swarmauri_agent_texttospeech 0.11.0.dev4__tar.gz → 0.11.0.dev6__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.
- {swarmauri_agent_texttospeech-0.11.0.dev4 → swarmauri_agent_texttospeech-0.11.0.dev6}/PKG-INFO +49 -12
- {swarmauri_agent_texttospeech-0.11.0.dev4 → swarmauri_agent_texttospeech-0.11.0.dev6}/README.md +46 -9
- {swarmauri_agent_texttospeech-0.11.0.dev4 → swarmauri_agent_texttospeech-0.11.0.dev6}/pyproject.toml +9 -3
- {swarmauri_agent_texttospeech-0.11.0.dev4 → swarmauri_agent_texttospeech-0.11.0.dev6}/swarmauri_agent_texttospeech/TextToSpeechAgent.py +55 -22
- {swarmauri_agent_texttospeech-0.11.0.dev4 → swarmauri_agent_texttospeech-0.11.0.dev6}/LICENSE +0 -0
- {swarmauri_agent_texttospeech-0.11.0.dev4 → swarmauri_agent_texttospeech-0.11.0.dev6}/swarmauri_agent_texttospeech/__init__.py +0 -0
{swarmauri_agent_texttospeech-0.11.0.dev4 → swarmauri_agent_texttospeech-0.11.0.dev6}/PKG-INFO
RENAMED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.4
|
|
2
2
|
Name: swarmauri_agent_texttospeech
|
|
3
|
-
Version: 0.11.0.
|
|
3
|
+
Version: 0.11.0.dev6
|
|
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
|

|
|
@@ -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
|
|
58
|
+
Install a TTS provider package alongside the agent. For the alpha OpenAI provider:
|
|
59
59
|
|
|
60
60
|
```bash
|
|
61
|
-
uv add
|
|
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
|
|
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
|
|
76
|
+
from swarmauri_tts_openai import OpenaiTTS
|
|
77
77
|
|
|
78
|
-
tts =
|
|
79
|
-
api_key=os.environ["
|
|
80
|
-
|
|
81
|
-
|
|
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
|
|
@@ -100,3 +133,7 @@ should follow the repository contribution and style guides.
|
|
|
100
133
|
|
|
101
134
|
Apache-2.0
|
|
102
135
|
|
|
136
|
+
Provider instances are validated against TTSBase at call-time construction, so
|
|
137
|
+
providers imported after the agent class remain usable. Serialized providers
|
|
138
|
+
retain their concrete fields; import that provider before deserialization.
|
|
139
|
+
|
{swarmauri_agent_texttospeech-0.11.0.dev4 → swarmauri_agent_texttospeech-0.11.0.dev6}/README.md
RENAMED
|
@@ -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
|
|
32
|
+
Install a TTS provider package alongside the agent. For the alpha OpenAI provider:
|
|
33
33
|
|
|
34
34
|
```bash
|
|
35
|
-
uv add
|
|
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
|
|
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
|
|
50
|
+
from swarmauri_tts_openai import OpenaiTTS
|
|
51
51
|
|
|
52
|
-
tts =
|
|
53
|
-
api_key=os.environ["
|
|
54
|
-
|
|
55
|
-
|
|
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
|
|
@@ -73,3 +106,7 @@ should follow the repository contribution and style guides.
|
|
|
73
106
|
## License
|
|
74
107
|
|
|
75
108
|
Apache-2.0
|
|
109
|
+
|
|
110
|
+
Provider instances are validated against TTSBase at call-time construction, so
|
|
111
|
+
providers imported after the agent class remain usable. Serialized providers
|
|
112
|
+
retain their concrete fields; import that provider before deserialization.
|
{swarmauri_agent_texttospeech-0.11.0.dev4 → swarmauri_agent_texttospeech-0.11.0.dev6}/pyproject.toml
RENAMED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
[project]
|
|
2
2
|
name = "swarmauri_agent_texttospeech"
|
|
3
|
-
version = "0.11.0.
