livekit-plugins-difyai 1.0.22__py3-none-any.whl

livekit/plugins/dify/__init__.py

@@ -0,0 +1,25 @@
+from .llm import LLM
+from .version import __version__
+
+__all__ = ["LLM", "__version__"]
+
+from livekit.agents import Plugin
+
+from .log import logger
+
+
+class DifyPlugin(Plugin):
+    def __init__(self):
+        super().__init__(__name__, __version__, __package__, logger)
+
+
+Plugin.register_plugin(DifyPlugin())
+
+# Cleanup docs of unexported modules
+_module = dir()
+NOT_IN_ALL = [m for m in _module if m not in __all__]
+
+__pdoc__ = {}
+
+for n in NOT_IN_ALL:
+    __pdoc__[n] = False

livekit/plugins/dify/llm.py

@@ -0,0 +1,298 @@
+from __future__ import annotations
+
+import json
+import os
+from dataclasses import dataclass
+from typing import Any, Dict, Optional
+
+import aiohttp
+from livekit.agents import (
+    APIConnectionError,
+    APIStatusError,
+    llm,
+)
+from livekit.agents.types import (
+    DEFAULT_API_CONNECT_OPTIONS,
+    APIConnectOptions,
+    NotGivenOr,
+    NOT_GIVEN,
+)
+from livekit.agents.llm import FunctionTool, ToolChoice
+
+from .log import logger
+
+
+@dataclass
+class LLMOptions:
+    api_key: str
+    api_base: str = "https://api.dify.ai/v1"
+    user: str | None = None
+    _conversation_id: str = ""
+
+    def get_parameters_url(self) -> str:
+        return f"{self.api_base}/parameters"
+
+    def get_message_history_url(self) -> str:
+        return f"{self.api_base}/messages"
+
+    def get_conversation_history_url(self) -> str:
+        return f"{self.api_base}/conversations"
+
+    def get_headers(self) -> Dict[str, str]:
+        return {
+            "Authorization": f"Bearer {self.api_key}",
+            "Content-Type": "application/json",
+        }
+
+
+class LLM(llm.LLM):
+    def __init__(
+        self,
+        *,
+        api_key: str | None = None,
+        api_base: str | None = None,
+        user: str | None = None,
+    ) -> None:
+        """
+        Create a new instance of Dify LLM.
+
+        api_key (str | None): The Dify API key. Defaults to DIFY_API_KEY env var.
+        api_base (str | None): The base URL for the Dify API. Defaults to https://api.dify.ai/v1.
+        temperature (float | None): The temperature for generation. Defaults to None.
+        conversation_id (str | None): The conversation ID to continue. Defaults to None.
+        """
+        super().__init__()
+        self._session: Optional[aiohttp.ClientSession] = None
+
+        api_key = api_key or os.environ.get("DIFY_API_KEY")
+        if api_key is None:
+            raise ValueError("Dify API key is required")
+
+        api_base = api_base or os.environ.get("DIFY_API_BASE", "https://api.dify.ai/v1")
+        
+        if user is None:
+            user = os.environ.get("DIFY_USER", "test_user")
+
+        self._opts = LLMOptions(
+            api_key=api_key,
+            api_base=api_base,
+            user=user,
+        )
+
+    def chat(
+        self,
+        *,
+        chat_ctx: llm.ChatContext,
+        conn_options: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,
+        tools: list[FunctionTool] | None = None,
+        parallel_tool_calls: NotGivenOr[bool] = NOT_GIVEN,
+        tool_choice: NotGivenOr[ToolChoice] = NOT_GIVEN,
+        extra_kwargs: NotGivenOr[dict[str, Any]] = NOT_GIVEN,
+    ) -> "LLMStream":
+        """Start a chat completion stream"""
+        # Extract the last user message
+        last_message = next(
+            (msg for msg in reversed(chat_ctx.items) if msg.role == "user"), None
+        )
+        if not last_message:
+            raise ValueError("No user message found in chat context")
+
+        # Prepare the request payload
+        payload = {
+            "inputs": {},
+            "query": last_message.content[0],
+            "response_mode": "streaming",
+            "conversation_id": self._opts._conversation_id,
+            "user": self._opts.user,
+        }
+        # Create headers
+        headers = {
+            "Authorization": f"Bearer {self._opts.api_key}",
+            "Content-Type": "application/json",
+        }
+
+        # Create or reuse the session
+        if self._session is None:
+            self._session = aiohttp.ClientSession()
+        logger.info("llm start", extra={"query": last_message.content[0]})
+        # Create the stream
+        stream = self._session.post(
+            f"{self._opts.api_base}/chat-messages", headers=headers, json=payload
+        )
+
+        return LLMStream(
+            self,
+            dify_stream=stream,
+            chat_ctx=chat_ctx,
+            conn_options=conn_options,
+            tools=tools,
+        )
+
+    def ensure_session(self) -> None:
+        """Ensure that the session is created"""
+        if self._session is None:
+            self._session = aiohttp.ClientSession()
+
+    async def get_parameters(self) -> Dict:
+        """Get the parameters for the LLM"""
+        self.ensure_session()
+        res = await self._session.get(
+            url=self._opts.get_parameters_url(), headers=self._opts.get_headers()
+        )
+        return await res.json()
+
+    async def get_opening_words(self) -> str:
+        """Get the introduction text for the LLM"""
+        params = await self.get_parameters()
+        return params.get("opening_statement", "")
+
+    async def is_chatable(self) -> bool:
+        """Check if the LLM is chatable"""
+        url = self._opts.get_conversation_history_url()
+        url = f"{url}?&user={self._opts.user}"
+        self.ensure_session()
+        res = await self._session.get(url=url, headers=self._opts.get_headers())
