swarmauri_agent_texttospeech 0.11.0.dev4__tar.gz → 0.11.0.dev5__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: swarmauri_agent_texttospeech
3
- Version: 0.11.0.dev4
3
+ Version: 0.11.0.dev5
4
4
  Summary: Stateless Swarmauri agent for converting independent text prompts into audio through any TTSBase provider.
5
5
  License-Expression: Apache-2.0
6
6
  License-File: LICENSE
@@ -20,8 +20,8 @@ Classifier: Programming Language :: Python :: 3.11
20
20
  Classifier: Programming Language :: Python :: 3.12
21
21
  Classifier: Topic :: Multimedia :: Sound/Audio :: Speech
22
22
  Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
23
- Requires-Dist: swarmauri_base
24
- Requires-Dist: swarmauri_core
23
+ Requires-Dist: swarmauri_base (>=0.11.0.dev12,<0.12)
24
+ Requires-Dist: swarmauri_core (>=0.11.0.dev7,<0.12)
25
25
  Description-Content-Type: text/markdown
26
26
 
27
27
  ![Swarmauri Logo](https://github.com/swarmauri/swarmauri-sdk/blob/3d4d1cfa949399d7019ae9d8f296afba773dfb7f/assets/swarmauri.brand.theme.svg)
@@ -55,16 +55,16 @@ uv add swarmauri_agent_texttospeech
55
55
  pip install swarmauri_agent_texttospeech
56
56
  ```
57
57
 
58
- Install a TTS provider package alongside the agent. For PlayHT:
58
+ Install a TTS provider package alongside the agent. For the alpha OpenAI provider:
59
59
 
60
60
  ```bash
61
- uv add swarmauri_tts_playht
61
+ uv add --prerelease allow swarmauri_tts_openai
62
62
  ```
63
63
 
64
64
  Or with `pip`:
65
65
 
66
66
  ```bash
67
- pip install swarmauri_tts_playht
67
+ pip install --pre swarmauri_tts_openai
68
68
  ```
69
69
 
70
70
  ## Usage
@@ -73,13 +73,12 @@ pip install swarmauri_tts_playht
73
73
  import os
74
74
 
75
75
  from swarmauri_agent_texttospeech import TextToSpeechAgent
76
- from swarmauri_tts_playht import PlayHTModel
76
+ from swarmauri_tts_openai import OpenaiTTS
77
77
 
78
- tts = PlayHTModel(
79
- api_key=os.environ["PLAYHT_API_KEY"],
80
- user_id=os.environ["PLAYHT_USER_ID"],
81
- name="Play3.0-mini",
82
- voice="Adolfo",
78
+ tts = OpenaiTTS(
79
+ api_key=os.environ["OPENAI_API_KEY"],
80
+ name="gpt-4o-mini-tts",
81
+ voice="alloy",
83
82
  )
84
83
  agent = TextToSpeechAgent(tts=tts, output_path="greeting.mp3")
85
84
 
@@ -90,6 +89,40 @@ print(audio_path)
90
89
  Each call is independent. The agent passes only that call's text to the TTS
91
90
  provider and never constructs a conversation or system message.
92
91
 
92
+ ## Structured dialogue and bounded batches
93
+
94
+ The agent also forwards a nonempty tuple of typed dialogue turns unchanged:
95
+
96
+ ```python
97
+ import os
98
+
99
+ from swarmauri_agent_texttospeech import TextToSpeechAgent
100
+ from swarmauri_tts_gemini_dialogue import (
101
+ GeminiDialogueTTS, GeminiDialogueTurn,
102
+ )
103
+
104
+ provider = GeminiDialogueTTS(
105
+ api_key=os.environ["GEMINI_API_KEY"],
106
+ speaker_voices={"A": "Puck", "B": "Kore"},
107
+ )
108
+ agent = TextToSpeechAgent(tts=provider, output_path="dialogue.wav")
109
+ turns = (
110
+ GeminiDialogueTurn(text="Hello", speaker="A"),
111
+ GeminiDialogueTurn(text="Hello back", speaker="B"),
112
+ )
113
+ agent.exec(turns)
114
+ # await agent.abatch([turns], max_concurrent=2)
115
+ ```
116
+
117
+ Speaker and voice validation belongs to the selected provider. A JSON string
118
+ is not parsed into turns. Set filenames to match the codec; Gemini/native
119
+ DeepInfra WAV defaults require .wav paths. Async batches default to at most
120
+ five concurrent calls, with a positive integer max_concurrent override.
121
+
122
+ The OpenAI Speech example above is temporary compatibility: its models have a
123
+ scheduled 2027-01-06 shutdown. All new provider packages remain alpha and live
124
+ verification is pending; see the SDK migration and verification guides.
125
+
93
126
  ## Contributing
94
127
 
95
128
  This package is part of the
@@ -29,16 +29,16 @@ uv add swarmauri_agent_texttospeech
29
29
  pip install swarmauri_agent_texttospeech
30
30
  ```
31
31
 
