and modify state data everywhere in Workflows, including: Agents, Teams, tools, steps and custom Python functions used in steps. You can read more in the Workflow state docs.yield_run_output flag on Agent/Team run/arun functions to be used in favor of yield_run_response . yield_run_response to be deprecatedFull Changelog: https://github.com/agno-agi/agno/compare/v2.2.9...v2.2.10
Strict output on Models: Added the strict_output parameter to all models that can work with it. When set, this will guarantee the model’s generated re
strict_output parameter to all models that can work with it. When set, this will guarantee the model’s generated response adheres to the contextual output_schema. It defaults to True.session_state when running the Agno Agents or Teams.CustomEvent events will now be better mapped into AG-UI custom events containing all relevant fields.knowledge_filters parameter is now supported. Also improve handling of boolean and null fields presented in non-Python format.MongoVectorDb and RedisVectorDbFull Changelog: https://github.com/agno-agi/agno/compare/v2.2.8...v2.2.9
Remove Redis dependency: Remove redis and redisvl as dependencies of the base VectorDb class.
redis and redisvl as dependencies of the base VectorDb class.Full Changelog: https://github.com/agno-agi/agno/compare/v2.2.7...v2.2.8
Run context: You can now gain access to the run context in your tools, hooks, tool hooks, dependency functions and instructions functions, by adding a
run_context parameter to them. This streamlines access and management of state data, and fixes some edge cases. Read more in our docs.tool_name_prefix as a parameter in the MCPTools class, allowing to add a prefix to all tool names related to the contextual MCP server. This is useful to avoid name collisions when using multiple MCP servers.update_cultural_knowledge in Culture not working with async db: Fix async CultureManager functions to use correct functions for async db.highlights from ExaTools: Highlights was removed as a parameter from Exa from v2.0.0 onwards. For newer versionsupdate_cultural_knowledge with async db by @kausmeows in https://github.com/agno-agi/agno/pull/5284Full Changelog: https://github.com/agno-agi/agno/compare/v2.2.6...v2.2.7
Conversational Workflows: Workflows can now have a chat like experience similar to Agent/Team. See more in our docs
openai:gpt-5FileTools toolkit. You can now use it to read large files in chunks, perform partial updates to large files, and delete files.RunOutput and RunCompleted event for Agents and TeamRunOutput and TeamRunCompleted event for teamsFull Changelog: https://github.com/agno-agi/agno/compare/v2.2.5...v2.2.6
Handle non built-in routers when reprovisioning the AgentOS: When using a lifespan function to update and reprovision the AgentOS, support all previou
Full Changelog: https://github.com/agno-agi/agno/compare/v2.2.4...v2.2.5
Granular control of history messages: Add num_history_messages to control how many messages to consider when getting them from history.
num_history_messages to control how many messages to consider when getting them from history.OllamaEmbedder not respecting custom dimensions parameter: Fixed OllamaEmbedder to properly pass the dimensions parameter to the Ollama API during embedding generation.: in its name.Full Changelog: https://github.com/agno-agi/agno/compare/v2.2.3...v2.2.4
Support for Async MongoDB: You can now use your MongoDB databases asynchronously, with the AsyncMongoDb class. See the docs.
AsyncMongoDb class. See the docs.Full Changelog: https://github.com/agno-agi/agno/compare/v2.2.2...v2.2.3
LLM Response Caching: You can now cache model responses by setting cache_response=True on the model class. Useful to reduce costs and speed-up develop
cache_response=True on the model class. Useful to reduce costs and speed-up development and testing scenarios. See the docs for more details.AsyncSqliteDb class. See the docs.TavilyReader for knowledge base integration and TavilyTools for enhanced URL content extraction with full async support.summary_request_message on SessionSummaryManager: This allows you to override the user instruction given to the LLM when generating session summaries.model, reasoning_model, parser_model, and output_model.MCPTools and MultiMCPTools to your Agent or Team and the connection lifecycle will be handled automatically. This does mean it will connect and reconnect on each run. See the docs for best practices.refresh_connection on MCPTools and MultiMCPTools that allows connections to be refreshed. See the docs.JsonDb or GCSJsonDb database implementationssend_media_to_model flag set on Team to its member agents..run and .arun properlysend_media_to_model from Team to its members by @kausmeows in https://github.com/agno-agi/agno/pull/5199Full Changelog: https://github.com/agno-agi/agno/compare/v2.2.1...v2.2.2
PPTX Reader: Added support for reading Microsoft PowerPoint (.pptx) files by implementing a dedicated PPTXReader class.
