liveavatar-platform-rtc 1.0.0__tar.gz

liveavatar_platform_rtc-1.0.0/.codegraph/.gitignore

@@ -0,0 +1,16 @@
+# CodeGraph data files
+# These are local to each machine and should not be committed
+
+# Database
+*.db
+*.db-wal
+*.db-shm
+
+# Cache
+cache/
+
+# Logs
+*.log
+
+# Hook markers
+.dirty

liveavatar_platform_rtc-1.0.0/.gitattributes

@@ -0,0 +1,2 @@
+# Auto detect text files and perform LF normalization
+* text=auto

liveavatar_platform_rtc-1.0.0/.gitignore

@@ -0,0 +1,198 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# UV
+#   Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#uv.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/latest/usage/project/#working-with-version-control
+.pdm.toml
+.pdm-python
+.pdm-build/
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+
+# Ruff stuff:
+.ruff_cache/
+
+# PyPI configuration file
+.pypirc
+
+# Cursor  
+#  Cursor is an AI-powered code editor.`.cursorignore` specifies files/directories to 
+#  exclude from AI features like autocomplete and code analysis. Recommended for sensitive data
+#  refer to https://docs.cursor.com/context/ignore-files
+.cursorignore
+.cursorindexingignore
+
+# macOS
+.DS_Store
+
+# Claude Code local settings
+.claude/
+
+# oh-my-claudecode runtime state
+.omc/
+
+# Local isolated development worktrees
+.worktrees/
+
+# Accidental nested project copy (tracked by mistake)
+/liveavatar-platform-rtc-python/
+ragflow/
+.codegraph/daemon.pid

liveavatar_platform_rtc-1.0.0/.superpowers/sdd/.gitignore

@@ -0,0 +1 @@
+*

liveavatar_platform_rtc-1.0.0/.superpowers/sdd/final-fix-report.md

@@ -0,0 +1,66 @@
+# RTC chat final fix report
+
+## Status
+
+Implemented in `/tmp/rtc-chat-final-fix` with surgical changes to `agent.py`, `llm.py`, `tts.py`, `QUICKSTART.md`, and focused tests.
+
+- Chat readiness now has explicit state/event gating. Handlers register early; audio is dropped before all waits and ASR connect complete; text waits for readiness and is released without hanging by close/failure. Ready is set only after successful ASR connect.
+- TTS terminal items retain per-generation provider failures. JSON `error` and abnormal close raise from `frames()` with provider detail; normal done/close completes normally.
+- Response failures send the request-context error event and do not append conversation or log completion.
+- LLM turns use candidate history and commit only after a successful stream/callback; failure and cancellation preserve the pre-turn history. `close()` supports sync/async client close and is idempotent.
+- Pipeline close includes LLM close in best-effort parallel cleanup.
+- Runtime readiness failure schedules active cleanup, guarded by readiness-task/pipeline/client identity; the done callback never directly awaits or cancels itself.
+- RTC sample-rate publish/format failures propagate `avatar configured for 24 kHz output / align avatar configuration`; QUICKSTART uses the same wording.
+- Provider construction failure before pipeline ownership best-effort closes every already-created ASR/LLM/TTS provider before disconnecting the RTC client.
+- Invalid Base64 or malformed TTS PCM is retained as a generation terminal error (`invalid TTS PCM payload from provider`), raised by `frames()`, and cannot produce a completed response.
+- Added race/failure tests for readiness gating and close release, stale readiness cleanup, provider errors, sample-rate guidance, transactional LLM failures, and client close.
+
+## Verification command summaries
+
+TDD red run:
+
+```text
+PYTHONPATH=... python -m pytest -q tests/test_llm.py tests/test_tts.py tests/test_agent.py
+9 failed, 50 passed in 0.93s
+```
+
+Focused green run after implementation:
+
+```text
+PYTHONPATH=... python -m pytest -q tests/test_agent.py tests/test_llm.py tests/test_tts.py
+60 passed, 1 failed (the remaining failure exposed a bare re-raise of the original AudioPublishError)
+```
+
+Residual-review red/green run:
+
+```text
+python -m pytest -q tests/test_agent.py::test_tts_factory_failure_closes_created_llm_and_other_resources tests/test_tts.py::test_invalid_audio_payload_raises_in_frames
+RED: 2 failed in 0.61s
+GREEN: 2 passed in 0.50s
+```
+
+Final full offline suite after the two residual review fixes (combined venv/site-packages; localhost bind allowed for aiohttp route tests):
+
+```text
+PYTHONPATH=/tmp/rtc-chat-final-fix:/Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/.venv/lib/python3.14/site-packages:/Users/luffer/workspace/liveavatar-platform-rtc-python/src /Users/luffer/workspace/liveavatar-platform-rtc-python/.venv/bin/python -m pytest -q
+........................................................................ [ 92%]
+......                                                                   [100%]
+78 passed in 0.84s
+```
+
+Compile and independence/forbidden-symbol scan:
+
+```text
+PYTHONPATH=... python -m compileall -q agent.py asr.py llm.py tts.py tests
+! rg -n "liveavatar_channel_sdk|AvatarAgent|developer_tts|voice_id|voiceId" . -g '*.py'
+PYTHONPATH=... python -c 'import agent, asr, llm, tts; print("independent imports: ok")'
+independent imports: ok
+```
+
+The forbidden-symbol scan produced no matches and the combined command exited 0.
+
+## Findings
+
+- Existing offline coverage remains green; total suite is now 78 tests (the prior 64 plus focused additions and existing review additions).
+- One sandbox-only run could not bind `127.0.0.1` for six aiohttp route tests; rerunning the identical full suite with localhost bind permission passed all 78 tests.
+- Live smoke: **NOT RUN**, as requested. It requires real provider credentials/platform access.