+        if res.status != 200:
+            return False
+        return True
+
+    async def required_inputs(self) -> bool:
+        """Check if the LLM requires inputs"""
+        paras = await self.get_parameters()
+        user_input_form = paras.get("user_input_form", [])
+        if len(user_input_form) == 0:
+            return False
+        else:
+            required = False
+            for item in user_input_form:
+                key = list(item.items())[0][0]
+                _item = item.get(key, {})
+                required = _item.get("required", False)
+                break
+            return required
+
+    async def close(self) -> None:
+        """Close the LLM client and cleanup resources"""
+        if self._session is not None:
+            await self._session.close()
+            self._session = None
+
+    @classmethod
+    def from_env(cls) -> "LLM":
+        """Create a DifyLLM instance from environment variables"""
+        api_key = os.getenv("DIFY_API_KEY")
+        if not api_key:
+            raise ValueError("DIFY_API_KEY environment variable is required")
+
+        api_base = os.getenv("DIFY_API_BASE", "app-BbmsdDbay9Js28Ku780EDZRV")
+
+        return cls(
+            api_key=api_key,
+            api_base=api_base,
+        )
+
+
+class LLMStream(llm.LLMStream):
+    def __init__(
+        self,
+        llm: LLM,
+        *,
+        dify_stream: aiohttp.ClientResponse,
+        chat_ctx: llm.ChatContext,
+        conn_options: APIConnectOptions,
+        tools: list[FunctionTool] | None = None,
+    ) -> None:
+        super().__init__(llm, chat_ctx=chat_ctx, conn_options=conn_options, tools=tools)
+        self._awaitable_dify_stream = dify_stream
+        self._dify_stream: aiohttp.ClientResponse | None = None
+        self._request_id: str = ""
+        self._input_tokens = 0
+        self._output_tokens = 0
+
+    async def _run(self) -> None:
+        """Run the LLM stream"""
+        is_chatable = await self._llm.is_chatable()
+        required_inputs = await self._llm.required_inputs()
+        is_available = is_chatable and (not required_inputs)
+        if not is_available:
+            error_text = "Dify is not available"
+            raise APIConnectionError(error_text, retryable=False)
+
+        retryable = True
+        first_response = True
+        try:
+            if not self._dify_stream:
+                self._dify_stream = await self._awaitable_dify_stream
+            async with self._dify_stream as response:
+                if response.status != 200:
+                    error_text = await response.text()
+                    raise APIStatusError(
+                        f"Dify API error: {error_text}",
+                        status_code=response.status,
+                        body=error_text,
+                    )
+                async for line in response.content:
+                    if line:
+                        try:
+                            line = line.decode("utf-8").strip()
+                            if line.startswith("data: "):
+                                data = json.loads(line[6:])
+                                chat_chunk = self._parse_event(data, self._llm._opts)
+                                if chat_chunk is not None:
+                                    self._event_ch.send_nowait(chat_chunk)
+                                    retryable = False
+                                    if first_response:
+                                        logger.info("llm first response")
+                                        first_response = False
+                        except Exception as e:
+                            logger.error(f"Error processing stream: {e}")
+                            continue
+
+                # Send final usage stats
+                self._event_ch.send_nowait(
+                    llm.ChatChunk(
+                        id=self._request_id,
+                        usage=llm.CompletionUsage(
+                            completion_tokens=self._output_tokens,
+                            prompt_tokens=self._input_tokens,
+                            total_tokens=self._input_tokens + self._output_tokens,
+                        ),
+                    )
+                )
+                logger.info("llm end")
+
+        except aiohttp.ClientError as e:
+            raise APIConnectionError(retryable=retryable) from e
+        except Exception as e:
+            raise APIConnectionError(retryable=retryable) from e
+
+    def _parse_event(
+        self, event: Dict[str, Any], ops: LLMOptions
+    ) -> llm.ChatChunk | None:
+        """Parse a Dify event into a ChatChunk"""
+        event_type = event.get("event")
+        if event_type == "message_end":
+            # Update usage statistics
+            if "metadata" in event and "usage" in event["metadata"]:
+                usage = event["metadata"]["usage"]
+                self._input_tokens = usage.get("prompt_tokens", 0)
+                self._output_tokens = usage.get("completion_tokens", 0)
+            ops._conversation_id = event.get("conversation_id", "")
+            return None
+
+        elif event_type == "agent_message" or event_type == "message":
+            # Extract message content
+            answer = event.get("answer", "")
+            if not answer:
+                return None
+
+            return llm.ChatChunk(
+                id=event.get("message_id", ""),
+                delta=llm.ChoiceDelta(
+                    role="assistant",
+                    content=answer,
+                ),
+            )
+
+        return None
+