max_tool_calls_from_history. Helps manage context size and reduce tokens. See the Docs for more information.Full Changelog: https://github.com/agno-agi/agno/compare/v2.2.0...v2.2.1
stream_intermediate_steps has been deprecated in favour of stream_events
.arun method of Workflow now returns an async iterator, similar to Agent and Team. Examplein our docs.PostHookStarted, PostHookCompleted for both Agent and Team eventsSessionSummaryCreationStarted, SessionSummaryCreationCompleted for both Agent and Team eventsRunContentCompleted for both Agent and Team → This signals that the content stream has ended, which is useful for building UIs using the eventsget_run_output and get_last_run_output on Agent/Team to allow retrieval of runs from members. Therefore after a team has executed you can get the last run straight from the member agent.expected_output as a default parameter on the delegate_task_to_member tool call for Team. This is done to increase flexibility with teams and to avoid incorrect behaviour on the team leader.add_team_history_to_members to share team-level request/response history with team members.share_member_interactions work to make sharing interactions during a run more accurate.stream_events flag: Add a stream_events flag to Agent, Team, Workflow and all run methods to emit all events whens streaming a response.update_session_state and aupdate_session_state functions on Agent and Team to allow direct updating of the session state that is persisted in the DB.None values in Ollama metrics: Fix a problem calculating metrics with Ollama when certain values were None instead of 0.Workflow.arun(..., stream=True) now return an AsyncIterator, i.e. which means await is not required, but instead async for event in workflow.arun(...) is the correct syntax.RunContent events (the actual model content chunks) are now behind the stream_events flag and will only be yielded where stream_events=True . Previously tool-call events were yielded regardless.stream_intermediate_steps has been deprecated in favour of stream_eventsFull Changelog: https://github.com/agno-agi/agno/compare/v2.1.10...v2.2.0
[Experimental feature] Culture: You can now provide your agents with a shared space to think, write, and build on each other's ideas. This will give y
Full Changelog: https://github.com/agno-agi/agno/compare/v2.1.9...v2.1.10
`id` on Message: Add id field on the Message class that is used internally for messages sent to LLMs, but also for the list of messages on the RunOutp
id on Message: Add id field on the Message class that is used internally for messages sent to LLMs, but also for the list of messages on the RunOutput objects. This can be used for tracking messages in storage.session_state in Workflows: Provide access to the contextual session_state to the evaluator and selector functions of the Condition and Router Workflow Step classes.metadata on run and arun for teams.async_add_contents: Fixed missing reader attribute not being passed to forknowledge loader functionsid on Message by @dirkbrnd in https://github.com/agno-agi/agno/pull/5110session_state to condition/router evaluator function by @kausmeows in https://github.com/agno-agi/agno/pull/5107Full Changelog: https://github.com/agno-agi/agno/compare/v2.1.8...v2.1.9
Class-based executors in Workflows: Enables to use a class-based executor by defining a class that implements the __call__ method. See docs
__call__ method. See docsGoogleCalenderTools to Notify attendees when creating,updating or deleting event in calendarsearch_vectors to search_knowledgeJiraTools toolkit to enable creating worklogs.acontinue_run() or acontinue_run_stream() methods with only run_id parameter and run_response=None. The logging statements were trying to access run_response.run_id when run_response could be None.Full Changelog: https://github.com/agno-agi/agno/compare/v2.1.7...v2.1.8
Memory with async databases: Fixed an error updating user memories when using an async database implementation, like PostgresDb.