liveavatar_platform_rtc-1.0.0/.superpowers/sdd/progress.md

@@ -0,0 +1,6 @@
+Task 1: complete (staged implementation, 10 tests passed, review clean)
+Task 2: complete (staged implementation, 7 tests passed, review clean)
+Task 3: complete (staged implementation, 17 tests passed, review clean)
+Task 4: complete (staged implementation, 19 tests passed, review clean)
+Task 5: complete (staged implementation, 34 tests passed, review clean)
+Task 6: complete (final implementation, 78 tests passed, whole-project review clean)

liveavatar_platform_rtc-1.0.0/.superpowers/sdd/task-1-brief.md

@@ -0,0 +1,102 @@
+### Task 1: Standalone configuration and LLM adapter
+
+**Files:**
+- Create: `/Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/rtc-chat/agent.py`
+- Create: `/Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/rtc-chat/llm.py`
+- Create: `/Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/rtc-chat/requirements.txt`
+- Create: `/Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/rtc-chat/tests/test_llm.py`
+
+**Interfaces:**
+- Produces: `Settings.from_env() -> Settings`, `Settings.validate() -> None`.
+- Produces: `LlmClient.stream_reply(user_text: str, on_delta: Callable[[str], Awaitable[None]]) -> str` and `reset() -> None`.
+
+- [ ] **Step 1: Write failing configuration and LLM tests**
+
+```python
+# tests/test_llm.py
+import pytest
+from agent import Settings
+from llm import LlmClient
+
+def test_settings_reject_missing_credentials():
+    settings = Settings("", "", None, "", "", "", "", "", "", 8080)
+    with pytest.raises(ValueError, match="LIVEAVATAR_API_KEY"):
+        settings.validate()
+
+@pytest.mark.asyncio
+async def test_stream_reply_forwards_deltas_and_retains_history(fake_openai):
+    client = LlmClient(api_key="key", base_url="https://llm.test", model="model",
+                       system_prompt="system", client=fake_openai(["你", "好"]))
+    deltas = []
+    reply = await client.stream_reply("hello", deltas.append)
+    assert reply == "你好"
+    assert deltas == ["你", "好"]
+    assert client.messages[-2:] == [
+        {"role": "user", "content": "hello"},
+        {"role": "assistant", "content": "你好"},
+    ]
+```
+
+- [ ] **Step 2: Run the focused tests and verify failure**
+
+Run: `cd /Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/rtc-chat && pytest tests/test_llm.py -q`
+
+Expected: import failure because `agent.py` and `llm.py` do not exist.
+
+- [ ] **Step 3: Implement settings and the minimal streaming LLM client**
+
+```python
+# llm.py core interface
+class LlmClient:
+    def __init__(self, *, api_key, base_url, model, system_prompt, client=None):
+        self._client = client or AsyncOpenAI(api_key=api_key, base_url=base_url)
+        self._model = model
+        self._system_prompt = system_prompt
+        self.messages = [{"role": "system", "content": system_prompt}]
+        self._lock = asyncio.Lock()
+
+    async def stream_reply(self, user_text, on_delta):
+        async with self._lock:
+            self.messages.append({"role": "user", "content": user_text})
+            self.messages = [self.messages[0]] + self.messages[-20:]
+            stream = await self._client.chat.completions.create(
+                model=self._model, messages=self.messages, stream=True,
+                max_tokens=512, temperature=0.7)
+            parts = []
+            async for chunk in stream:
+                text = chunk.choices[0].delta.content
+                if text:
+                    parts.append(text)
+                    result = on_delta(text)
+                    if inspect.isawaitable(result):
+                        await result
+            reply = "".join(parts)
+            if reply.strip():
+                self.messages.append({"role": "assistant", "content": reply})
+            return reply
+
+    def reset(self):
+        self.messages = [{"role": "system", "content": self._system_prompt}]
+```
+
+Implement `Settings` as a frozen dataclass populated from the exact environment variables in the design. `validate()` checks required values in this order: `LIVEAVATAR_API_KEY`, `LIVEAVATAR_AVATAR_ID`, `DASHSCOPE_API_KEY`, `DEEPSEEK_API_KEY`.
+
+Set `requirements.txt` to:
+
+```text
+aiohttp>=3.9
+dashscope>=1.25.11
+openai>=1.0.0
+pytest>=8.0
+pytest-asyncio>=0.23
+liveavatar-platform-rtc @ file:///Users/luffer/workspace/liveavatar-platform-rtc-python
+```
+
+- [ ] **Step 4: Run focused tests**
+
+Run: `pytest tests/test_llm.py -q`
+
+Expected: all tests pass.
+
+---
+

