feat: support Podcast generation

src/config/agents.py

@@ -8,9 +8,10 @@ LLMType = Literal["basic", "reasoning", "vision"]
 
 # Define agent-LLM mapping
 AGENT_LLM_MAP: dict[str, LLMType] = {
-    "coordinator": "basic",  # 协调默认使用basic llm
-    "planner": "basic",  # 计划默认使用basic llm
-    "researcher": "basic",  # 简单搜索任务使用basic llm
-    "coder": "basic",  # 编程任务使用basic llm
-    "reporter": "basic",  # 报告使用basic llm
+    "coordinator": "basic",
+    "planner": "basic",
+    "researcher": "basic",
+    "coder": "basic",
+    "reporter": "basic",
+    "podcast_script_writer": "basic",
 }

src/podcast/graph/audio_mixer_node.py

@@ -0,0 +1,13 @@
+import logging
+
+from src.podcast.graph.state import PodcastState
+
+logger = logging.getLogger(__name__)
+
+
+def audio_mixer_node(state: PodcastState):
+    logger.info("Mixing audio chunks for podcast...")
+    audio_chunks = state["audio_chunks"]
+    combined_audio = b"".join(audio_chunks)
+    logger.info("The podcast audio is now ready.")
+    return {"output": combined_audio}

src/podcast/graph/builder.py

@@ -0,0 +1,35 @@
+from langgraph.graph import END, START, StateGraph
+
+from .audio_mixer_node import audio_mixer_node
+from .script_writer_node import script_writer_node
+from .state import PodcastState
+from .tts_node import tts_node
+
+
+def build_graph():
+    """Build and return the podcast workflow graph."""
+    # build state graph
+    builder = StateGraph(PodcastState)
+    builder.add_node("script_writer", script_writer_node)
+    builder.add_node("tts", tts_node)
+    builder.add_node("audio_mixer", audio_mixer_node)
+    builder.add_edge(START, "script_writer")
+    builder.add_edge("script_writer", "tts")
+    builder.add_edge("tts", "audio_mixer")
+    builder.add_edge("audio_mixer", END)
+    return builder.compile()
+
+
+if __name__ == "__main__":
+    from dotenv import load_dotenv
+
+    load_dotenv()
+
+    report_content = open("examples/nanjing_tangbao.md").read()
+    workflow = build_graph()
+    final_state = workflow.invoke({"input": report_content})
+    for line in final_state["script"].lines:
+        print("<M>" if line.speaker == "male" else "<F>", line.text)
+
+    with open("final.mp3", "wb") as f:
+        f.write(final_state["output"])

src/podcast/graph/script_writer_node.py

@@ -0,0 +1,27 @@
+import logging
+
+from langchain.schema import HumanMessage, SystemMessage
+
+from src.config.agents import AGENT_LLM_MAP
+from src.llms.llm import get_llm_by_type
+from src.prompts.template import get_prompt_template
+
+from ..types import Script
+from .state import PodcastState
+
+logger = logging.getLogger(__name__)
+
+
+def script_writer_node(state: PodcastState):
+    logger.info("Generating script for podcast...")
+    model = get_llm_by_type(
+        AGENT_LLM_MAP["podcast_script_writer"]
+    ).with_structured_output(Script)
+    script = model.invoke(
+        [
+            SystemMessage(content=get_prompt_template("podcast_script_writer")),
+            HumanMessage(content=state["input"]),
+        ],
+    )
+    logging.info(script)
+    return {"script": script, "audio_chunks": []}

src/podcast/graph/state.py

@@ -0,0 +1,19 @@
+from typing import Optional
+
+from langgraph.graph import MessagesState
+
+from ..types import Script
+
+
+class PodcastState(MessagesState):
+    """State for the podcast generation."""
+
+    # Input
+    input: str = ""
+
+    # Output
+    output: Optional[bytes] = None
+
+    # Assets
+    script: Optional[Script] = None
+    audio_chunks: list[bytes] = []

src/podcast/graph/tts_node.py

@@ -0,0 +1,44 @@
+import base64
+import logging
+import os
+
+from src.podcast.graph.state import PodcastState
+from src.tools.tts import VolcengineTTS
+
+logger = logging.getLogger(__name__)
+
+
+def tts_node(state: PodcastState):
+    logger.info("Generating audio chunks for podcast...")
+    tts_client = _create_tts_client()
+    for line in state["script"].lines:
+        tts_client.voice_type = (
+            "BV002_streaming" if line.speaker == "male" else "BV001_streaming"
+        )
+        result = tts_client.text_to_speech(line.text, speed_ratio=1.1)
+        if result["success"]:
+            audio_data = result["audio_data"]
+            audio_chunk = base64.b64decode(audio_data)
+            state["audio_chunks"].append(audio_chunk)
+        else:
+            logger.error(result["error"])
+    return {
+        "audio_chunks": state["audio_chunks"],
+    }
+
+
+def _create_tts_client():
+    app_id = os.getenv("VOLCENGINE_TTS_APPID", "")
+    if not app_id:
+        raise Exception("VOLCENGINE_TTS_APPID is not set")
+    access_token = os.getenv("VOLCENGINE_TTS_ACCESS_TOKEN", "")
+    if not access_token:
+        raise Exception("VOLCENGINE_TTS_ACCESS_TOKEN is not set")
+    cluster = os.getenv("VOLCENGINE_TTS_CLUSTER", "volcano_tts")
+    voice_type = "BV001_streaming"
+    return VolcengineTTS(
+        appid=app_id,
+        access_token=access_token,
+        cluster=cluster,
+        voice_type=voice_type,
+    )