PostgresDb.Full Changelog: https://github.com/agno-agi/agno/compare/v2.1.6...v2.1.7
Support for SurrealDB: Add complete support to use a SurrealDB database with your Agents, Teams and Workflows, via the new SurrealDb class.
SurrealDb class.store_tool_results flag:
store_tool_messages for better claritystore_history_messages and store_tool_messages by @kausmeows in https://github.com/agno-agi/agno/pull/5053Full Changelog: https://github.com/agno-agi/agno/compare/v2.1.5...v2.1.6
Async Postgres (docs): Added support for async Postgres all across the library.
store_history_messages and store_tool_messages to control whether these extra messages should be stored or not.stream_executor_events to filter workflow executor events (docs):
stream_intermediate_events we control to switch on-off the workflow events like WorkflowStarted, StepStarted etc.stream_executor_events now similar to stream_member_events in Team.serve(access_log=True).GoogleSheetsTools./mcp segment.updated_at field is now preserved when migrating sessions and memories.pymongo.stream_executor_events to filter workflow executor events by @kausmeows in https://github.com/agno-agi/agno/pull/5038get_run_input logic by @kausmeows in https://github.com/agno-agi/agno/pull/5035Full Changelog: https://github.com/agno-agi/agno/compare/v2.1.4...v2.1.5
Workflow History: Support to add workflow history to all or individual steps for a better continuous conversational experience
GoogleDriveTools class, to provide your Agents with read and write access to Google Drive.agno/tools/function.py used str() to serialize dictionary arguments for cache key generation. Since str() preserves dictionary key order, identical data with different key orders produced different cache keys. Replaced str() with json.dumps(sort_keys=True) to ensure serialization regardless of dictionary key orderFull Changelog: https://github.com/agno-agi/agno/compare/v2.1.3...v2.1.4
Claude Vertex AI: Implemented a new Claude model class to be served through Vertex AI.
enable_mcp_server parameter.Full Changelog: https://github.com/agno-agi/agno/compare/v2.1.2...v2.1.3
fix: h11 0.14.0 has a known CVE vulnerability by @fatelei in https://github.com/agno-agi/agno/pull/4920
A2A interface for the AgentOS../script to run MCP servers with your MCPTools or MultiMCPTools classes.user_id in the forwarded_props field when initiating a run.user_id.rename_session method in the SQLite DB implementation./sessions api for workflows with custom function step: the case of a custom function step being the first step was not being handled.session_state is always passed as a dictionary (never None) to custom functions during parallel execution, preventing parameter errors when session_state is required.Full Changelog: https://github.com/agno-agi/agno/compare/v2.1.1...v2.1.2
AgentOS Interface Route Prefix: Added the option to specify a prefix for AGUI, Whatsapp, and Slack interfaces. This enables running multiple interface
prefix for AGUI, Whatsapp, and Slack interfaces. This enables running multiple interfaces on the same AgentOS instance.WebSocketHandler.handle_event() and handle_dict() methods in types.py failed to serialize datetime/date objects when sending workflow events via WebSocket, causing "Object of type datetime is not JSON serializable" errors.Full Changelog: https://github.com/agno-agi/agno/compare/v2.1.0...v2.1.1
None of the below deprecated parameters have been removed, only new attributes added. It will likely be removed in v2.3.0 and after.
fastapi_app and added base_app → For configuring a custom FastAPI app for the AgentOS.replace_routes and added on_route_conflict → For how to handle conflicting routes when you set base_app.enable_mcp and added enable_mcp_server → For converting the AgentOS into an MCP server.os_id and added id → The ID of the OS.async and yield content or events), they are now executed in parallel.Team.arun(..., stream=True)) is now also concurrent.
