setup text to speech models in Podcasts

open_notebook/domain/models.py

@@ -16,6 +16,13 @@ class Model(ObjectModel):
     provider: str
     type: str
 
+    @classmethod
+    def get_models_by_type(cls, model_type):
+        models = repo_query(
+            "SELECT * FROM model WHERE type=$model_type;", {"model_type": model_type}
+        )
+        return [Model(**model) for model in models]
+
 
 class DefaultModels(BaseModel):
     default_chat_model: Optional[str] = None

open_notebook/models/__init__.py

@@ -16,6 +16,10 @@ from open_notebook.models.llms import (
     VertexAnthropicLanguageModel,
 )
 from open_notebook.models.speech_to_text_models import OpenAISpeechToTextModel
+from open_notebook.models.text_to_speech_models import (
+    ElevenLabsTextToSpeechModel,
+    OpenAITextToSpeechModel,
+)
 
 # Unified model class map with type information
 MODEL_CLASS_MAP = {
@@ -38,7 +42,10 @@ MODEL_CLASS_MAP = {
     "speech_to_text": {
         "openai": OpenAISpeechToTextModel,
     },
-    "text_to_speech": {"openai": None, "elevenlabs": None},
+    "text_to_speech": {
+        "openai": OpenAITextToSpeechModel,
+        "elevenlabs": ElevenLabsTextToSpeechModel,
+    },
 }
 
 

open_notebook/models/text_to_speech_models.py

@@ -0,0 +1,26 @@
+"""
+Classes for supporting different text to speech models
+"""
+
+from abc import ABC
+from dataclasses import dataclass
+from typing import Optional
+
+
+@dataclass
+class TextToSpeechModel(ABC):
+    """
+    Abstract base class for text to speech models.
+    """
+
+    model_name: Optional[str] = None
+
+
+@dataclass
+class OpenAITextToSpeechModel(TextToSpeechModel):
+    model_name: str
+
+
+@dataclass
+class ElevenLabsTextToSpeechModel(TextToSpeechModel):
+    model_name: str

open_notebook/plugins/podcasts.py

@@ -1,4 +1,4 @@
-from typing import ClassVar, List, Literal, Optional
+from typing import ClassVar, List, Optional
 
 from loguru import logger
 from podcastfy.client import generate_podcast
@@ -31,7 +31,7 @@ class PodcastConfig(ObjectModel):
     ending_message: Optional[str] = None
     wordcount: int = Field(ge=400, le=10000)
     creativity: float = Field(ge=0, le=1)
-    provider: Literal["openai", "elevenlabs", "edge"] = Field(default="openai")
+    provider: str = Field(default="openai")
     voice1: Optional[str] = None
     voice2: Optional[str] = None
     model: str

pages/5_🎙️_Podcasts.py

@@ -1,6 +1,9 @@
+from typing import Dict, List
+
 import streamlit as st
 from streamlit_tags import st_tags
 
+from open_notebook.domain.models import Model
 from open_notebook.plugins.podcasts import (
     PodcastConfig,
     PodcastEpisode,
@@ -17,6 +20,21 @@ st.set_page_config(
 
 version_sidebar()
 
+text_to_speech_models = Model.get_models_by_type("text_to_speech")
+
+
+provider_models: Dict[str, List[str]] = {}
+
+for model in text_to_speech_models:
+    if model.provider not in provider_models:
+        provider_models[model.provider] = []
+    provider_models[model.provider].append(model.name)
+
+
+if len(text_to_speech_models) == 0:
+    st.error("No text to speech models found. Please set one up in the Settings page.")
+    st.stop()
+
 episodes_tab, templates_tab = st.tabs(["Episodes", "Templates"])
 
 with episodes_tab:
@@ -76,7 +94,7 @@ with templates_tab:
         pd_cfg["ending_message"] = st.text_input(
             "Ending Message", placeholder="Thank you for listening!"
         )
-        pd_cfg["provider"] = st.selectbox("Provider", ["openai", "elevenlabs", "edge"])
+        pd_cfg["provider"] = st.selectbox("Provider", provider_models.keys())
         pd_cfg["voice1"] = st.text_input(
             "Voice 1", help="You can use Elevenlabs voice ID"
         )
@@ -86,7 +104,8 @@ with templates_tab:
         pd_cfg["voice2"] = st.text_input(
             "Voice 2", help="You can use Elevenlabs voice ID"
         )
-        pd_cfg["model"] = st.text_input("Model")
+
+        pd_cfg["model"] = st.selectbox("Model", provider_models[pd_cfg["provider"]])
         st.caption(
             "OpenAI: tts-1 or tts-1-hd, Elevenlabs: eleven_multilingual_v2, eleven_turbo_v2_5"
         )
@@ -183,8 +202,8 @@ with templates_tab:
             )
             pd_config.provider = st.selectbox(
                 "Provider",
-                ["openai", "elevenlabs", "edge"],
-                index=["openai", "elevenlabs", "edge"].index(pd_config.provider),
+                list(provider_models.keys()),
+                index=list(provider_models.keys()).index(pd_config.provider),
                 key=f"provider_{pd_config.id}",
             )
             pd_config.voice1 = st.text_input(
@@ -202,8 +221,11 @@ with templates_tab:
                 key=f"voice2_{pd_config.id}",
                 help="You can use Elevenlabs voice ID",
             )
-            pd_config.model = st.text_input(
-                "Model", value=pd_config.model, key=f"model_{pd_config.id}"
+            pd_config.model = st.selectbox(
+                "Model",
+                provider_models[pd_config.provider],
+                index=provider_models[pd_config.provider].index(pd_config.model),
+                key=f"model_{pd_config.id}",
             )
             st.caption(
                 "OpenAI: tts-1 or tts-1-hd, Elevenlabs: eleven_multilingual_v2, eleven_turbo_v2_5"

pages/9_⚙️_Settings.py

@@ -45,7 +45,7 @@ provider_status["openrouter"] = (
     and os.environ.get("OPENROUTER_BASE_URL") is not None
 )
 provider_status["anthropic"] = os.environ.get("ANTHROPIC_API_KEY") is not None
-provider_status["eleven_labs"] = os.environ.get("ELEVENLABS_API_KEY") is not None
+provider_status["elevenlabs"] = os.environ.get("ELEVENLABS_API_KEY") is not None
 provider_status["litellm"] = (
     provider_status["ollama"]
     or provider_status["vertexai"]
@@ -104,7 +104,6 @@ with model_tab:
             st.warning(f"No models available for {model_type}")
 
 
-# todo: check for each type of model
 def get_selected_index(models, model_id, default=0):
     """Returns the index of the selected model in the list of models"""
     if not model_id or not models:
@@ -216,6 +215,3 @@ with model_defaults_tab:
     if st.button("Save Defaults", key="save_defaults"):
         DefaultModels.update(defs)
         st.rerun()
-
-# todo: return an error if a selected model is no longer supported
-# todo: do this check on the app homepage as well