32
- Install a TTS provider package alongside the agent. For PlayHT:
32
+ Install a TTS provider package alongside the agent. For the alpha OpenAI provider:
33
33
 
34
34
  ```bash
35
- uv add swarmauri_tts_playht
35
+ uv add --prerelease allow swarmauri_tts_openai
36
36
  ```
37
37
 
38
38
  Or with `pip`:
39
39
 
40
40
  ```bash
41
- pip install swarmauri_tts_playht
41
+ pip install --pre swarmauri_tts_openai
42
42
  ```
43
43
 
44
44
  ## Usage
@@ -47,13 +47,12 @@ pip install swarmauri_tts_playht
47
47
  import os
48
48
 
49
49
  from swarmauri_agent_texttospeech import TextToSpeechAgent
50
- from swarmauri_tts_playht import PlayHTModel
50
+ from swarmauri_tts_openai import OpenaiTTS
51
51
 
52
- tts = PlayHTModel(
53
- api_key=os.environ["PLAYHT_API_KEY"],
54
- user_id=os.environ["PLAYHT_USER_ID"],
55
- name="Play3.0-mini",
56
- voice="Adolfo",
52
+ tts = OpenaiTTS(
53
+ api_key=os.environ["OPENAI_API_KEY"],
54
+ name="gpt-4o-mini-tts",
55
+ voice="alloy",
57
56
  )
58
57
  agent = TextToSpeechAgent(tts=tts, output_path="greeting.mp3")
59
58
 
@@ -64,6 +63,40 @@ print(audio_path)
64
63
  Each call is independent. The agent passes only that call's text to the TTS
65
64
  provider and never constructs a conversation or system message.
66
65
 
66
+ ## Structured dialogue and bounded batches
67
+
68
+ The agent also forwards a nonempty tuple of typed dialogue turns unchanged:
69
+
70
+ ```python
71
+ import os
72
+
73
+ from swarmauri_agent_texttospeech import TextToSpeechAgent
74
+ from swarmauri_tts_gemini_dialogue import (
75
+ GeminiDialogueTTS, GeminiDialogueTurn,
76
+ )
77
+
78
+ provider = GeminiDialogueTTS(
79
+ api_key=os.environ["GEMINI_API_KEY"],
80
+ speaker_voices={"A": "Puck", "B": "Kore"},
81
+ )
82
+ agent = TextToSpeechAgent(tts=provider, output_path="dialogue.wav")
83
+ turns = (
84
+ GeminiDialogueTurn(text="Hello", speaker="A"),
85
+ GeminiDialogueTurn(text="Hello back", speaker="B"),
86
+ )
87
+ agent.exec(turns)
88
+ # await agent.abatch([turns], max_concurrent=2)
89
+ ```
90
+
91
+ Speaker and voice validation belongs to the selected provider. A JSON string
92
+ is not parsed into turns. Set filenames to match the codec; Gemini/native
93
+ DeepInfra WAV defaults require .wav paths. Async batches default to at most
94
+ five concurrent calls, with a positive integer max_concurrent override.
95
+
96
+ The OpenAI Speech example above is temporary compatibility: its models have a
97
+ scheduled 2027-01-06 shutdown. All new provider packages remain alpha and live
98
+ verification is pending; see the SDK migration and verification guides.
99
+
67
100
  ## Contributing
68
101
 
69
102
  This package is part of the
@@ -1,6 +1,6 @@
1
1
  [project]
2
2
  name = "swarmauri_agent_texttospeech"
3
- version = "0.11.0.dev4"
3
+ version = "0.11.0.dev5"
4
4
  description = "Stateless Swarmauri agent for converting independent text prompts into audio through any TTSBase provider."
5
5
  license = "Apache-2.0"
6
6
  readme = "README.md"
@@ -22,8 +22,8 @@ classifiers = [
22
22
  ]
23
23
  authors = [{ name = "Jacob Stewart", email = "jacob@swarmauri.com" }]
24
24
  dependencies = [
25
- "swarmauri_core",
26
- "swarmauri_base",
25
+ "swarmauri_core>=0.11.0.dev7,<0.12",
26
+ "swarmauri_base>=0.11.0.dev12,<0.12",
27
27
  ]
28
28
  keywords = [
29
29
  "swarmauri",
@@ -39,6 +39,8 @@ keywords = [
39
39
  [tool.uv.sources]
40
40
  swarmauri_core = { workspace = true }
41
41
  swarmauri_base = { workspace = true }
42
+ swarmauri_tts_elevenlabs_dialogue = { workspace = true }
43
+ swarmauri_tts_gemini_dialogue = { workspace = true }
42
44
 
43
45
  [project.entry-points.'swarmauri.agents']
44
46
  TextToSpeechAgent = "swarmauri_agent_texttospeech:TextToSpeechAgent"
@@ -54,6 +56,8 @@ build-backend = "poetry.core.masonry.api"
54
56
 
