agents

agents copied to clipboard

### Feature Type Nice to have ### Feature Description When using the `conversation_item_added` event in console mode to print custom output (e.g., Rich panels), assistant messages render incorrectly because they...

MSameerAbbas

Add minimum confidence level filter for deepgram STT.

5

Updated to send out stt results for deepgram that has a minimum amount of confidence to avoid transcriptions generated without user actually speaking.

meetakshay99

KeyError in Room._on_room_event for local_track_published (track SID missing from local_participant.track_publications)

2

### Bug Description KeyError logged and bubbles through _listen_task’s error logging error running user callback for local_track_published: room_handle: 6 local_track_published { track_sid: "TR_AMLxYxTEBo2L7c" } Traceback (most recent call last): File...

MonkeyLeeT

feat: add tool_calling_sound to BackgroundAudioPlayer

2

Add support for separate audio feedback during function tool execution: - Add FunctionToolsExecutingEvent to signal when tool execution starts - Add tool_calling_sound parameter to BackgroundAudioPlayer - Auto-switch from thinking_sound to...

AKomplished-bug

fix: resolve audio dropout after AgentTask completion

2

Fixed critical bugs causing agent to go silent after nested workflow agents (AgentTasks) complete: 1. Audio recognition not recreated on resume: - When activity is paused for AgentTask handover, _close_session()...

tomc98

preemptive_generation=True causes duplicate LLM requests and doubled token costs

1

### Bug Description When `preemptive_generation=True` is enabled in `AgentSession`, two separate LLM requests are made per user turn, both completing successfully (both with `cancelled=False`), resulting in **doubled token consumption and...

javguitor

OpenAI tools (`WebSearch`, `FileSearch`, `ComputerUse`) to follow soon after. To use: `llm=openai.responses.LLM()`

tinalenguyen

### Bug Description The VideoAvatar logic assumes that a single AgentSession is tied to one video avatar. This design prevents assigning separate avatars to different agents within the same session,...

Viktoriagrg

fix(llm): sanitize control characters in function call JSON arguments

1

## Description ### Problem LLMs sometimes generate function call JSON with literal control characters (e.g., newlines, tabs) inside string values. For example: ```json {"prompt": "A timeline showing: - Event 1...

ArpitKotecha

feat: add Azure AI Foundry support for Claude models

4

- Add AzureFoundryModels type definitions - Implement with_azure_foundry() static method - Support Azure endpoint authentication - Addresses issue #4071

satheeshKOLA532