Spaces:
Running
Running
Update agents/tools.py
Browse files- agents/tools.py +12 -11
agents/tools.py
CHANGED
|
@@ -23,6 +23,7 @@ def _get_llm_client() -> OpenAI | None:
|
|
| 23 |
api_key=token,
|
| 24 |
)
|
| 25 |
|
|
|
|
| 26 |
LLM_MODEL = "Qwen/Qwen2.5-7B-Instruct:together"
|
| 27 |
|
| 28 |
|
|
@@ -84,7 +85,7 @@ PERSONAS = {
|
|
| 84 |
"name": "Dave",
|
| 85 |
"show": "The Weekly Wheeze",
|
| 86 |
"desc": "42, raspy, laughs mid-sentence",
|
| 87 |
-
"voice": "
|
| 88 |
"style": "comedic stand-up energy, warm and relatable",
|
| 89 |
"sfx": ["drumroll.wav", "crowd_react.wav"],
|
| 90 |
},
|
|
@@ -92,7 +93,7 @@ PERSONAS = {
|
|
| 92 |
"name": "Elena",
|
| 93 |
"show": "Corporate Tears",
|
| 94 |
"desc": "31, soft, melancholic sighs",
|
| 95 |
-
"voice": "
|
| 96 |
"style": "melancholic NPR storyteller, gentle and empathetic",
|
| 97 |
"sfx": ["violin_sting.wav"],
|
| 98 |
},
|
|
@@ -100,7 +101,7 @@ PERSONAS = {
|
|
| 100 |
"name": "Marcus",
|
| 101 |
"show": "Officially Unhinged",
|
| 102 |
"desc": "55, baritone, barely-contained rage",
|
| 103 |
-
"voice": "
|
| 104 |
"style": "barely restrained fury, dark comedy, gravelly intensity",
|
| 105 |
"sfx": ["thunder.wav", "glass_break.wav", "dramatic_sting.wav"],
|
| 106 |
},
|
|
@@ -108,7 +109,7 @@ PERSONAS = {
|
|
| 108 |
"name": "Priya",
|
| 109 |
"show": "Please Stop Talking",
|
| 110 |
"desc": "27, fast talker, nervous giggle",
|
| 111 |
-
"voice": "
|
| 112 |
"style": "rapid nervous energy, second-hand cringe expert",
|
| 113 |
"sfx": [],
|
| 114 |
},
|
|
@@ -116,7 +117,7 @@ PERSONAS = {
|
|
| 116 |
"name": "Rex",
|
| 117 |
"show": "Total System Failure",
|
| 118 |
"desc": "???, manic, breaks 4th wall",
|
| 119 |
-
"voice": "
|
| 120 |
"style": "completely unhinged, 4th-wall breaking, reality-questioning",
|
| 121 |
"sfx": ["static.wav", "dial_up.wav", "system_error.wav", "vhs_rewind.wav"],
|
| 122 |
},
|
|
@@ -283,7 +284,7 @@ def parse_script(script: str) -> dict[str, str]:
|
|
| 283 |
# "hf" β HuggingFace InferenceClient (hexgrad/Kokoro-82M via API)
|
| 284 |
# "chatterbox" β Chatterbox Turbo (local, needs GPU for speed)
|
| 285 |
# "voxcpm2" β VoxCPM2 (local, voice design mode)
|
| 286 |
-
# "elevenlabs" β ElevenLabs API (needs ELEVENLABS_API_KEY)
|
| 287 |
#
|
| 288 |
# If TTS_BACKEND is not set, backends are tried in this order:
|
| 289 |
# hf β voxcpm2 β chatterbox β elevenlabs
|
|
@@ -300,11 +301,11 @@ VOICE_DESCRIPTIONS = {
|
|
| 300 |
|
| 301 |
# HF InferenceClient voice IDs (Kokoro voices, used with hf backend)
|
| 302 |
HF_VOICE_IDS = {
|
| 303 |
-
"hilarious": "
