index.yaml

version: '1.0'
description: Index of all LiveKit Agent examples with metadata
total_examples: 77
examples:
- file_path: complex-agents/avatars/hedra/dynamically_created_avatar/agent.py
  title: Dynamically Created Avatar
  category: avatars
  tags:
  - avatar
  - openai
  - deepgram
  difficulty: intermediate
  description: Shows how to create an avatar dynamically in an agent.
  demonstrates:
  - Making parallel calls to an LLM while speaking to the voice agent
  - Creating an avatar dynamically in an agent
  - Creating a new hedra avatar session using a custom image
  - A function tool that uses a docstring to describe the tool for the LLM to use
- file_path: complex-agents/avatars/hedra/education_avatar/agent.py
  title: Education Avatar
  category: avatars
  tags:
  - avatar
  - openai
  - deepgram
  - hedra
  difficulty: advanced
  description: Shows how to create an avatar that can help a user learn about the Fall of the Roman Empire using flash cards
    and quizzes.
  demonstrates:
  - Creating a new hedra avatar session using a custom image
  - Using RPC to send messages to the client for flash cards and quizzes using `perform_rpc`
  - Using `register_rpc_method` to register the RPC methods so that the agent can receive messages from the client
  - Using UserData to store state for the cards and the quizzes
  - Using custom data classes to represent the flash cards and quizzes
- file_path: complex-agents/avatars/hedra/pipeline_avatar/agent.py
  title: Hedra Avatar with Pipeline
  category: avatars
  tags:
  - hedra
  - avatar
  - static_image
  - pipeline
  - inworld_tts
  - multilingual
  difficulty: intermediate
  description: Visual avatar using Hedra with static image, pipeline architecture, and Inworld TTS
  demonstrates:
  - Hedra avatar session with static image loading
  - Pipeline architecture with separate STT/LLM/TTS components
  - Inworld TTS voice integration
  - Multilingual turn detection model
  - Noise cancellation with BVC
- file_path: complex-agents/avatars/hedra/realtime_avatar/agent.py
  title: Hedra Avatar with Realtime
  category: avatars
  tags:
  - hedra
  - avatar
  - static_image
  - openai_realtime
  difficulty: intermediate
  description: Visual avatar using Hedra with OpenAI Realtime model integration
  demonstrates:
  - Hedra avatar session with dynamic image selection
  - OpenAI Realtime model for low-latency conversation
  - Minimal agent configuration with realtime model
  - Avatar participant identity management
- file_path: complex-agents/avatars/tavus/tavus.py
  title: Tavus Avatar
  category: avatars
  tags:
  - avatar
  - openai
  - deepgram
  - tavus
  difficulty: intermediate
  description: Shows how to create a tavus avatar that can help a user learn about the Fall of the Roman Empire using flash
    cards and quizzes.
  demonstrates:
  - Creating a new tavus avatar session
  - Using RPC to send messages to the client for flash cards and quizzes using `perform_rpc`
  - Using `register_rpc_method` to register the RPC methods so that the agent can receive messages from the client
  - Using UserData to store state for the cards and the quizzes
  - Using custom data classes to represent the flash cards and quizzes
- file_path: change_agent_instructions/page.mdoc
  title: Change Agent Instructions
  category: basics
  tags:
  - instructions
  - assemblyai
  - openai
  - cartesia
  difficulty: beginner
  description: Shows how to change the instructions of an agent at runtime.
  demonstrates:
  - Changing agent instructions after the agent has started using update_instructions()
  - Conditional logic based on participant attributes
- file_path: context_variables/page.mdoc
  title: Context Variables
  category: basics
  tags:
  - context
  - variables
  - assemblyai
  - openai
  - cartesia
  difficulty: beginner
  description: Shows how to give an agent context about the user using simple variables.
  demonstrates:
  - Using context variables from a simple dictionary
  - Formatting instructions with user-specific data
- file_path: label_messages/page.mdoc
  title: Conversation Event Monitoring Agent
  category: basics
  tags:
  - events
  - conversation-monitoring
  - logging
  - deepgram
  - openai
  difficulty: beginner
  description: Shows how to monitor and log conversation events as they occur, useful for debugging and understanding agent-user
    interactions.
  demonstrates:
  - Conversation event handling and logging
- file_path: echo_transcriber_agent/page.mdoc
  title: Echo Transcriber Agent
  category: basics
  tags:
  - echo
  - transcriber
  - assemblyai
  - silero
  difficulty: beginner
  description: Shows how to create an agent that can transcribe audio and echo it back.
  demonstrates:
  - Transcribing audio
  - Echoing audio back that's stored in a buffer
  - Custom STT node processing
  - Custom VAD stream handling
- file_path: exit_message/page.mdoc
  title: Exit Message
  category: basics
  tags:
  - exit
  - message
  - assemblyai
  - openai
  - cartesia
  difficulty: beginner
  description: Shows how to use the on_exit method to take an action when the agent exits.