liveavatar_platform_rtc-1.0.0/.superpowers/sdd/task-1-report.md

@@ -0,0 +1,104 @@
+# Task 1 Report
+
+## Status
+
+DONE_WITH_CONCERNS
+
+Implementation is complete in the approved staging location `/tmp/rtc-chat-task1`. The requested destination is outside this agent's writable roots, so the main agent must copy the staged files into `/Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/rtc-chat`.
+
+## Files
+
+- `/tmp/rtc-chat-task1/agent.py`
+- `/tmp/rtc-chat-task1/llm.py`
+- `/tmp/rtc-chat-task1/requirements.txt`
+- `/tmp/rtc-chat-task1/tests/test_llm.py`
+
+## TDD and test results
+
+RED command:
+
+`pytest tests/test_llm.py -q`
+
+Raw result summary: exit 2 during collection with `ModuleNotFoundError: No module named 'agent'`, the expected missing-production-module failure before implementation.
+
+GREEN command attempted after implementation:
+
+`pytest tests/test_llm.py -q`
+
+Raw result summary: exit 2 during collection. In this environment the `pytest` console entry point did not add the working directory to `sys.path`, so it still reported `ModuleNotFoundError: No module named 'agent'` even though `agent.py` exists.
+
+Focused retry:
+
+`PYTHONPATH=. pytest tests/test_llm.py -q`
+
+Raw result summary: exit 2 during collection with `ModuleNotFoundError: No module named 'openai'`. Dependencies were not installed, per the instruction not to install anything.
+
+Compilation check:
+
+`python3 -m py_compile agent.py llm.py tests/test_llm.py`
+
+Raw result summary: exit 0, no output.
+
+## Review Follow-up (2026-07-15)
+
+- Removed the remaining forbidden-field assertion from `/tmp/rtc-chat-task1/tests/test_llm.py` so the staged source tree contains no match for that identifier.
+
+Command:
+
+`/Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/.venv/bin/python -m pytest -q /tmp/rtc-chat-task1/tests/test_llm.py`
+
+Raw output:
+
+```text
+..........                                                               [100%]
+10 passed in 0.25s
+```
+
+## Self-review
+
+- `Settings` is a frozen dataclass with the ten-field positional interface required by the supplied test.
+- `from_env()` reads the LiveAvatar, DashScope, DeepSeek, system prompt, and HTTP port environment variables with the specified demo defaults.
+- `validate()` checks missing credentials in the required order: `LIVEAVATAR_API_KEY`, `LIVEAVATAR_AVATAR_ID`, `DASHSCOPE_API_KEY`, `DEEPSEEK_API_KEY`.
+- `LlmClient` serializes turns with an asyncio lock, retains the system prompt plus the latest 20 conversation messages, forwards non-empty deltas to sync or async callbacks, stores non-blank assistant replies, and resets history to the system prompt.
+- `requirements.txt` exactly matches the brief.
+- No existing WebSocket chat, shared frontend, shared LLM client, or SDK source was modified.
+
+## Concerns
+
+- Main-agent authorization is required to copy the four staged files to the requested external directory.
+- The focused test could not reach GREEN in the current interpreter because `openai` is not installed. After copying, run the brief's test command in the demo environment with requirements installed; if its `pytest` entry point has the same path behavior, use `PYTHONPATH=.` or `python -m pytest`.
+
+## Review Fix (2026-07-15)
+
+### Fixes
+
+- Removed `Settings.liveavatar_voice_id` and the `LIVEAVATAR_VOICE_ID` environment read from `/tmp/rtc-chat-task1/agent.py`, because RTC start must not accept `voiceId`.
+- Corrected history trimming in `/tmp/rtc-chat-task1/llm.py` to retain the original system message plus only the latest 20 non-system messages, without duplicating the system entry.
+- Expanded `/tmp/rtc-chat-task1/tests/test_llm.py` to cover `Settings.from_env()`, all four required credential failures in validation order, `reset()`, async delta callbacks, and history trimming/system-message uniqueness.
+
+### RED result
+
+Command:
+
+`/Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/.venv/bin/python -m pytest -q /tmp/rtc-chat-task1/tests/test_llm.py`
+
+Raw result summary: exit 1; `6 failed, 4 passed in 0.27s`. The failures showed that `Settings` still required the removed voice field. The short-history assertion was then tightened to enforce exactly one system message before applying the production fix.
+
+### GREEN result
+
+Command:
+
+`/Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/.venv/bin/python -m pytest -q /tmp/rtc-chat-task1/tests/test_llm.py`
+
+Raw output:
+
+```text
+..........                                                               [100%]
+10 passed in 0.23s
+```
+
+Compilation check:
+
+`/Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/.venv/bin/python -m py_compile /tmp/rtc-chat-task1/agent.py /tmp/rtc-chat-task1/llm.py /tmp/rtc-chat-task1/tests/test_llm.py`
+
+Raw result summary: exit 0, no output.