use_batch parameter from PGVectorDB classfiles in the async pipeline input creation, enhancing the flexibility of input handling.InMemoryDb mutable references: Fixes a bug where some DB methods unexpectedly returned shared object references.tool_choice bug on Gemini: Agno now convert tool_choice values for none and auto correctly to what Gemini expects.to_json method was being overridden in the BaseWorkflowRunOutputEvent class which is not really required.search to relevant ones by @Mustafa-Esoofally in https://github.com/agno-agi/agno/pull/4863Full Changelog: https://github.com/agno-agi/agno/compare/v2.0.11...v2.1.0
LiteLLM: Add metadata and additional fields to LiteLLM Model class
user_id not being passed when creating RunOutput for response of agent.run() by @nebulaanish in https://github.com/agno-agi/agno/pull/4765Full Changelog: https://github.com/agno-agi/agno/compare/v2.0.10...v2.0.11
Session State: Adds a new flag overwrite_db_session_state to overwrite the session state persisted in the database
overwrite_db_session_state to overwrite the session state persisted in the databaseFull Changelog: https://github.com/agno-agi/agno/compare/v2.0.9...v2.0.10
MCP Toolbox: new Toolkit to provide your Agents with access to Google’s MCP Toolbox for Databases.
session_state and dependencies parametersCONTRIBUTING.md by @b41sh in https://github.com/agno-agi/agno/pull/4752Full Changelog: https://github.com/agno-agi/agno/compare/v2.0.8...v2.0.9
CometAPI: Added support for CometAPI as a model provider.
dependencies are now available on custom tools as a built-in argument. See an example.add_contents() in Knowledge.allow_partial_failure flag to MultiMCPTools toolkittool_call_limit can be passed to the default reasoning agent, avoiding potential costly tool call loops.run_input not being populated in session runs endpoint: The /sessions/<session_id>/runs endpoint does not return run_input in case of error as it references it from step_executor_runs and its empty if the workflow hasn't run yet, so in this case we fallback to directly get it from input.agent.run.OpenAIResponses to remove duplicate messages for reasoning models:
OpenAIResponses API has the (default enabled) feature that they store the conversation server-side and we send a previous_response_id with any followup questions.OpenAIResponses class implements this feature by looking through the message history, finding the most recent assistant response, and using that message ID as previous_response_id. This part works. However, we were then not deleting the prior messages from the request it sends to OpenAI. As a result OpenAI concatenates its own message history with the message history sent by Agno, resulting in an exponential number of repeated messages as the conversation continues.Full Changelog: https://github.com/agno-agi/agno/compare/v2.0.7...v2.0.8
LlamaCpp Model: Added a new LlamaCpp Model class for Llama CPP support!
LlamaCpp Model class for Llama CPP support!input_schema in UI breaking workflows: For some workflow configs where the steps are not wrapped in Step class, the UI does not show the input_schema form for them. This has been resolved.send_media_to_model flag for the case where the input image to agent is to be accessed by a tool and should not be available to the model.send_media_to_model=False we now exclude these results when sending messages to the model.Full Changelog: https://github.com/agno-agi/agno/compare/v2.0.6...v2.0.7
Nexus Model Class: Adds a Model class for Nexus Router
FileGenerationTools to give agents the ability to generate file artifacts! PDFs, CSVs, JSON and TXT. See the docs for more information.WorkflowRunOutput and correspondingly in DB.ChromaDb.search_session_history should be user specific: fix on ****enabling search_session_history=True we are not fetching user specific sessions.get_session() by @guyrosin in https://github.com/agno-agi/agno/pull/4625Full Changelog: https://github.com/agno-agi/agno/compare/v2.0.5...v2.0.6
AgentOS Home: Added route for / to avoid 404 errors.
/ to avoid 404 errors.replace_routes=False .Full Changelog: https://github.com/agno-agi/agno/compare/v2.0.4...v2.0.5
`TypedDict` support in `input_schema` for agent/team/workflow with structured input: Along with Pydantic support, TypedDict support is also available.
TypedDict support in input_schema for agent/team/workflow with structured input: Along with Pydantic support, TypedDict support is also available.