  demonstrates:
  - Using the on_exit method to take an action when the agent exits
  - Using function tools to end sessions gracefully
- file_path: say_in_voice/page.mdoc
  title: Function Tool Voice Switching Agent
  category: basics
  tags:
  - tts
  - voice-switching
  - function-tools
  - inworld
  - assemblyai
  - openai
  difficulty: beginner
  description: Demonstrates how to create an agent that can dynamically switch between different voices during a conversation
    using function tools.
  demonstrates:
  - Dynamic TTS voice switching
  - Function tool integration
  - Multiple TTS provider support (Inworld)
- file_path: listen_and_respond/page.mdoc
  title: Listen and Respond
  category: basics
  tags:
  - basics
  - assemblyai
  - openai
  - cartesia
  difficulty: beginner
  description: Shows how to create an agent that can listen to the user and respond.
  demonstrates:
  - This is the most basic agent that can listen to the user and respond. This is a good starting point for any agent.
- file_path: playing_audio/page.mdoc
  title: Playing Audio
  category: basics
  tags:
  - audio
  - assemblyai
  - openai
  - cartesia
  difficulty: beginner
  description: Shows how to play audio from a file in an agent.
  demonstrates:
  - Playing audio from a file
- file_path: repeater/page.mdoc
  title: Repeater
  category: basics
  tags:
  - repeater
  - assemblyai
  - openai
  - cartesia
  difficulty: beginner
  description: Shows how to create an agent that can repeat what the user says.
  demonstrates:
  - Using the `on_user_input_transcribed` event to listen to the user's input
  - Using the `say` method to respond to the user with the same input
- file_path: tool_calling/page.mdoc
  title: Tool Calling
  category: basics
  tags:
  - tool-calling
  - assemblyai
  - openai
  - cartesia
  difficulty: beginner
  description: Shows how to use tool calling in an agent.
  demonstrates:
  - Using the most basic form of tool calling in an agent to print to the console
- file_path: uninterruptable/page.mdoc
  title: Uninterruptable Agent
  category: basics
  tags:
  - interruptions
  - allow_interruptions
  - agent_configuration
  - assemblyai
  - openai
  - cartesia
  difficulty: beginner
  description: Agent configured to complete responses without user interruptions
  demonstrates:
  - Setting allow_interruptions=False in agent configuration
  - Testing interruption handling behavior
  - Agent-initiated conversation with on_enter
- file_path: complex-agents/role-playing/agents/base_agent.py
  title: Base Game Agent
  category: complex-agents
  tags:
  - rpg
  - game-state
  - agent-switching
  - context-preservation
  - rpc-communication
  difficulty: advanced
  description: Base class for RPG game agents with context preservation and state management
  demonstrates:
  - Agent inheritance patterns for game systems
  - Chat context preservation across agent switches
  - Game state integration and management
  - RPC communication for client updates
  - Agent lifecycle management with session data
  - Context truncation and memory management
- file_path: complex-agents/role-playing/agents/combat_agent.py
  title: Combat Agent
  category: complex-agents
  tags:
  - rpg
  - combat-system
  - turn-based-combat
  - npc-ai
  - function-tools
  difficulty: advanced
  description: Specialized agent for handling turn-based combat encounters in RPG games
  demonstrates:
  - Turn-based combat management
  - Combat action queueing and processing
  - NPC AI for automated combat turns
  - Real-time combat state updates via RPC
  - Experience and loot distribution
  - Dynamic combat flow with player/NPC interactions
  - Combat action validation and execution
- file_path: complex-agents/role-playing/agent.py
  title: D&D Role-Playing Game
  category: complex-agents
  tags:
  - rpg
  - game_state
  - rpc_methods
  - item_generation
  - combat_system
  - npc_interaction
  difficulty: advanced
  description: Dungeons & Dragons role-playing game with narrator and combat agents
  demonstrates:
  - Complex game state management
  - Multiple RPC methods for game queries
  - Dynamic NPC and item generation
  - Combat system with initiative tracking
  - Character creation and stats management
  - Inventory and equipment system
  - Voice acting for different NPCs
- file_path: complex-agents/doheny-surf-desk/agent.py
  title: Doheny Surf Desk Booking Agent
  category: complex-agents
  tags:
  - multi_agent
  - tasks
  - task_groups
  - observer_pattern
  - guardrails
  - llm_evaluation
  - context_injection
  - phone_receptionist
  - booking_system
  - telephony
  difficulty: advanced
  description: Phone receptionist agent for surf school bookings with background observer and task groups - great example
    for building phone-based booking systems
  requires: livekit-agents>=1.3.0
  demonstrates:
  - Complete phone receptionist workflow for appointment booking (you need to connect your calendar) and payment processing
    (demo, not real payment processing)
  - Background observer agent that monitors conversation transcripts in real-time
  - LLM-based evaluation every 3 user turns to detect safety issues (minors, injuries, weather concerns, skill mismatches,
    VIP customers)
  - Context injection pattern - observer injects system messages into active agent's chat context as guardrails
  - Task groups for sequential task execution with structured return values (profile collection, payment processing)