liveavatar_platform_rtc-1.0.0/.superpowers/sdd/task-2-brief.md

@@ -0,0 +1,63 @@
+### Task 2: Qwen realtime ASR adapter
+
+**Files:**
+- Create: `/Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/rtc-chat/asr.py`
+- Create: `/Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/rtc-chat/tests/test_asr.py`
+
+**Interfaces:**
+- Consumes: RTC `AudioFrame.data` containing 16 kHz mono PCM16.
+- Produces: `QwenAsr.connect()`, `feed_audio(frame: AudioFrame)`, and `close()`.
+- Produces callbacks `on_speech_started`, `on_speech_stopped`, `on_partial(text)`, `on_final(text)`, and `on_error(message)` on the owning asyncio loop.
+
+- [ ] **Step 1: Write failing callback and audio tests**
+
+```python
+@pytest.mark.asyncio
+async def test_feed_audio_encodes_pcm(fake_conversation, audio_frame):
+    asr = make_asr(fake_conversation)
+    await asr.connect()
+    asr.feed_audio(audio_frame)
+    fake_conversation.append_audio.assert_called_once_with(
+        base64.b64encode(audio_frame.data.tobytes()).decode("ascii"))
+
+@pytest.mark.asyncio
+async def test_callback_dispatches_final_on_event_loop(callback, final_texts):
+    callback.on_event({"type": "conversation.item.input_audio_transcription.completed",
+                       "transcript": "你好"})
+    await asyncio.sleep(0)
+    assert final_texts == ["你好"]
+```
+
+- [ ] **Step 2: Run tests and verify failure**
+
+Run: `pytest tests/test_asr.py -q`
+
+Expected: import failure for `asr.QwenAsr`.
+
+- [ ] **Step 3: Implement the adapter around DashScope**
+
+Use `OmniRealtimeConversation`, `OmniRealtimeCallback`, and `TranscriptionParams`. In `connect()`, run blocking `conversation.connect` with `asyncio.to_thread`, then configure:
+
+```python
+conversation.update_session(
+    output_modalities=[MultiModality.TEXT],
+    enable_turn_detection=True,
+    turn_detection_type="server_vad",
+    turn_detection_threshold=0.0,
+    turn_detection_silence_duration_ms=400,
+    enable_input_audio_transcription=True,
+    transcription_params=TranscriptionParams(
+        language="zh", sample_rate=16000, input_audio_format="pcm"),
+)
+```
+
+Implement callback dispatch with `asyncio.run_coroutine_threadsafe()` for coroutine callbacks and `loop.call_soon_threadsafe()` for synchronous callbacks. Ignore empty partial/final strings. Make `close()` idempotent and run the blocking provider close in a thread.
+
+- [ ] **Step 4: Run focused tests**
+
+Run: `pytest tests/test_asr.py -q`
+
+Expected: all tests pass.
+
+---
+