Added SiliconFlow : New model class added for Silicon Flow.
session_state as param in custom function step for workflow:
You can now do-
# on the function of the custom python function step for a workflow
def custom_function_step(step_input: StepInput, session_state):
session_state["test"] = test_1 # updates the workflow session state
extra_body on OpenAI: Added extra_body parameter on OpenAIChat and OpenAILike.ag infra up and ag infra down when run outside of the project.input_schema by @kausmeows in https://github.com/agno-agi/agno/pull/4534Full Changelog: https://github.com/agno-agi/agno/compare/v2.0.3...v2.0.4
WorkflowTools: Added WorkflowTools as a tool for agents and teams that allows the agent/team to run a workflow. Also has reasoning support for think/a
WorkflowTools as a tool for agents and teams that allows the agent/team to run a workflow. Also has reasoning support for think/analyze.MemoryTools as a tool for agents and teams to add, update, delete memories. Also has reasoning support for think/analyze.encoding to be passed on reader class.kwargs on runs: Added support for additional arguments on running an agent/team/workflow via AgentOS API.parallel_tool_calls in Cerebras model.store=False in OpenAIResponses class.slack/slack/events, it is now exposed on slack/eventsmemory_manager parameter to Teams.print_response_stream: In streaming case the structured output from parser model was not being displayed json formatted in cli.Full Changelog: https://github.com/agno-agi/agno/compare/v2.0.2...v2.0.3
chore: Release 2.0.2 by @ysolanky in https://github.com/agno-agi/agno/pull/4498
Full Changelog: https://github.com/agno-agi/agno/compare/v2.0.1...v2.0.2
fix: knowledge cookbooks by @willemcdejongh in https://github.com/agno-agi/agno/pull/4491
Full Changelog: https://github.com/agno-agi/agno/compare/v2.0.0...v2.0.1
Github Changelog: https://github.com/agno-agi/agno/compare/v1.8.2...v2.0.0
See the full changelog
Github Changelog: https://github.com/agno-agi/agno/compare/v1.8.2...v2.0.0
Nothing published for this version
Nothing published for this version
Nothing published for this version
Aws Claude Model Class: Fixed a regression caused in 1.8.3 leading to tool call failure for AWS Claude Model Class
Full Changelog: https://github.com/agno-agi/agno/compare/v1.8.3...v1.8.4
Anthropic Claude: Removes legacy tool call formatting
Full Changelog: https://github.com/agno-agi/agno/compare/v1.8.2...v1.8.3
Structured Output for AG-UI: Added response_model support for AG-UI.
response_model support for AG-UI.response_model support for the DiscordClient.updated_at to AgentSessionResponse, TeamSessionResponse by @Zulko in https://github.com/agno-agi/agno/pull/3848Full Changelog: https://github.com/agno-agi/agno/compare/v1.8.1...v1.8.2
…latest e2b_code_interpreter SDK, replacing the deprecated Sandbox() initialization. No other functionality changed.
Neo4jTools to explore and manipulate graphs in Neo4j.timezone_identifier to Team to match Agent functionality.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.filters was not correctly applied on all documents when inserting into the knowledge base.Full Changelog: https://github.com/agno-agi/agno/compare/v1.8.0...v1.8.1
Memori ToolKit: Added `MemoriTools`, a ToolKit for Agents and Teams to use GibsonAI’s Memori.
MemoriTools, a ToolKit for Agents and Teams to use GibsonAI’s Memori.files=[] can be provided to Workflow.run or Workflow.arun.agentic_crawler to ScrapeGraphTools.DuckDuckGoTools to work with the ddgs package. This requires users to install ddgs instead of duckduckgo-search.Full Changelog: https://github.com/agno-agi/agno/compare/v1.7.12...v1.8.0
Youtube Transcripts: Removed deprecated implementation in YoutubeTools.
collaborate mode.verbosity parameter to OpenAIChat and OpenAIResponses..load() functioned incorrectly.YoutubeTools.Full Changelog: https://github.com/agno-agi/agno/compare/v1.7.11...v1.7.12
In Memory Storage: Added InMemoryStorage for a simpler storage option useful in rapid prototyping.
InMemoryStorage for a simpler storage option useful in rapid prototyping.TrafilaturaTools SDK for web scraping and text extraction using the Trafilatura library, with 15 comprehensive cookbook examples.DashScope integration.File for compatible AWS Bedrock models.Full Changelog: https://github.com/agno-agi/agno/compare/v1.7.10...v1.7.11
One column per month.