  - Multi-agent workflow with 5 specialized agents (IntakeAgent, SchedulerAgent, GearAgent, BillingAgent, FrontDeskAgent)
  - Observer can improve conversation quality by monitoring outputs in background
- file_path: complex-agents/exa-deep-researcher/agent.py
  title: EXA Deep Researcher
  category: complex-agents
  tags:
  - exa
  - research
  - voice_controlled
  - background_jobs
  - rpc_streaming
  difficulty: advanced
  description: Voice-controlled deep research agent using EXA for web intelligence
  demonstrates:
  - Voice-only control
  - Single background research job with state management
  - EXA API integration for search and content fetching
  - RPC streaming to UI for status, notes, and reports
  - Token-aware compression and concurrency control
  - Cited final reports saved to disk
- file_path: complex-agents/role-playing/core/game_state.py
  title: Game State Management
  category: complex-agents
  tags:
  - rpg
  - state-management
  - dataclass
  - session-data
  - type-safety
  difficulty: intermediate
  description: Centralized game state management for RPG sessions with type-safe data structures
  demonstrates:
  - Dataclass-based state management
  - Session data persistence across agent switches
  - Type-safe context handling with generics
  - Game progression tracking and history
  - Multi-agent state coordination
  - Combat state integration
- file_path: complex-agents/ivr-agent/agent.py
  title: IVR Phone System Navigator
  category: complex-agents
  tags:
  - ivr
  - dtmf
  - telephony
  - sip
  - participant_attributes
  - cooldown
  difficulty: advanced
  description: Agent that navigates phone IVR systems using DTMF codes
  demonstrates:
  - DTMF code transmission for phone navigation
  - SIP participant detection and handling
  - Task extraction from participant attributes
  - Cooldown mechanism for DTMF presses
  - Dynamic instruction updates based on task
  - Function tool for sending DTMF codes
- file_path: complex-agents/role-playing/generators/item_generator.py
  title: Item Generator
  category: complex-agents
  tags:
  - rpg
  - procedural-generation
  - llm-generation
  - yaml-configuration
  - item-creation
  difficulty: advanced
  description: AI-powered procedural item generation system for RPG games
  demonstrates:
  - LLM-driven content generation with structured prompts
  - YAML-based rule configuration for generation
  - Context-aware item creation based on NPC and location
  - Automated mechanical property assignment
  - Fallback systems for generation failures
  - JSON parsing and validation for generated content
- file_path: complex-agents/nova-sonic/form_agent.py
  title: Job Application Form Agent
  category: complex-agents
  tags:
  - aws_realtime
  - form_filling
  - rpc_frontend
  - interview
  - structured_data
  difficulty: advanced
  description: Interactive interview agent for job applications with AWS Realtime
  demonstrates:
  - AWS Realtime model integration
  - Structured form data collection
  - RPC communication with frontend for live updates
  - Multi-section interview process
  - Field validation and capitalization
  - Application status tracking
  - Frontend form highlighting and updates
- file_path: complex-agents/medical_office_triage/triage.py
  title: Medical Office Triage System
  category: complex-agents
  tags:
  - multi_agent
  - agent_transfer
  - medical
  - context_preservation
  - chat_history
  difficulty: advanced
  description: Multi-agent medical triage system with specialized departments
  demonstrates:
  - Multiple specialized agents (triage, support, billing)
  - Agent-to-agent transfer with context preservation
  - Chat history truncation and management
  - Shared userdata across agent transfers
  - Room attribute updates for agent tracking
  - YAML prompt loading for agent instructions
- file_path: complex-agents/role-playing/generators/npc_generator.py
  title: NPC Generator
  category: complex-agents
  tags:
  - rpg
  - procedural-generation
  - character-creation
  - personality-generation
  - dialogue-system
  difficulty: advanced
  description: AI-powered NPC generation system with personality, backstory, and dynamic dialogue
  demonstrates:
  - Dynamic NPC creation with context-aware traits
  - Parallel content generation for efficiency
  - YAML rule-based character archetypes
  - Personality and backstory generation
  - Dynamic dialogue creation and management
  - Integration with item generation systems
- file_path: complex-agents/role-playing/agents/narrator_agent.py
  title: Narrator Agent
  category: complex-agents
  tags:
  - rpg
  - storytelling
  - npc-interaction
  - voice-acting
  - exploration
  difficulty: advanced
  description: Main storytelling agent for RPG games with voice acting and world interaction
  demonstrates:
  - Dynamic storytelling and narrative flow
  - Multi-voice character portrayal with TTS switching
  - NPC interaction and dialogue management
  - World exploration and location transitions
  - Character creation and progression
  - Trading and inventory management
  - Skill check resolution and dice rolling
- file_path: complex-agents/note-taking-assistant/agent.py
  title: Note Taking Assistant
  category: complex-agents
  tags:
  - complex-agents
  - cerebras
  - deepgram
  difficulty: intermediate
  description: Shows how to use the Note Taking Assistant.
  demonstrates:
  - Using the Note Taking Assistant.
