2 Releases published by 1 person

• v0.2.5

  published Aug 7, 2025

• v0.2.6

  published Aug 11, 2025

26 Pull requests merged by 14 people

• Realtime: one guardrail trip event per response

  #1458 merged Aug 14, 2025

• Update all translated document pages

  #1473 merged Aug 14, 2025

• docs: Clarify the is_enabled parameter in handoff() documentation

  #1472 merged Aug 14, 2025

• Update all translated document pages

  #1471 merged Aug 14, 2025

• Migrate document translation script to gpt-5

  #1470 merged Aug 14, 2025

• Update all translated document pages

  #1469 merged Aug 14, 2025

• docs: Add a comprehensive example for handling function tool errors.

  #1354 merged Aug 14, 2025

• Add input validation and type conversion for user input

  #1462 merged Aug 14, 2025

• Fix: Clarify random_number function docstring for inclusive range

  #1461 merged Aug 14, 2025

• Add a simple gpt-oss example

  #1440 merged Aug 14, 2025

• Fix #1407 Add reasoning.effort="minimal" and "verbosity" params to ModelSettings

  #1439 merged Aug 13, 2025

• Fix: enforce strict instructions function signature in get_system_prompt

  #1426 merged Aug 13, 2025

• feat(tests): add tests for unicode content, special characters, and concurrent access in SQLiteSession

  #1399 merged Aug 11, 2025

• Realtime: add handoff prompt prefix

  #1382 merged Aug 11, 2025

• Added runtime validation for Agent constructor arguments

  #998 merged Aug 11, 2025

• v0.2.6

  #1423 merged Aug 11, 2025

• chore(deps): Bump "openai>=1.99.6,<2"

  #1415 merged Aug 11, 2025

• Updated the default values for two parameters in the ModelSettings class to align with their documented behavior:

  #1392 merged Aug 11, 2025

• Add streaming function call arguments example

  #1052 merged Aug 11, 2025

• Add logprobs to ModelSettings

  #971 merged Aug 11, 2025

• Update all translated document pages

  #1413 merged Aug 11, 2025

• Improve accuracy of guardrails and lifecycle hook descriptions

  #1386 merged Aug 11, 2025

• Enhanced FAQ lookup tool in examples/customer_service app

  #1063 merged Aug 11, 2025

• feat: support agent output guardrails in realtime

  #1381 merged Aug 8, 2025

• v0.2.5

  #1398 merged Aug 7, 2025

• Realtime: send session update before tool response

  #1388 merged Aug 7, 2025

7 Pull requests opened by 6 people

• Add first-class verbosity support for GPT-5

  #1403 opened Aug 8, 2025

• Implement bidirectional handoff system (#1376)

  #1410 opened Aug 8, 2025

• Add an example using Ollama Turbo

  #1460 opened Aug 13, 2025

• Fix a bug where this SDK's internals fail to handle items restored using Pydantic etc.

  #1467 opened Aug 14, 2025

• Feat: Add Optional Structured Session Storage

  #1474 opened Aug 14, 2025

• Fix: Clarify random_number function docstring for inclusive range

  #1475 opened Aug 14, 2025

• Add input validation and type conversion for user input

  #1476 opened Aug 14, 2025

54 Issues closed by 13 people

• Add `timeout` feature for agent runs

  #1425 closed Aug 14, 2025

• Error code: 400 "No tool output found for function call"

  #673 closed Aug 14, 2025

• Very high response times at random during hand-offs

  #717 closed Aug 14, 2025

• Add reasoning.effort="minimal" and "verbosity" params to ModelSettings

  #1407 closed Aug 13, 2025

• Engineering review request — persistent emergent optimization layer ("NEO Core")

  #1441 closed Aug 13, 2025

• Gemini 2.5 does not work well with function call

  #1420 closed Aug 13, 2025

• Built-in Support for Sticky Session Routing / Direct Agent Handoff

  #1436 closed Aug 13, 2025

• Processing image/ multi modal responses in function tool results?

  #787 closed Aug 13, 2025

• Prompt the handed model to begin a structured conversation.

  #1428 closed Aug 13, 2025

• Agent keep retrying until hit max_retires on function calling

  #981 closed Aug 12, 2025

• Return

  #259 closed Aug 12, 2025

• Support for OpenAI Realtime Models in Agents SDK

  #239 closed Aug 12, 2025

• Agents SDK v0.0.4: Local Traces Missing Critical LLM Outputs and Tool Interaction Data

  #222 closed Aug 12, 2025

• Tracing for the realtime api

  #220 closed Aug 12, 2025

• Clarification on the need for RunContextWrapper[UserContext] and Agent[UserContext]

  #219 closed Aug 12, 2025

• Discrepancy in Documentation: strict_mode Parameter Mentioned but Not Implemented in function_tool Decorator

  #215 closed Aug 12, 2025

• 404 Error During Agent Handoff in Multi-Agent System

  #206 closed Aug 12, 2025

• Got this error with example file {'error': {'message': 'The model `computer-use-preview-2025-03-11` does not exist or you do not have access to it.', 'type': 'invalid_request_error', 'param': None, 'code': 'model_not_found'}}

  #183 closed Aug 12, 2025

• How to use the Custom Web Search tool

  #170 closed Aug 12, 2025

• collective Multi-agent reinforcement learning

  #160 closed Aug 12, 2025

• The output panel shows "Request failed" after the task is completed

  #157 closed Aug 12, 2025

• Are there some other way to set up the api token besides setting it as an environment variable?