|
|
3
|
+
version = "0.11.0.dev6"
|
|
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",
|
|
@@ -38,7 +38,10 @@ keywords = [
|
|
|
38
38
|
|
|
39
39
|
[tool.uv.sources]
|
|
40
40
|
swarmauri_core = { workspace = true }
|
|
41
|
+
swarmauri_tts_openai = { workspace = true }
|
|
41
42
|
swarmauri_base = { workspace = true }
|
|
43
|
+
swarmauri_tts_elevenlabs_dialogue = { workspace = true }
|
|
44
|
+
swarmauri_tts_gemini_dialogue = { workspace = true }
|
|
42
45
|
|
|
43
46
|
[project.entry-points.'swarmauri.agents']
|
|
44
47
|
TextToSpeechAgent = "swarmauri_agent_texttospeech:TextToSpeechAgent"
|
|
@@ -54,6 +57,9 @@ build-backend = "poetry.core.masonry.api"
|
|
|
54
57
|
|
|
55
58
|
[dependency-groups]
|
|
56
59
|
dev = [
|
|
60
|
+
"swarmauri_tts_openai>=0.1.0a2,<0.2",
|
|
61
|
+
"swarmauri_tts_elevenlabs_dialogue>=0.1.0a1,<0.2",
|
|
62
|
+
"swarmauri_tts_gemini_dialogue>=0.1.0a1,<0.2",
|
|
57
63
|
"pytest>=8.0",
|
|
58
64
|
"pytest-asyncio>=0.24.0",
|
|
59
65
|
"ruff>=0.9.9",
|
|
@@ -7,9 +7,15 @@ 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
|
|
11
|
-
|
|
12
|
-
|
|
10
|
+
from pydantic import (
|
|
11
|
+
BaseModel,
|
|
12
|
+
ConfigDict,
|
|
13
|
+
Field,
|
|
14
|
+
SerializeAsAny,
|
|
15
|
+
field_validator,
|
|
16
|
+
)
|
|
17
|
+
|
|
18
|
+
from swarmauri_base.ComponentBase import ComponentBase
|
|
13
19
|
from swarmauri_base.agents.AgentBase import AgentBase
|
|
14
20
|
from swarmauri_base.tts.TTSBase import TTSBase
|
|
15
21
|
from swarmauri_core.messages.IMessage import IMessage
|
|
@@ -21,27 +27,32 @@ class TextToSpeechAgent(AgentBase):
|
|
|
21
27
|
|
|
22
28
|
llm: None = Field(default=None, exclude=True)
|
|
23
29
|
llm_kwargs: None = Field(default=None, exclude=True)
|
|
24
|
-
tts:
|
|
30
|
+
tts: SerializeAsAny[TTSBase]
|
|
25
31
|
output_path: str = "output.mp3"
|
|
26
32
|
tts_kwargs: Dict[str, Any] = Field(default_factory=dict)
|
|
27
33
|
model_config = ConfigDict(extra="forbid", arbitrary_types_allowed=True)
|
|
28
34
|
type: Literal["TextToSpeechAgent"] = "TextToSpeechAgent"
|
|
29
35
|
|
|
30
|
-
|
|
31
|
-
|
|
32
|
-
|
|
33
|
-
|
|
34
|
-
|
|
36
|
+
@field_validator("tts", mode="before")
|
|
37
|
+
@classmethod
|
|
38
|
+
def _restore_tts(cls, value):
|
|
39
|
+
"""Resolve serialized providers at validation time, after discovery."""
|
|
40
|
+
if not isinstance(value, dict):
|
|
41
|
+
return value
|
|
42
|
+
type_name = value.get("type")
|
|
35
43
|
registry = TTSBase._registry.get("TTSBase", {})
|
|
36
44
|
tts_class = registry.get("subtypes", {}).get(type_name)
|
|
37
45
|
if tts_class is None and type_name == "TTSBase":
|
|
38
46
|
tts_class = registry.get("model_cls", TTSBase)
|
|
39
|
-
if tts_class is
|
|
40
|
-
|
|
47
|
+
if tts_class is None:
|
|
48
|
+
raise ValueError(
|
|
49
|
+
"Import the serialized TTS provider before loading it."