- file_path: complex-agents/nutrition-assistant/agent.py
  title: Nutrition Tracker Assistant
  category: complex-agents
  tags:
  - sqlite_database
  - nutrition
  - food_tracking
  - rpc_updates
  - thread_pool
  difficulty: advanced
  description: Nutrition tracking assistant with SQLite database and real-time updates
  demonstrates:
  - SQLite database for persistent food tracking
  - Thread pool executor for non-blocking database operations
  - Nutritional data calculation and aggregation
  - RPC updates to frontend for live nutrition display
  - Daily summaries and food consumption history
  - Fixed participant identity for testing
  - Initial data sync on connection
- file_path: complex-agents/personal_shopper/personal_shopper.py
  title: Personal Shopper Multi-Agent
  category: complex-agents
  tags:
  - customer_database
  - multi_agent_transfer
  - order_management
  - customer_identification
  difficulty: advanced
  description: E-commerce personal shopper with triage, sales, and returns departments
  demonstrates:
  - Customer identification and database management
  - Three specialized agents with different responsibilities
  - Order creation and management with pricing
  - Order history retrieval for returns
  - Context-aware agent transfers with personalization
  - Customer database with order tracking
  - Persistent customer data across sessions
- file_path: complex-agents/shopify-voice-shopper/shopify.py
  title: Shopify Voice Shopping Agent
  category: complex-agents
  tags:
  - mcp_server
  - shopify
  - dynamic_agent_switching
  - rpc_navigation
  - fast_llm_response
  difficulty: advanced
  description: Voice shopping assistant for Shopify stores with MCP server integration
  demonstrates:
  - MCP server integration for Shopify operations
  - Dynamic shop switching with agent updates
  - RPC navigation to frontend browser
  - Fast LLM for quick acknowledgments
  - Voice-optimized response formatting
  - Factory pattern for agent creation
  - Shop URL handling and validation
- file_path: complex-agents/teleprompter/cartesia-ink.py
  title: Teleprompter Transcription Agent
  category: complex-agents
  tags:
  - rpc_transcript
  - cartesia_stt
  - user_input_transcribed
  - frontend_communication
  difficulty: intermediate
  description: Real-time teleprompter that sends transcriptions to frontend via RPC
  demonstrates:
  - Cartesia STT for transcription
  - RPC method registration for status checks
  - Event handler for final transcriptions
  - Async RPC calls to send transcripts
  - Frontend communication pattern
  - Real-time transcript streaming
- file_path: complex-agents/turn-taking/agent.py
  title: Turn-Taking Detection Agent
  category: complex-agents
  tags:
  - eou_probability
  - turn_detection
  - gladia_stt
  - multilingual
  - rpc_eou_updates
  difficulty: advanced
  description: Agent that exposes end-of-utterance probability for turn-taking research
  demonstrates:
  - Custom turn detector wrapper exposing EOU probability
  - Real-time EOU probability RPC updates to frontend
  - Multilingual STT with Gladia
  - Percentage formatting for tiny probabilities
  - Session userdata for RPC communication
  - STT event processing for transcript logging
- file_path: complex-agents/vision/agent.py
  title: Vision-Enabled Agent
  category: complex-agents
  tags:
  - video_stream
  - grok_vision
  - x_ai
  - frame_capture
  - image_content
  difficulty: intermediate
  description: Agent with camera vision capabilities using Grok-2 Vision model
  demonstrates:
  - Video stream processing from remote participants
  - Frame buffering from video tracks
  - X.AI Grok-2 Vision model integration
  - Dynamic video track subscription
  - Image content injection into chat context
  - Track publication event handling
- file_path: recording_agent/page.mdoc
  title: Recording Agent
  category: egress
  tags:
  - recording
  - assemblyai
  - openai
  - cartesia
  difficulty: intermediate
  description: Shows how to create an agent that can record the input to a room and save it to a file.
  demonstrates:
  - Using egress to record the input to a room
- file_path: basic_event/page.mdoc
  title: Basic Event
  category: events
  tags:
  - events
  - openai
  - assemblyai
  - cartesia
  difficulty: beginner
  description: Shows how to use events in an agent to trigger actions.