55
57
  [dependency-groups]
56
58
  dev = [
59
+ "swarmauri_tts_elevenlabs_dialogue>=0.1.0a1,<0.2",
60
+ "swarmauri_tts_gemini_dialogue>=0.1.0a1,<0.2",
57
61
  "pytest>=8.0",
58
62
  "pytest-asyncio>=0.24.0",
59
63
  "ruff>=0.9.9",
@@ -7,7 +7,7 @@ from collections.abc import Sequence
7
7
  from pathlib import Path
8
8
  from typing import Any, Dict, List, Literal, Optional, Union
9
9
 
10
- from pydantic import ConfigDict, Field
10
+ from pydantic import BaseModel, ConfigDict, Field
11
11
 
12
12
  from swarmauri_base.ComponentBase import ComponentBase, SubclassUnion
13
13
  from swarmauri_base.agents.AgentBase import AgentBase
@@ -41,7 +41,7 @@ class TextToSpeechAgent(AgentBase):
41
41
 
42
42
  def exec(
43
43
  self,
44
- input_str: Optional[Union[str, IMessage]] = "",
44
+ input_str: Optional[Union[str, IMessage, tuple[BaseModel, ...]]] = "",
45
45
  llm_kwargs: Optional[Dict] = None,
46
46
  *,
47
47
  audio_path: Optional[str] = None,
@@ -58,7 +58,7 @@ class TextToSpeechAgent(AgentBase):
58
58
 
59
59
  async def aexec(
60
60
  self,
61
- input_str: Optional[Union[str, IMessage]] = "",
61
+ input_str: Optional[Union[str, IMessage, tuple[BaseModel, ...]]] = "",
62
62
  llm_kwargs: Optional[Dict] = None,
63
63
  *,
64
64
  audio_path: Optional[str] = None,
@@ -75,7 +75,7 @@ class TextToSpeechAgent(AgentBase):
75
75
 
76
76
  def batch(
77
77
  self,
78
- inputs: List[Union[str, IMessage]],
78
+ inputs: List[Union[str, IMessage, tuple[BaseModel, ...]]],
79
79
  llm_kwargs: Optional[Dict] = None,
80
80
  *,
81
81
  audio_paths: Optional[Sequence[str]] = None,
@@ -95,35 +95,57 @@ class TextToSpeechAgent(AgentBase):
95
95
 
96
96
  async def abatch(
97
97
  self,
98
- inputs: List[Union[str, IMessage]],
98
+ inputs: List[Union[str, IMessage, tuple[BaseModel, ...]]],
99
99
  llm_kwargs: Optional[Dict] = None,
100
100
  *,
101
101
  audio_paths: Optional[Sequence[str]] = None,
102
102
  tts_kwargs: Optional[Dict[str, Any]] = None,
103
+ max_concurrent: int = 5,
103
104
  ) -> List[str]:
104
- """Concurrently synthesize independent inputs."""
105
+ """Synthesize independent inputs with bounded concurrency."""
106
+ if (
107
+ isinstance(max_concurrent, bool)
108
+ or not isinstance(max_concurrent, int)
109
+ or max_concurrent < 1
110
+ ):
111
+ raise ValueError("max_concurrent must be a positive integer.")
105
112
  paths = self._resolve_batch_paths(len(inputs), audio_paths)
106
- return await asyncio.gather(
107
- *[
108
- self.aexec(
113
+ semaphore = asyncio.Semaphore(max_concurrent)
114
+
115
+ async def generate(input_data, path):
116
+ async with semaphore:
117
+ return await self.aexec(
109
118
  input_data,
110
119
  llm_kwargs=llm_kwargs,
111
120
  audio_path=path,
112
121
  tts_kwargs=tts_kwargs,
113
122
  )
123
+
124
+ return await asyncio.gather(
125
+ *(
126
+ generate(input_data, path)
114
127
  for input_data, path in zip(inputs, paths)
115
- ]
128
+ )
116
129
  )
117
130
 
118
131
  @staticmethod
119
- def _input_text(input_data: Optional[Union[str, IMessage]]) -> str:
132
+ def _input_text(
133
+ input_data: Optional[Union[str, IMessage, tuple[BaseModel, ...]]],
134
+ ) -> Union[str, tuple[BaseModel, ...]]:
135
+ if (
136
+ isinstance(input_data, tuple)
137
+ and input_data
138
+ and all(isinstance(turn, BaseModel) for turn in input_data)
139
+ ):
140
+ return input_data
120
141
  if isinstance(input_data, str):
121
142
  text = input_data
122
143
  elif isinstance(input_data, IMessage):
123
144
  text = input_data.content
124
145
  else:
125
146
  raise TypeError(
126
- "Input must be text or an IMessage with text content."
147
+ "Input must be text, an IMessage, "
148
+ "or a nonempty tuple of typed dialogue turns."
127
149
  )
128
150
 
129
151
  if not isinstance(text, str):