|
| 304 |
-
"tragic": "
|
| 305 |
-
"unhinged": "
|
| 306 |
-
"awkward": "
|
| 307 |
-
"chaotic": "
|
| 308 |
}
|
| 309 |
|
| 310 |
|
|
|
|
| 23 |
api_key=token,
|
| 24 |
)
|
| 25 |
|
| 26 |
+
LLM_MODEL_Nemo ="nvidia/Llama-3.1-Nemotron-8B-UltraLong-1M-Instruct:featherless-ai"
|
| 27 |
LLM_MODEL = "Qwen/Qwen2.5-7B-Instruct:together"
|
| 28 |
|
| 29 |
|
|
|
|
| 85 |
"name": "Dave",
|
| 86 |
"show": "The Weekly Wheeze",
|
| 87 |
"desc": "42, raspy, laughs mid-sentence",
|
| 88 |
+
"voice": "Mr_Quagmire",
|
| 89 |
"style": "comedic stand-up energy, warm and relatable",
|
| 90 |
"sfx": ["drumroll.wav", "crowd_react.wav"],
|
| 91 |
},
|
|
|
|
| 93 |
"name": "Elena",
|
| 94 |
"show": "Corporate Tears",
|
| 95 |
"desc": "31, soft, melancholic sighs",
|
| 96 |
+
"voice": "Sarah",
|
| 97 |
"style": "melancholic NPR storyteller, gentle and empathetic",
|
| 98 |
"sfx": ["violin_sting.wav"],
|
| 99 |
},
|
|
|
|
| 101 |
"name": "Marcus",
|
| 102 |
"show": "Officially Unhinged",
|
| 103 |
"desc": "55, baritone, barely-contained rage",
|
| 104 |
+
"voice": "Cho",
|
| 105 |
"style": "barely restrained fury, dark comedy, gravelly intensity",
|
| 106 |
"sfx": ["thunder.wav", "glass_break.wav", "dramatic_sting.wav"],
|
| 107 |
},
|
|
|
|
| 109 |
"name": "Priya",
|
| 110 |
"show": "Please Stop Talking",
|
| 111 |
"desc": "27, fast talker, nervous giggle",
|
| 112 |
+
"voice": "nicole",
|
| 113 |
"style": "rapid nervous energy, second-hand cringe expert",
|
| 114 |
"sfx": [],
|
| 115 |
},
|
|
|
|
| 117 |
"name": "Rex",
|
| 118 |
"show": "Total System Failure",
|
| 119 |
"desc": "???, manic, breaks 4th wall",
|
| 120 |
+
"voice": "max",
|
| 121 |
"style": "completely unhinged, 4th-wall breaking, reality-questioning",
|
| 122 |
"sfx": ["static.wav", "dial_up.wav", "system_error.wav", "vhs_rewind.wav"],
|
| 123 |
},
|
|
|
|
| 284 |
# "hf" β HuggingFace InferenceClient (hexgrad/Kokoro-82M via API)
|
| 285 |
# "chatterbox" β Chatterbox Turbo (local, needs GPU for speed)
|
| 286 |
# "voxcpm2" β VoxCPM2 (local, voice design mode)
|
| 287 |
+
# Fallback "elevenlabs" β ElevenLabs API (needs ELEVENLABS_API_KEY)
|
| 288 |
#
|
| 289 |
# If TTS_BACKEND is not set, backends are tried in this order:
|
| 290 |
# hf β voxcpm2 β chatterbox β elevenlabs
|
|
|
|
| 301 |
|
| 302 |
# HF InferenceClient voice IDs (Kokoro voices, used with hf backend)
|
| 303 |
HF_VOICE_IDS = {
|
| 304 |
+
"hilarious": "Mr_Quagmire",
|
| 305 |
+
"tragic": "Sarah",
|
| 306 |
+
"unhinged": "Cho",
|
| 307 |
+
"awkward": "nicole",
|
| 308 |
+
"chaotic": "max",
|
| 309 |
}
|
| 310 |
|
| 311 |
|