liveavatar_platform_rtc-1.0.0/.superpowers/sdd/task-2-report.md

@@ -0,0 +1,86 @@
+# Task 2 Report: Qwen realtime ASR adapter
+
+## Status
+
+Complete in the required staging area only:
+
+- `/tmp/rtc-chat-task2/asr.py`
+- `/tmp/rtc-chat-task2/tests/test_asr.py`
+
+No Task 2 files were copied into the formal demo or SDK source trees.
+
+## TDD evidence
+
+### RED
+
+Command:
+
+```text
+cd /tmp/rtc-chat-task2
+/Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/.venv/bin/python -m pytest tests/test_asr.py -q
+```
+
+Result: collection failed with `ModuleNotFoundError: No module named 'asr'`, which is the expected missing-feature failure.
+
+### GREEN
+
+Command:
+
+```text
+cd /tmp/rtc-chat-task2
+PYTHONPATH=/Users/luffer/workspace/liveavatar-platform-rtc-python/src /Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/.venv/bin/python -m pytest tests/test_asr.py -q
+```
+
+Initial result before review fixes: `6 passed in 0.29s`.
+
+### Final verification
+
+Command:
+
+```text
+cd /tmp/rtc-chat-task2
+PYTHONPATH=/Users/luffer/workspace/liveavatar-platform-rtc-python/src /Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/.venv/bin/python -m pytest tests/test_asr.py -q
+/Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/.venv/bin/python -m py_compile asr.py tests/test_asr.py
+```
+
+Initial result before review fixes: `6 passed in 0.29s`; bytecode compilation
+exited successfully.
+
+## Implemented behavior
+
+- Injected DashScope provider boundary for network-free tests.
+- Blocking provider `connect()` and `close()` run through `asyncio.to_thread()`.
+- Session configuration matches the brief: text modality, server VAD, threshold `0.0`, 400 ms silence, and Chinese 16 kHz PCM transcription.
+- RTC audio uses `frame.to_pcm_bytes()` when available and falls back to `frame.data.tobytes()`, then base64-encodes for `append_audio()`.
+- Speech start/stop, partial/final transcript, and provider error events map to owner callbacks.
+- Provider-thread callbacks use `loop.call_soon_threadsafe()` to invoke on the
+  owning loop; returned awaitables are scheduled with `loop.create_task()`.
+- Empty partial/final values and unknown events are ignored.
+- Concurrent and repeated `close()` calls invoke provider close exactly once.
+
+## Self-check
+
+- Changes are limited to the required `/tmp` staging files and this report.
+- No dependency was installed and no network call was made.
+- Formal demo path `/Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/rtc-chat/asr.py` remains absent.
+- Existing Task 1 files and old demo/SDK files were not modified.
+
+## Concerns
+
+- DashScope in the supplied virtual environment defines `TranscriptionParams` in `dashscope.audio.qwen_omni.omni_realtime` but does not export it from `dashscope.audio.qwen_omni`; the implementation uses that official submodule import for compatibility.
+- Provider event schemas were implemented exactly from the supplied design/brief. A live credential smoke test was intentionally not run because this task requires network-free provider injection tests.
+
+## Review Fix Verification
+
+- Added the official Qwen ASR interim event contract
+  `conversation.item.input_audio_transcription.text` with `text`/`stash`, while
+  retaining compatible `.delta` handling.
+- Added an async callable-object regression test for loop-side awaitable
+  scheduling.
+- Fresh verification command run by the main agent:
+
+```text
+PYTHONPATH=/tmp/rtc-chat-task2:/Users/luffer/workspace/liveavatar-platform-rtc-python/src /Users/luffer/workspace/liveavatar-ws-integration-demo/python-agent/.venv/bin/python -m pytest -q /tmp/rtc-chat-task2/tests/test_asr.py
+```
+
+Result: `7 passed in 0.22s`.