  demonstrates:
  - Using events in an agent to trigger actions
  - Using on() to register an event listener
  - Using off() to unregister an event listener
  - Using once() to register an event listener that will only be triggered once
- file_path: event_emitters/page.mdoc
  title: Event Emitters
  category: events
  tags:
  - events
  - assemblyai
  - openai
  - cartesia
  difficulty: beginner
  description: Shows how to use event emitters in an agent to trigger actions.
  demonstrates:
  - Using event emitters in an agent to trigger custom actions like welcome and farewell messages
  - Custom event handling with EventEmitter
- file_path: update_tools/page.mdoc
  title: Dynamic Tool Updates
  category: function-calling
  tags:
  - dynamic-tools
  - tool-updates
  - runtime-modification
  - function-composition
  - assemblyai
  - openai
  - cartesia
  difficulty: intermediate
  description: Demonstrates dynamically adding function tools to agents at runtime
  demonstrates:
  - Dynamic function tool creation and addition
  - Runtime agent tool modification with update_tools
  - External function wrapping with function_tool decorator
  - Tool composition and agent enhancement
  - Combining static and dynamic function tools
- file_path: pi_zero_transcriber/page.mdoc
  title: Pi Zero Transcriber
  category: hardware
  tags:
  - hardware
  - deepgram
  difficulty: beginner
  description: Shows how to create a simple transcriber that uses the LiveKit SDK to transcribe audio from the microphone.
  demonstrates:
  - Using the LiveKit SDK to transcribe audio from the microphone.
  - Displaying the transcribed text on a Pirate Audio display on a Raspberry Pi Zero 2 W.
- file_path: home_assistant/page.mdoc
  title: Home Automation
  category: home-automation
  tags:
  - home-automation
  - openai
  - assemblyai
  difficulty: intermediate
  description: Shows how to create an agent that can control home automation devices.
  demonstrates:
  - Using function tools to control home automation devices.
  - Using a wake word to trigger the agent.
- file_path: stdio_mcp_client/page.mdoc
  title: MCP Agent
  category: mcp
  tags:
  - mcp
  - openai
  - assemblyai
  - cartesia
  difficulty: beginner
  description: Shows how to use a LiveKit Agent as an MCP client.
  demonstrates:
  - Connecting to a local MCP server as a client.
  - Connecting to a remote MCP server as a client.
  - Using a function tool to retrieve data from the MCP server.
- file_path: http_mcp_client/page.mdoc
  title: MCP Agent
  category: mcp
  tags:
  - mcp
  - assemblyai
  - openai
  - cartesia
  difficulty: beginner
  description: Shows how to use a LiveKit Agent as an MCP client.
  demonstrates:
  - Connecting to a remote MCP server as a client
  - Using MCP tools with voice-based interaction
- file_path: metrics_llm/page.mdoc
  title: LLM Metrics
  category: metrics
  tags:
  - metrics
  - assemblyai
  - openai
  - cartesia
  difficulty: beginner
  description: Shows how to use the LLM metrics to log metrics to the console for all of the different LLM models.
  demonstrates:
  - Using the LLM metrics to log metrics to the console
  - Inspecting fields like type, label, request ID, timestamps, duration, time to first token, cancellation, token counts,
    and throughput
- file_path: langfuse_tracing/page.mdoc
  title: Langfuse Tracing
  category: metrics
  tags:
  - metrics
  - openai
  - deepgram
  difficulty: intermediate
  description: Shows how to use the langfuse tracer to trace the agent session.
  demonstrates:
  - Using the langfuse tracer to trace the agent session.
  - Using the metrics_collected event to log metrics to langfuse.
- file_path: metrics_stt/page.mdoc
  title: STT Metrics
  category: metrics
  tags:
  - metrics
  - assemblyai
  - openai
  - cartesia
  difficulty: beginner
  description: Shows how to use the STT metrics to log metrics to the console.
  demonstrates:
  - Using the STT metrics to log metrics to the console.
  - This includes:
    - Type
    - Label
    - Request ID
    - Timestamp
    - Duration
    - Speech ID
    - Error
- file_path: metrics_tts/page.mdoc
  title: TTS Metrics
  category: metrics
  tags:
  - metrics
  - assemblyai
  - openai
  - cartesia
  difficulty: beginner
  description: Shows how to use the TTS metrics to log metrics to the console.
  demonstrates:
  - Using the TTS metrics to log metrics to the console.
  - This includes:
    - TTFB
    - Duration
    - Audio Duration
    - Cancelled
    - Characters Count
    - Streamed
    - Speech ID
    - Error
- file_path: metrics_vad/page.mdoc
  title: VAD Metrics
  category: metrics
  tags:
  - metrics
  - assemblyai
  - openai
  - cartesia
  difficulty: beginner
  description: Shows how to use the VAD metrics to log metrics to the console.
  demonstrates:
  - Using the VAD metrics to log metrics to the console.