  #154 closed Aug 12, 2025

• Is it possible to use a locally privatized model?

  #136 closed Aug 12, 2025

• Hosted tool 'computer_use_preview' is not supported

  #132 closed Aug 12, 2025

• from LangChain and Langgraph to openai-agents

  #128 closed Aug 12, 2025

• Changing the src/agents/models/openai_chatcompletions.py into smaller parts

  #119 closed Aug 12, 2025

• use localmodel with openai format does not work

  #118 closed Aug 12, 2025

• Proper way of managing large context window for ComputerTool

  #111 closed Aug 12, 2025

• Support for Gemini Flash

  #106 closed Aug 12, 2025

• Allow agent to return logprobs

  #715 closed Aug 12, 2025

• Agent session error in OpenAI Agents SDK when handling tool output

  #1429 closed Aug 12, 2025

• No validation on the Agent constructor arguments

  #996 closed Aug 11, 2025

• Latest `openai` release (`1.99.2`) breaks `LitellmModel`.

  #1395 closed Aug 11, 2025

• ChatCompletionMessageToolCallParam can't not be converted with openai>=1.99.0

  #1401 closed Aug 11, 2025

• `TypeError: Cannot instantiate typing.Union` when running with latest `openai` Python package

  #1417 closed Aug 11, 2025

• TypeError: Cannot instantiate typing.Union when calling tools due to misuse of TypeAlias in chatcmpl_converter.py

  #1416 closed Aug 11, 2025

• Import error with openai package 1.99.0

  #1378 closed Aug 11, 2025

• Reasoning summary output items from Azure OpenAI models

  #1389 closed Aug 11, 2025

• How to use non-async agents when tools respond synchronously (no /responses endpoint)?

  #1391 closed Aug 11, 2025

• clear the memory of agents between handoffs

  #1244 closed Aug 11, 2025

• How do I obtain the assistant and input audio transcription events using RealtimeAgent?

  #1380 closed Aug 11, 2025

• LiteLLM: extra_body with guardrails is not taken into account

  #1324 closed Aug 11, 2025

• Discussion: Folder and files structure for context session management

  #1365 closed Aug 11, 2025

• agents.exceptions.UserError: Unhandled item type or structure when use reasoning model with tool call.

  #1004 closed Aug 11, 2025

• Will an agent call a tool multiple times?

  #1394 closed Aug 10, 2025

• TypeError: Cannot instantiate typing.Union

  #1411 closed Aug 10, 2025

• agent stop when no tool called

  #1279 closed Aug 10, 2025

• Best way to store structured conversation context

  #1299 closed Aug 10, 2025

• include_usage in ModelSettings class

  #1311 closed Aug 10, 2025

• How to show Metadata/Sessions/Tags when using Litellm version sdk

  #1210 closed Aug 9, 2025

• Agent-level MCP server tool filtering

  #1234 closed Aug 9, 2025

• How to Retrieve Tool Name from `tool_call_item` Event in Streaming?

  #1264 closed Aug 9, 2025

• How to handle files

  #1281 closed Aug 9, 2025

• Is it possible to silence `OPENAI_API_KEY is not set, skipping trace export`, without silencing third-party tracers?

  #1387 closed Aug 7, 2025

22 Issues opened by 17 people

• How to disable default OpenAI tracer and use only Datadog ddtrace for tracing

  #1477 opened Aug 14, 2025

• openai.BadRequestError with gpt-oss-120b - Expected 2 output messages (reasoning and final), but got 3.

  #1468 opened Aug 14, 2025

• RealtimeAgent Error(message='Conversation already has an active response in progress: Wait until the response is finished')

  #1466 opened Aug 14, 2025

• Tool not invoked on consecutive calls when using SQLiteSession in OpenAI Agent SDK

  #1465 opened Aug 14, 2025

• How can you add session (memory) to realtime agents

  #1464 opened Aug 14, 2025

• Add Flag for Priority Processing Tier (or Tier in general)

  #1459 opened Aug 13, 2025

• `RunResult` missing `.usage` attribute in latest OpenAI Agents SDK

  #1457 opened Aug 13, 2025

• `UserError` not raised when invalid tool type is passed to `tools` parameter

  #1443 opened Aug 13, 2025

• agents.exceptions.MaxTurnsExceeded with Llama 3.3

  #1442 opened Aug 13, 2025

• Fail to load traces when using a different API key for tracing

  #1437 opened Aug 13, 2025

• Model parameter compatibility documentation

  #1435 opened Aug 13, 2025

• in history_updated events, the status of 'assistant' is always 'in_progress'. After the assistant has spoken or generated audio, it must say 'Completed'