livekit/plugins/dify/log.py

@@ -0,0 +1,3 @@
+from logging import getLogger
+
+logger = getLogger("livekit.plugins.dify")

livekit/plugins/dify/version.py

@@ -0,0 +1 @@
+__version__ = "1.0.22"

livekit_plugins_difyai-1.0.22.dist-info/METADATA

@@ -0,0 +1,61 @@
+Metadata-Version: 2.4
+Name: livekit-plugins-difyai
+Version: 1.0.22
+Summary: LiveKit Agent Plugins for Dify
+Author-email: wangmengdi <790990241@qq.com>
+Keywords: audio,livekit,realtime,video,webrtc
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: Apache Software License
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3 :: Only
+Classifier: Programming Language :: Python :: 3.9
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Topic :: Multimedia :: Sound/Audio
+Classifier: Topic :: Multimedia :: Video
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.9
+Requires-Dist: livekit-agents~=1.0.22
+Description-Content-Type: text/markdown
+
+# LiveKit Plugins Dify
+
+Agent Framework plugin for Dify.
+
+## Installation
+```python
+pip install livekit-plugins-dify
+```
+
+## Pre-requisites
+
+- Dify API Key environment variables: `DIFY_API_KEY`
+
+## Usage
+
+
+This example shows how to use the Dify plugin.
+
+```python
+from livekit.agents import Agent, AgentSession, JobContext, cli, WorkerOptions
+from livekit.plugins import dify
+from dotenv import load_dotenv
+
+
+async def entry_point(ctx: JobContext):
+    
+    await ctx.connect()
+    
+    agent = Agent(instructions="You are a helpful assistant.")
+
+    session = AgentSession(
+        llm = dify.LLM(user="xxx")
+    )
+    
+    await session.start(agent=agent, room=ctx.room)
+
+
+if __name__ == "__main__":
+    load_dotenv()
+    cli.run_app(WorkerOptions(entrypoint_fnc=entry_point))
+```
+

livekit_plugins_difyai-1.0.22.dist-info/RECORD

@@ -0,0 +1,8 @@
+livekit/plugins/dify/__init__.py,sha256=sItl-nm33GTWN3M7Oavv8Ybe2l83U75_w4TQI0ubU74,477
+livekit/plugins/dify/llm.py,sha256=a-SL2fdrfAp2xKi9JqhO_O-_N1J5LmuRQ9iDIcYCG3Q,10331
+livekit/plugins/dify/log.py,sha256=Ok6vysHtAP2pArHHRCH0VMYMzjvMq_GxkvVhBT-p8Kg,74
+livekit/plugins/dify/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+livekit/plugins/dify/version.py,sha256=3qPJsI_cs0SlJn5DWM459KiP-4DSaTWmW-7lyF1jPMo,23
+livekit_plugins_difyai-1.0.22.dist-info/METADATA,sha256=zwnX4ioW8EwTtVDWXhFfk5HwXe5LdUfYeK-4i69Xgic,1554
+livekit_plugins_difyai-1.0.22.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+livekit_plugins_difyai-1.0.22.dist-info/RECORD,,

livekit_plugins_difyai-1.0.22.dist-info/WHEEL

@@ -0,0 +1,4 @@
+Wheel-Version: 1.0
+Generator: hatchling 1.27.0
+Root-Is-Purelib: true
+Tag: py3-none-any