  - This includes:
    - Idle Time
    - Inference Duration Total
    - Inference Count
    - Speech ID
    - Error
- file_path: agent_transfer/page.mdoc
  title: Agent Transfer
  category: multi-agent
  tags:
  - multi-agent
  - assemblyai
  - openai
  - cartesia
  difficulty: intermediate
  description: Shows how to switch between agents mid-call using function tools.
  demonstrates:
  - Agent transfer using update_agent()
  - Function tools for agent switching
  - Lightweight agent design with instructions and tools only
  - Shared AgentSession across agent swaps
  style: two-column
- file_path: interrupt_user/page.mdoc
  title: Interrupt User
  category: pipeline-llm
  tags:
  - pipeline-llm
  - openai
  - deepgram
  difficulty: intermediate
  description: Shows how to interrupt the user if they've spoken too much.
  demonstrates:
  - Using the `session.say` method to interrupt the user.
  - Using the `allow_interruptions` parameter to prevent the user from interrupting the agent.
  - Once the agent has spoken, it will allow the user to interrupt again.
- file_path: replacing_llm_output/page.mdoc
  title: LLM Output Replacement
  category: pipeline-llm
  tags:
  - deepseek
  - groq
  - deepgram
  - openai
  difficulty: intermediate
  description: Replaces Deepseek thinking tags with custom messages for TTS
  demonstrates:
  - Groq integration with Deepseek model
  - Real-time stream processing
  - Text replacement in LLM output
  - Custom llm_node for output manipulation
  - Handling model-specific output formats
- file_path: llm_powered_content_filter/page.mdoc
  title: LLM-Powered Content Filter
  category: pipeline-llm
  tags:
  - content_moderation
  - assemblyai
  - openai
  - cartesia
  difficulty: advanced
  description: Content filter using a separate LLM for real-time moderation decisions
  demonstrates:
  - Dual LLM setup (main + moderator)
  - Sentence-level content buffering
  - Stream processing with moderation checks
  - Custom llm_node override for filtering
  - Handling different chunk formats
  - Real-time content evaluation
- file_path: large_context/page.mdoc
  title: Large Context Window LLM
  category: pipeline-llm
  tags:
  - gemini_2_flash
  - large_context
  - book_analysis
  - war_and_peace
  difficulty: intermediate
  description: Agent using Gemini 2.0 Flash to analyze War and Peace with large context window
  demonstrates:
  - Loading large text files into LLM context
  - Google Gemini 2.0 Flash model for large contexts
  - Book analysis and discussion capabilities
  - Direct text quotation from context
  - Custom TTS instructions for literary tone
- file_path: simple_content_filter/page.mdoc
  title: Simple Content Filter
  category: pipeline-llm
  tags:
  - pipeline-llm
  - openai
  - deepgram
  difficulty: beginner
  description: Basic keyword-based content filter with inline replacement
  demonstrates:
  - Simple keyword filtering approach
  - Inline content replacement
  - Custom llm_node override
  - Static offensive terms list
  - Stream processing with substitution
- file_path: transcription_node/page.mdoc
  title: Transcription Node Modifier
  category: pipeline-llm
  tags:
  - transcription_modification
  - word_replacement
  - emoji_injection
  - assemblyai
  - openai
  - cartesia
  difficulty: intermediate
  description: Modifies transcriptions by replacing words with custom versions
  demonstrates:
  - Custom transcription_node override
  - Word replacement in transcriptions
  - Emoji injection in text
  - Async stream processing for text
  - Model settings usage
- file_path: keyword-detection/page.mdoc
  title: Keyword Detection
  category: pipeline-stt
  tags:
  - pipeline-stt
  - assemblyai
  - openai
  - cartesia
  difficulty: intermediate
  description: Shows how to detect keywords in user speech.
  demonstrates:
  - If the user says a keyword, the agent will log the keyword to the console.
  - Using the `user_input_transcribed` event to inspect transcripts.
- file_path: transcriber/page.mdoc
  title: Transcriber
  category: pipeline-stt
  tags:
  - pipeline-stt
  - assemblyai
  difficulty: beginner
  description: Shows how to transcribe user speech to text without TTS or an LLM.
  demonstrates:
  - Saving transcripts to a file.
  - An Agent that does not have TTS or an LLM. This is STT only.