  #1434 opened Aug 13, 2025

• Unexpected behavior with MCP server tools

  #1430 opened Aug 12, 2025

• RuntimeError: aclose(): asynchronous generator is already running w/ realtime/app example

  #1427 opened Aug 12, 2025

• Eval integration

  #1418 opened Aug 11, 2025

• Cannot use outputType with gpt-oss:20b

  #1414 opened Aug 11, 2025

• LLM-friendly Documentation

  #1412 opened Aug 11, 2025

• gpt-5 model fails to complete my tasks while gpt-4.1 completes

  #1409 opened Aug 8, 2025

• How to track finish_reason when using the Runner class

  #1406 opened Aug 8, 2025

• if mcp initialize fails, the streaming hangs

  #1404 opened Aug 8, 2025

• LiteLLM calls not working because of improper imports

  #1402 opened Aug 8, 2025

• GPT-5 does not work well compared to 4.1

  #1397 opened Aug 7, 2025

33 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• Initial pass at PostgreSQL session management

  #1163 commented on Aug 10, 2025 • 4 new comments

• Proposal: Return results to the parent Agent

  #1376 commented on Aug 7, 2025 • 0 new comments

• No OpenAI tracing with Azure Foundry models

  #1051 commented on Aug 13, 2025 • 0 new comments

• from agents.extensions.models.litellm_model import LitellmModel

  #666 commented on Aug 14, 2025 • 0 new comments

• FileSearchTool runs despite InputGuardrailTripwireTriggered

  #889 commented on Aug 14, 2025 • 0 new comments

• Human-In-The-Loop Architecture should be implemented on top priority!

  #636 commented on Aug 14, 2025 • 0 new comments

• Ensure that input_guardrails can block tools from running

  #931 commented on Aug 11, 2025 • 0 new comments

• feat(agents): Add on_llm_start and on_llm_end Lifecycle Hooks

  #987 commented on Aug 14, 2025 • 0 new comments

• Add run configuration inheritance for agent-as-tool workflows

  #1006 commented on Aug 10, 2025 • 0 new comments

• feat: #864 support streaming nested tool events in Agent.as_tool

  #1057 commented on Aug 13, 2025 • 0 new comments

• Feature: Make trace_include_sensitive_data configurable via environment variable

  #1192 commented on Aug 9, 2025 • 0 new comments

• Add conditional tool enabling feature to agent as tool.

  #1193 commented on Aug 12, 2025 • 0 new comments

• Enh - Use handoff_description as Default Fallback for Tool Description in as_tool() of agents

  #1199 commented on Aug 14, 2025 • 0 new comments

• Allow to pass both session and input list

  #1298 commented on Aug 13, 2025 • 0 new comments

• Fix: Emit tool_called events immediately in streaming runs

  #1300 commented on Aug 13, 2025 • 0 new comments

• Add SQLAlchemy session backend for conversation history management

  #1357 commented on Aug 13, 2025 • 0 new comments

• Mongodb memory session

  #1364 commented on Aug 11, 2025 • 0 new comments

• SQLAlchemy Session Store for Production Databases

  #1328 commented on Aug 10, 2025 • 0 new comments

• MongoDB Session Store

  #1362 commented on Aug 11, 2025 • 0 new comments

• Handoffs to deterministic workflows

  #1274 commented on Aug 11, 2025 • 0 new comments

• RunContextWrapper in MCP server tools

  #1383 commented on Aug 11, 2025 • 0 new comments

• Does gpt-oss support Agents SDK tools?

  #1379 commented on Aug 11, 2025 • 0 new comments

• MCPServerStreamableHttp connection error corrupt next async calls

  #1288 commented on Aug 11, 2025 • 0 new comments

• Issue with output_type and Custom LLM Providers in openai-agents-python

  #1390 commented on Aug 11, 2025 • 0 new comments

• how to cleanup mcp servers correctly?

  #1393 commented on Aug 11, 2025 • 0 new comments

• will there be a Java version of openai-agents-java in the future?

  #115 commented on Aug 11, 2025 • 0 new comments

• Suggestion: Database schema for conversations could be better

  #1385 commented on Aug 11, 2025 • 0 new comments

• Tool call parsed as regular message or raises errors (litellm+Mistral)

  #1196 commented on Aug 12, 2025 • 0 new comments

• Why have agents (or GPT 4.1) slowed down so much?

  #1249 commented on Aug 12, 2025 • 0 new comments

• BadRequestError while trying out input guardrails w/ Groq

  #129 commented on Aug 12, 2025 • 0 new comments

• Run `custom_output_extractor` AFTER `RunHooks.on_tool_end`

  #1375 commented on Aug 13, 2025 • 0 new comments

• Figuring out how to use streamable http mcp server

  #1316 commented on Aug 13, 2025 • 0 new comments

• Tool calling with LiteLLM and thinking models fail

  #765 commented on Aug 13, 2025 • 0 new comments