prepare for release (#1007)

.changeset/wild-cougars-decide.md

@@ -0,0 +1,6 @@
+---
+"livekit-agents": minor
+"livekit-plugins-llama-index": minor
+---
+
+prepare for release

.github/workflows/check-types.yml → .github/workflows/ci,yml

@@ -1,4 +1,4 @@
-name: Check Types
+name: CI
 
 on:
   push:
@@ -12,9 +12,29 @@ on:
   workflow_dispatch:
 
 jobs:
-  type-check:
+  ruff:
     runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/setup-python@v4
+        with:
+          python-version: "3.9"
+          cache: "pip"
 
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install ruff
+
+      - name: Ruff
+        run: ruff check --output-format=github .
+
+      - name: Check format
+        run: ruff format --check .
+
+  type-check:
+    runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
         with:
@@ -53,8 +73,7 @@ jobs:
             types-psutil \
             types-pyOpenSSL \
             types-requests \
-            types-openpyxl \
-            types-requests
+            types-openpyxl
 
       - name: Check Types
         run: |
@@ -70,3 +89,4 @@ jobs:
                -p livekit.plugins.rag \
                -p livekit.plugins.azure \
                -p livekit.plugins.anthropic
+

.github/workflows/tests.yml

@@ -1,4 +1,4 @@
-name: Tests
+name: tests
 
 on:
   push:
@@ -13,25 +13,34 @@ on:
 
 jobs:
   tests:
-    # don't run tests for PRs on forks
+    # Don't run tests for PRs on forks
     if: github.event_name != 'pull_request' || github.event.pull_request.head.repo.fork == false
     strategy:
       fail-fast: false
       matrix:
         include:
+          # Base tests on all operating systems
           - os: macos-14-large
-            platform: macos
+            test_group: base
           - os: macos-14
-            platform: macos
+            test_group: base
           - os: windows-2019
-            platform: windows
+            test_group: base
           - os: ubuntu-20.04
-            platform: linux
+            test_group: base
           - os: namespace-profile-default-arm64
-            platform: linux
+            test_group: base
+
+          # Run llm, stt, and tts tests only on Ubuntu
+          - os: ubuntu-20.04
+            test_group: llm
+          - os: ubuntu-20.04
+            test_group: stt
+          - os: ubuntu-20.04
+            test_group: tts
 
     runs-on: ${{ matrix.os }}
-    name: Run tests (${{ matrix.os }})
+    name: ${{ matrix.test_group }} — ${{ matrix.os }})
     steps:
       - uses: actions/checkout@v4
         with:
@@ -50,14 +59,12 @@ jobs:
             ${{ runner.os }}-cache
 
       - uses: actions/setup-python@v5
-        # on mac, ffmpeg installs python
-        if: ${{ matrix.os != 'macos' }}
         with:
           python-version: "3.9"
           cache: "pip"
 
       - name: Install ffmpeg (Linux)
-        if: ${{ matrix.platform == 'linux' }}
+        if: ${{ matrix.os == 'ubuntu-20.04' || matrix.os == 'namespace-profile-default-arm64' }}
         run: sudo apt-get update && sudo apt-get install -y ffmpeg
 
       # Azure plugin fails with OpenSSL3, and Ubuntu 22.04 does not include libssl1.1 in its repos
@@ -70,11 +77,11 @@ jobs:
           sudo dpkg -i libssl-dev_1.1.1-1ubuntu2.1_arm64.deb
 
       - name: Install ffmpeg (macOS)
-        if: ${{ matrix.platform == 'macos' }}
+        if: ${{ startsWith(matrix.os, 'macos') }}
         run: brew install ffmpeg
 
       - name: Install ffmpeg (Windows)
-        if: ${{ matrix.platform == 'windows' }}
+        if: ${{ matrix.os == 'windows-2019' }}
         run: choco install ffmpeg
 
       - name: Install packages
@@ -108,6 +115,28 @@ jobs:
           GOOGLE_CREDENTIALS_JSON: ${{ secrets.GOOGLE_CREDENTIALS_JSON }}
           ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
           GOOGLE_APPLICATION_CREDENTIALS: google.json
+          PYTEST_ADDOPTS: "--color=yes"
+        working-directory: tests
         run: |
-          echo $GOOGLE_CREDENTIALS_JSON > google.json
-          pytest --asyncio-mode=auto --timeout=60 .
+          echo "$GOOGLE_CREDENTIALS_JSON" > google.json
+
+          case "${{ matrix.test_group }}" in
+            base)
+              test_files="test_aio.py test_tokenizer.py test_vad.py test_ipc.py"
+              ;;
+            llm)
+              test_files="test_llm.py"
+              ;;
+            stt)
+              test_files="test_stt.py"
+              ;;
+            tts)
+              test_files="test_tts.py"
+              ;;
+            *)
+              echo "Unknown test group: ${{ matrix.test_group }}"
+              exit 1
+              ;;
+          esac
+          pytest --asyncio-mode=auto --timeout=60 $test_files
+

examples/voice-pipeline-agent/llamaindex-rag/README.md

@@ -0,0 +1,11 @@
+# RAG Example using LlamaIndex
+
+This repository showcases three ways to build a voice assistant with Retrieval-Augmented Generation (RAG) using LlamaIndex:
+
+1. **`chat_engine.py`**: Utilizes LlamaIndex's `as_chat_engine` for a straightforward, integrated solution. **Trade-off**: Lacks function calling support, limiting advanced interactions.
+
+2. **`query_engine.py`**: Uses an LLM that supports function calling (e.g., OpenAI's models) to define custom functions like `query_info` for retrieval. **Trade-off**: Requires additional setup but offers greater flexibility.
+
+3. **`retrieval.py`**: Manually injects retrieved context into the system prompt using LlamaIndex's retriever. **Trade-off**: Provides fine-grained control but involves complex prompt engineering.
+
+**Current recommended way**: Use **`query_engine.py`** for its balance of flexibility and control, enabling function calling and custom behaviors without excessive complexity.