- file_path: changing_language/page.mdoc
  title: ElevenLabs Change Language
  category: pipeline-tts
  tags:
  - pipeline-tts
  - elevenlabs
  - deepgram
  - openai
  difficulty: intermediate
  description: Shows how to use the ElevenLabs TTS model to change the language of the agent.
  demonstrates:
  - Using the update_options() method to change the language of STT and TTS
  - Allowing agents to self-update their own options using function tools
  - Accessing session STT/TTS from within an agent
- file_path: short_replies_only/page.mdoc
  title: Short Replies Only
  category: pipeline-tts
  tags:
  - pipeline-tts
  - openai
  - deepgram
  - rime
  difficulty: beginner
  description: Shows how to override the default TTS node to only respond with short replies based on the number of chunks.
  demonstrates:
  - Using the `tts_node` method to override the default TTS node and add custom logic to only respond with short replies.
  - Using the `session.interrupt` method to interrupt the agent if it's taking too long to respond, and then informing the
    user with `session.say`
- file_path: tts_comparison/page.mdoc
  title: TTS Comparison
  category: pipeline-tts
  tags:
  - pipeline-tts
  - deepgram
  - openai
  - rime
  - elevenlabs
  - cartesia
  - playai
  difficulty: intermediate
  description: Switches between different TTS providers using function tools.
  demonstrates:
  - Using function tools to switch between different TTS providers.
  - Each function tool returns a new agent with the same instructions, but with a different TTS provider.
- file_path: tts_node_modifications/page.mdoc
  title: TTS Node Override
  category: pipeline-tts
  tags:
  - pipeline-tts
  - deepgram
  - openai
  - rime
  difficulty: intermediate
  description: Shows how to override the default TTS node to do replacements on the output.
  demonstrates:
  - Using the `tts_node` method to override the default TTS node and add custom logic to do replacements on the output, like
    replacing "lol" with "<laughs>".
- file_path: rag/page.mdoc
  title: RAG Voice Agent
  category: rag
  tags:
  - annoy_index
  - vector_search
  - embeddings
  - livekit_docs
  - deduplication
  difficulty: advanced
  description: RAG-enabled agent with vector search for LiveKit documentation
  demonstrates:
  - Annoy vector index loading and querying
  - OpenAI embeddings for semantic search
  - Result deduplication with seen tracking
  - Function tool for document search
  - Paragraph-based context retrieval
  - Noise cancellation with BVC
- file_path: gemini_live_vision/page.mdoc
  title: Gemini Realtime Agent with Live Vision
  category: realtime
  tags:
  - gemini_realtime
  - live_vision
  - google
  difficulty: beginner
  description: Minimal Gemini Realtime model agent setup with live vision capabilities
  demonstrates:
  - Gemini Realtime model basic usage
  - Live vision capabilities
  - Session-based generation
  - VAD with Silero
  style: two-column
- file_path: rpc_agent/page.mdoc
  title: RPC State Management Agent
  category: rpc
  tags:
  - rpc
  - state-management
  - crud-operations
  - session-data
  - json-handling
  difficulty: advanced
  description: Agent demonstrating RPC communication with comprehensive CRUD state management
  demonstrates:
  - RPC payload extraction and JSON parsing
  - Session-based CRUD operations (Create, Read, Update, Delete)
  - State management with UUID-based object tracking
  - RPC method registration and error handling
  - Function tools integrated with RPC state
  - Structured error responses and logging
  style: two-column
- file_path: state_tracking/page.mdoc
  title: NPC Character State Tracking
  category: state-management
  tags:
  - state-management
  - assemblyai
  - openai
  - cartesia
  difficulty: advanced
  description: Advanced NPC system with dynamic rapport tracking and conversation state management
  demonstrates:
  - Complex character state tracking with rapport system
  - Multi-agent conversation flows and switching
  - Topic-based conversation management
  - Dynamic response variation based on relationship state
  - Agent inheritance patterns for character consistency
  - Session data persistence across interactions
- file_path: make_call/page.mdoc
  title: Outbound Calling Script
  category: telephony
  tags:
  - telephony
  - sip
  - outbound-calls
  difficulty: beginner
  description: Script that makes outbound calls via LiveKit Telephony using the LiveKit API
  demonstrates:
  - Creating agent dispatches via the API
  - Dialing outbound numbers via SIP trunks
  - LiveKit API client usage
- file_path: sip_lifecycle/page.mdoc
  title: SIP Lifecycle Management Agent
  category: telephony
  tags:
  - telephony
  - sip
  - deepgram
  - openai
  - elevenlabs
  difficulty: advanced
  description: Advanced SIP agent demonstrating complete call lifecycle management
  demonstrates:
  - SIP participant management and addition
  - Call status tracking and event handling
  - Room management and deletion
  - Participant attribute monitoring
  - Call lifecycle event handlers
  - Function tools for call operations
- file_path: answer_call/page.mdoc
  title: Simple Call Answering Agent
  category: telephony
  tags:
  - telephony
  - assemblyai
  - openai
  - cartesia
  difficulty: beginner