|
|
50
|
+
)
|
|
51
|
+
return tts_class.model_validate(value)
|
|
41
52
|
|
|
42
53
|
def exec(
|
|
43
54
|
self,
|
|
44
|
-
input_str: Optional[Union[str, IMessage]] = "",
|
|
55
|
+
input_str: Optional[Union[str, IMessage, tuple[BaseModel, ...]]] = "",
|
|
45
56
|
llm_kwargs: Optional[Dict] = None,
|
|
46
57
|
*,
|
|
47
58
|
audio_path: Optional[str] = None,
|
|
@@ -58,7 +69,7 @@ class TextToSpeechAgent(AgentBase):
|
|
|
58
69
|
|
|
59
70
|
async def aexec(
|
|
60
71
|
self,
|
|
61
|
-
input_str: Optional[Union[str, IMessage]] = "",
|
|
72
|
+
input_str: Optional[Union[str, IMessage, tuple[BaseModel, ...]]] = "",
|
|
62
73
|
llm_kwargs: Optional[Dict] = None,
|
|
63
74
|
*,
|
|
64
75
|
audio_path: Optional[str] = None,
|
|
@@ -75,7 +86,7 @@ class TextToSpeechAgent(AgentBase):
|
|
|
75
86
|
|
|
76
87
|
def batch(
|
|
77
88
|
self,
|
|
78
|
-
inputs: List[Union[str, IMessage]],
|
|
89
|
+
inputs: List[Union[str, IMessage, tuple[BaseModel, ...]]],
|
|
79
90
|
llm_kwargs: Optional[Dict] = None,
|
|
80
91
|
*,
|
|
81
92
|
audio_paths: Optional[Sequence[str]] = None,
|
|
@@ -95,35 +106,57 @@ class TextToSpeechAgent(AgentBase):
|
|
|
95
106
|
|
|
96
107
|
async def abatch(
|
|
97
108
|
self,
|
|
98
|
-
inputs: List[Union[str, IMessage]],
|
|
109
|
+
inputs: List[Union[str, IMessage, tuple[BaseModel, ...]]],
|
|
99
110
|
llm_kwargs: Optional[Dict] = None,
|
|
100
111
|
*,
|
|
101
112
|
audio_paths: Optional[Sequence[str]] = None,
|
|
102
113
|
tts_kwargs: Optional[Dict[str, Any]] = None,
|
|
114
|
+
max_concurrent: int = 5,
|
|
103
115
|
) -> List[str]:
|
|
104
|
-
"""
|
|
116
|
+
"""Synthesize independent inputs with bounded concurrency."""
|
|
117
|
+
if (
|
|
118
|
+
isinstance(max_concurrent, bool)
|
|
119
|
+
or not isinstance(max_concurrent, int)
|
|
120
|
+
or max_concurrent < 1
|
|
121
|
+
):
|
|
122
|
+
raise ValueError("max_concurrent must be a positive integer.")
|
|
105
123
|
paths = self._resolve_batch_paths(len(inputs), audio_paths)
|
|
106
|
-
|
|
107
|
-
|
|
108
|
-
|
|
124
|
+
semaphore = asyncio.Semaphore(max_concurrent)
|
|
125
|
+
|
|
126
|
+
async def generate(input_data, path):
|
|
127
|
+
async with semaphore:
|
|
128
|
+
return await self.aexec(
|
|
109
129
|
input_data,
|
|
110
130
|
llm_kwargs=llm_kwargs,
|
|
111
131
|
audio_path=path,
|
|
112
132
|
tts_kwargs=tts_kwargs,
|
|
113
133
|
)
|
|
134
|
+
|
|
135
|
+
return await asyncio.gather(
|
|
136
|
+
*(
|
|
137
|
+
generate(input_data, path)
|
|
114
138
|
for input_data, path in zip(inputs, paths)
|
|
115
|
-
|
|
139
|
+
)
|
|
116
140
|
)
|
|
117
141
|
|
|
118
142
|
@staticmethod
|
|
119
|
-
def _input_text(
|
|
143
|
+
def _input_text(
|
|
144
|
+
input_data: Optional[Union[str, IMessage, tuple[BaseModel, ...]]],
|
|
145
|
+
) -> Union[str, tuple[BaseModel, ...]]:
|
|
146
|
+
if (
|
|
147
|
+
isinstance(input_data, tuple)
|
|
148
|
+
and input_data
|
|
149
|
+
and all(isinstance(turn, BaseModel) for turn in input_data)
|
|
150
|
+
):
|
|
151
|
+
return input_data
|
|
120
152
|
if isinstance(input_data, str):
|
|
121
153
|
text = input_data
|
|
122
154
|
elif isinstance(input_data, IMessage):
|
|
123
155
|
text = input_data.content
|
|
124
156
|
else:
|
|
125
157
|
raise TypeError(
|
|
126
|
-
"Input must be text
|
|
158
|
+
"Input must be text, an IMessage, "
|
|
159
|
+
"or a nonempty tuple of typed dialogue turns."
|
|
127
160
|
)
|
|
128
161
|
|
|
129
162
|
if not isinstance(text, str):
|
{swarmauri_agent_texttospeech-0.11.0.dev4 → swarmauri_agent_texttospeech-0.11.0.dev6}/LICENSE
RENAMED
|
File without changes
|