examples/voice-pipeline-agent/llamaindex-rag/chat_engine.py

@@ -27,8 +27,6 @@
     storage_context = StorageContext.from_defaults(persist_dir=PERSIST_DIR)
     index = load_index_from_storage(storage_context)
 
-chat_engine = index.as_chat_engine(chat_mode=ChatMode.CONTEXT)
-
 
 async def entrypoint(ctx: JobContext):
     initial_ctx = llm.ChatContext().append(
@@ -38,6 +36,7 @@ async def entrypoint(ctx: JobContext):
             "You should use short and concise responses, and avoiding usage of unpronouncable punctuation."
         ),
     )
+    chat_engine = index.as_chat_engine(chat_mode=ChatMode.CONTEXT)
 
     await ctx.connect(auto_subscribe=AutoSubscribe.AUDIO_ONLY)
 

livekit-agents/livekit/agents/cli/log.py

@@ -20,6 +20,14 @@
     "watchfiles",
 ]
 
+
+def _silence_noisy_loggers() -> None:
+    for noisy_logger in NOISY_LOGGERS:
+        logger = logging.getLogger(noisy_logger)
+        if logger.level == logging.NOTSET:
+            logger.setLevel(logging.WARN)
+
+
 # skip default LogRecord attributes
 # http://docs.python.org/library/logging.html#logrecord-attributes
 _RESERVED_ATTRS: Tuple[str, ...] = (
@@ -209,10 +217,7 @@ def setup_logging(log_level: str, devmode: bool) -> None:
     root.addHandler(handler)
     root.setLevel(log_level)
 
-    for noisy_logger in NOISY_LOGGERS:
-        logger = logging.getLogger(noisy_logger)
-        if logger.level == logging.NOTSET:
-            logger.setLevel(logging.WARN)
+    _silence_noisy_loggers()
 
     from ..log import logger
 

livekit-agents/livekit/agents/pipeline/pipeline_agent.py

@@ -110,7 +110,7 @@ class _ImplOptions:
     int_speech_duration: float
     int_min_words: int
     min_endpointing_delay: float
-    max_recursive_fnc_calls: int
+    max_nested_fnc_calls: int
     preemptive_synthesis: bool
     before_llm_cb: BeforeLLMCallback
     before_tts_cb: BeforeTTSCallback
@@ -161,7 +161,7 @@ def __init__(
         interrupt_speech_duration: float = 0.5,
         interrupt_min_words: int = 0,
         min_endpointing_delay: float = 0.5,
-        max_recursive_fnc_calls: int = 1,
+        max_nested_fnc_calls: int = 1,
         preemptive_synthesis: bool = False,
         transcription: AgentTranscriptionOptions = AgentTranscriptionOptions(),
         before_llm_cb: BeforeLLMCallback = _default_before_llm_cb,
@@ -186,6 +186,8 @@ def __init__(
             interrupt_min_words: Minimum number of words to consider for interruption.
                 Defaults to 0 as this may increase the latency depending on the STT.
             min_endpointing_delay: Delay to wait before considering the user finished speaking.
+            max_nested_fnc_calls: Maximum number of nested function calls allowed for chaining
+                function calls (e.g functions that depend on each other).
             preemptive_synthesis: Whether to preemptively synthesize responses.
             transcription: Options for assistant transcription.
             before_llm_cb: Callback called when the assistant is about to synthesize a reply.
@@ -216,7 +218,7 @@ def __init__(
             int_speech_duration=interrupt_speech_duration,
             int_min_words=interrupt_min_words,
             min_endpointing_delay=min_endpointing_delay,
-            max_recursive_fnc_calls=max_recursive_fnc_calls,
+            max_nested_fnc_calls=max_nested_fnc_calls,
             preemptive_synthesis=preemptive_synthesis,
             transcription=transcription,
             before_llm_cb=before_llm_cb,
@@ -734,7 +736,7 @@ def _commit_user_question_if_needed() -> None:
 
             new_function_calls = llm_stream.function_calls
 
-            for i in range(self._opts.max_recursive_fnc_calls):
+            for i in range(self._opts.max_nested_fnc_calls):
                 self.emit("function_calls_collected", new_function_calls)
 
                 called_fncs = []
@@ -788,7 +790,7 @@ def _commit_user_question_if_needed() -> None:
                 answer_llm_stream = self._llm.chat(
                     chat_ctx=chat_ctx,
                     fnc_ctx=self.fnc_ctx
-                    if i < self._opts.max_recursive_fnc_calls - 1
+                    if i < self._opts.max_nested_fnc_calls - 1
                     else None,
                 )
                 answer_synthesis = self._synthesize_agent_speech(

livekit-agents/livekit/agents/tokenize/tokenizer.py

@@ -21,7 +21,7 @@ class TokenData:
 
 class SentenceTokenizer(ABC):
     @abstractmethod
-    def tokenize(self, *, text: str, language: str | None = None) -> list[str]:
+    def tokenize(self, text: str, *, language: str | None = None) -> list[str]:
         pass
 
     @abstractmethod
@@ -62,7 +62,7 @@ def _check_not_closed(self) -> None:
 
 class WordTokenizer(ABC):
     @abstractmethod
-    def tokenize(self, *, text: str, language: str | None = None) -> list[str]:
+    def tokenize(self, text: str, *, language: str | None = None) -> list[str]:
         pass
 
     @abstractmethod