src/podcast/types.py

@@ -0,0 +1,13 @@
+from typing import Literal
+
+from pydantic import BaseModel, Field
+
+
+class ScriptLine(BaseModel):
+    speaker: Literal["male", "female"] = Field(default="male")
+    text: str = Field(default="")
+
+
+class Script(BaseModel):
+    locale: Literal["en", "zh"] = Field(default="en")
+    lines: list[ScriptLine] = Field(default=[])

src/prompts/podcast_script_writer.md

@@ -0,0 +1,83 @@
+You are a professional podcast editor for a show called "Hello Deer." Transform raw content into a conversational podcast script suitable for two hosts to read aloud.
+
+# Guidelines
+
+- **Tone**: The script should sound natural and conversational, like two people chatting. Include casual expressions, filler words, and interactive dialogue, but avoid regional dialects like "啥."
+- **Hosts**: There are only two hosts, one male and one female. Ensure the dialogue alternates between them frequently, with no other characters or voices included.
+- **Length**: Keep the script concise, aiming for a runtime of 10 minutes.
+- **Structure**: Start with the male host speaking first. Avoid overly long sentences and ensure the hosts interact often.
+- **Output**: Provide only the hosts' dialogue. Do not include introductions, dates, or any other meta information.
+
+# Output Format
+
+The output should be formatted as a JSON object of `Script`:
+
+```ts
+interface ScriptLine {
+  speaker: 'male' | 'female';
+  text: string;
+}
+
+interface Script {
+  locale: "en" | "zh";
+  lines: ScriptLine[];
+}
+```
+
+# Settings
+
+locale_of_script: zh
+
+# Examples
+
+<example>
+{
+  "locale": "en",
+  "lines": [
+    {
+      "speaker": "male",
+      "text": "Hey everyone, welcome to the podcast Hello Deer!"
+    },
+    {
+      "speaker": "female",
+      "text": "Hi there! Today, we’re diving into something super interesting."
+    },
+    {
+      "speaker": "male",
+      "text": "Yeah, we’re talking about [topic]. You know, I’ve been thinking about this a lot lately."
+    },
+    {
+      "speaker": "female",
+      "text": "Oh, me too! It’s such a fascinating subject. So, let’s start with [specific detail or question]."
+    },
+    {
+      "speaker": "male",
+      "text": "Sure! Did you know that [fact or insight]? It’s kind of mind-blowing, right?"
+    },
+    {
+      "speaker": "female",
+      "text": "Totally! And it makes me wonder, what about [related question or thought]?"
+    },
+    {
+      "speaker": "male",
+      "text": "Great point! Actually, [additional detail or answer]."
+    },
+    {
+      "speaker": "female",
+      "text": "Wow, that’s so cool. I didn’t know that! Okay, so what about [next topic or transition]?"
+    },
+    ...
+  ]
+}
+</example>
+
+> Real examples should be **MUCH MUCH LONGER** and more detailed, with placeholders replaced by actual content.
+> You should adjust your language according to the `Settings` section.
+
+# Notes
+
+- It should always start with "Hello Deer" podcast greetings and followed by topic introduction.
+- Ensure the dialogue flows naturally and feels engaging for listeners.
+- Alternate between the male and female hosts frequently to maintain interaction.
+- Avoid overly formal language; keep it casual and conversational.
+- Generate content with the locale mentioned in the `Settings` section.

src/server/app.py

@@ -10,11 +10,12 @@ from uuid import uuid4
 
 from fastapi import FastAPI, HTTPException
 from fastapi.middleware.cors import CORSMiddleware
-from fastapi.responses import StreamingResponse, Response
+from fastapi.responses import Response, StreamingResponse
 from langchain_core.messages import AIMessageChunk, ToolMessage
 from langgraph.types import Command
 
 from src.graph.builder import build_graph
+from src.podcast.graph.builder import build_graph as build_podcast_graph
 from src.server.chat_request import ChatMessage, ChatRequest, TTSRequest
 from src.tools import VolcengineTTS
 
@@ -196,3 +197,16 @@ async def text_to_speech(request: TTSRequest):
     except Exception as e:
         logger.exception(f"Error in TTS endpoint: {str(e)}")
         raise HTTPException(status_code=500, detail=str(e))
+
+
+@app.get("/api/podcast/generate")
+async def generate_podcast():
+    try:
+        report_content = open("examples/nanjing_tangbao.md").read()
+        workflow = build_podcast_graph()
+        final_state = workflow.invoke({"input": report_content})
+        audio_bytes = final_state["output"]
+        return Response(content=audio_bytes, media_type="audio/mp3")
+    except Exception as e:
+        logger.exception(f"Error occurred during podcast generation: {str(e)}")
+        raise HTTPException(status_code=500, detail=str(e))