  description: Basic agent for handling incoming phone calls with simple conversation
  style: two-column
  githubUrl: https://github.com/livekit-examples/python-agents-examples/docs/examples/answer_call/answer_call.py
  demonstrates:
  - Simple telephony agent setup
  - Basic call handling workflow
  - Standard STT/LLM/TTS configuration
  - Automatic greeting generation on entry
  - Clean agent session lifecycle
- file_path: survey_caller/page.mdoc
  title: Survey Calling Agent
  category: telephony
  tags:
  - surveys
  - data-collection
  - csv-handling
  - automated-calling
  - metadata-processing
  - assemblyai
  - openai
  - cartesia
  difficulty: intermediate
  description: Automated survey calling agent with CSV data management and response recording
  demonstrates:
  - Survey data collection via phone calls
  - CSV file integration for response tracking
  - Metadata-based call configuration
  - Automated room deletion after completion
  - Function tools for survey response recording
  - Professional survey caller persona
- file_path: warm_handoff/page.mdoc
  title: Warm Handoff Agent
  category: telephony
  tags:
  - call-transfer
  - warm-handoff
  - sip
  - agent-to-human
  - function-tools
  - deepgram
  - openai
  - elevenlabs
  difficulty: intermediate
  description: Agent demonstrating warm handoff functionality to transfer calls to human agents
  demonstrates:
  - Call transfer to human agents via SIP
  - Warm handoff implementation patterns
  - SIP participant creation and management
  - Function tools for call operations
  - Multi-participant call handling
  - Professional call transfer announcements
- file_path: complex-agents/testing/start_test.py
  title: Basic Agent Test Starter
  category: testing
  tags:
  - pytest
  - basic-testing
  - getting-started
  - agent-greeting
  difficulty: beginner
  description: Simple starting point for testing voice agents with basic greeting validation
  demonstrates:
  - Basic pytest setup for agent testing
  - Environment configuration for testing
  - Simple agent session creation
  - Basic greeting behavior testing
  - LLM-based response validation
- file_path: complex-agents/testing/agent_test.py
  title: Comprehensive Agent Testing
  category: testing
  tags:
  - pytest
  - agent-testing
  - function-mocking
  - conversation-testing
  - fixtures
  difficulty: advanced
  description: Complete test suite for voice agents with fixtures, mocks, and conversation flows
  demonstrates:
  - Comprehensive pytest fixtures for agent testing
  - Function tool mocking and validation
  - Conversation flow and context testing
  - Error handling and edge case coverage
  - Parameterized testing for multiple scenarios
  - Class-based test organization with shared setup
- file_path: complex-agents/testing/agent.py
  title: Function Calling Test Agent
  category: testing
  tags:
  - function_calling
  - console_print
  - agent_session_config
  difficulty: beginner
  description: Testing agent with single print_to_console function
  demonstrates:
  - Basic function tool implementation
  - Console printing functionality
  - Custom agent instructions
  - Agent-level STT/LLM/TTS/VAD configuration
  - on_enter event handler
  - Returning tuples from function tools
- file_path: complex-agents/testing/testing_test.py
  title: Testing Test
  category: testing
  tags:
  - pytest
  - test-validation
  - duplicate-test
  - agent-greeting
  difficulty: beginner
  description: Duplicate test file demonstrating basic agent testing patterns
  demonstrates:
  - Basic pytest async test structure
  - Agent session lifecycle management
  - Environment variable loading for tests
  - Simple greeting validation pattern
  - LLM judge pattern for response evaluation
- file_path: pipeline_translator/page.mdoc
  title: Pipeline Translator Agent
  category: translation
  tags:
  - translation
  - multilingual
  - french
  - elevenlabs
  - deepgram
  - openai
  difficulty: intermediate
  description: Simple translation pipeline that converts English speech to French
  demonstrates:
  - Direct language translation workflow
  - Multilingual TTS configuration with ElevenLabs
  - Simple translation-focused agent instructions
  - Clean input-to-output translation pipeline
  - Voice-to-voice translation system
  style: two-column
- file_path: tts_translator/page.mdoc
  title: TTS Translator with Gladia STT
  category: translation
  tags:
  - translation
  - gladia
  - elevenlabs
  - multilingual
  - code-switching
  - event-handling
  difficulty: advanced
  description: Advanced translation system using Gladia STT with code switching and event handling
  demonstrates:
  - Gladia STT integration with multiple languages
  - Code switching between French and English
  - Translation event handling and processing
  - Custom STT configuration with translation capabilities
  - Event-driven transcription and speech synthesis
  - Advanced multilingual processing pipeline
- file_path: moondream_vision/page.mdoc
  title: Moondream Vision Agent
  category: vision
  tags:
  - moondream
  - vision
  - assemblyai
  - openai
  - cartesia
  difficulty: intermediate
  description: Moondream Vision Agent
  demonstrates:
  - Adding vision capabilities to an agent when the LLM does not have vision capabilities
  style: two-column