Commit ·
ffcd201
1
Parent(s): e38afe2
Replace model selection with specialized coding models
Browse filesUpdated to 4 specialized models:
- WhiteRabbitNeo 2.5 Qwen 2.5 Coder 7B: Advanced coding assistant with security focus
- Gemma 3 Prompt Coder 270m: Ultra-fast lightweight coding specialist
- DeepSeek V4 Pro: Advanced multimodal reasoning model
- Qwen 3.6 35B A3B: Large model with aggressive reasoning
Each model has custom system prompts optimized for their capabilities.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
app.py
CHANGED
|
@@ -15,61 +15,33 @@ llm_model = None
|
|
| 15 |
|
| 16 |
# Comprehensive model configurations
|
| 17 |
MODELS = {
|
| 18 |
-
"
|
| 19 |
-
"filename": "
|
| 20 |
-
"repo_id": "
|
| 21 |
-
"system_prompt": "You are
|
| 22 |
"formatter": "CHATML",
|
| 23 |
-
"description": "
|
| 24 |
},
|
| 25 |
-
"
|
| 26 |
-
"filename": "
|
| 27 |
-
"repo_id": "
|
| 28 |
-
"system_prompt": "You are
|
| 29 |
"formatter": "CHATML",
|
| 30 |
-
"description": "
|
| 31 |
},
|
| 32 |
-
"
|
| 33 |
-
"filename": "
|
| 34 |
-
"repo_id": "
|
| 35 |
-
"system_prompt": "You are
|
| 36 |
"formatter": "CHATML",
|
| 37 |
-
"description": "
|
| 38 |
},
|
| 39 |
-
"
|
| 40 |
-
"filename": "
|
| 41 |
-
"repo_id": "
|
| 42 |
-
"system_prompt": "You are
|
| 43 |
"formatter": "CHATML",
|
| 44 |
-
"description": "
|
| 45 |
-
},
|
| 46 |
-
"Phi-3 Medium 128K": {
|
| 47 |
-
"filename": "Phi-3-medium-128k-instruct-Q5_K_M.gguf",
|
| 48 |
-
"repo_id": "microsoft/Phi-3-medium-128k-instruct-GGUF",
|
| 49 |
-
"system_prompt": "You are a helpful AI assistant with extensive knowledge. Provide detailed, accurate responses.",
|
| 50 |
-
"formatter": "CHATML",
|
| 51 |
-
"description": "Long context window (128K tokens)"
|
| 52 |
-
},
|
| 53 |
-
"Gemma 2 9B IT": {
|
| 54 |
-
"filename": "gemma-2-9b-it-Q5_K_M.gguf",
|
| 55 |
-
"repo_id": "mradermacher/gemma-2-9b-it-GGUF",
|
| 56 |
-
"system_prompt": "You are a helpful, harmless, and honest AI assistant. Provide clear, well-reasoned responses.",
|
| 57 |
-
"formatter": "CHATML",
|
| 58 |
-
"description": "Google's efficient instruction-tuned model"
|
| 59 |
-
},
|
| 60 |
-
"Qwen2.5 7B Instruct": {
|
| 61 |
-
"filename": "Qwen2.5-7B-Instruct-Q5_K_M.gguf",
|
| 62 |
-
"repo_id": "mradermacher/Qwen2.5-7B-Instruct-GGUF",
|
| 63 |
-
"system_prompt": "You are a helpful AI assistant named Qwen, created by Alibaba Cloud. Provide accurate, thoughtful responses.",
|
| 64 |
-
"formatter": "CHATML",
|
| 65 |
-
"description": "Strong multilingual capabilities"
|
| 66 |
-
},
|
| 67 |
-
"Yi 1.5 34B": {
|
| 68 |
-
"filename": "dolphin-2.9.1-yi-1.5-34b-Q6_K.gguf",
|
| 69 |
-
"repo_id": "bartowski/dolphin-2.9.1-yi-1.5-34b-GGUF",
|
| 70 |
-
"system_prompt": "You are Dolphin, an AI assistant trained for reasoning and analysis. Provide detailed, well-reasoned responses.",
|
| 71 |
-
"formatter": "CHATML",
|
| 72 |
-
"description": "Strong reasoning with 34B parameters"
|
| 73 |
}
|
| 74 |
}
|
| 75 |
|
|
|
|
| 15 |
|
| 16 |
# Comprehensive model configurations
|
| 17 |
MODELS = {
|
| 18 |
+
"WhiteRabbitNeo 2.5 Qwen 2.5 Coder 7B": {
|
| 19 |
+
"filename": "WhiteRabbitNeo-2.5-Qwen-2.5-Coder-7B-OBLITERATED-i1-Q5_K_M.gguf",
|
| 20 |
+
"repo_id": "mradermacher/WhiteRabbitNeo-2.5-Qwen-2.5-Coder-7B-OBLITERATED-i1-GGUF",
|
| 21 |
+
"system_prompt": "You are WhiteRabbitNeo, an advanced AI coding assistant with deep expertise in software development, security analysis, and problem-solving. You provide detailed, accurate responses with proper code examples and thorough explanations.",
|
| 22 |
"formatter": "CHATML",
|
| 23 |
+
"description": "Advanced coding assistant with security focus"
|
| 24 |
},
|
| 25 |
+
"Gemma 3 Prompt Coder 270m": {
|
| 26 |
+
"filename": "Gemma-3-Prompt-Coder-270m-it-Uncensored-Q8_0.gguf",
|
| 27 |
+
"repo_id": "mradermacher/Gemma-3-Prompt-Coder-270m-it-Uncensored-GGUF",
|
| 28 |
+
"system_prompt": "You are Gemma 3 Prompt Coder, a lightweight but powerful AI assistant specialized in coding and technical tasks. Provide clear, accurate responses with well-formatted code examples.",
|
| 29 |
"formatter": "CHATML",
|
| 30 |
+
"description": "Ultra-fast lightweight coding specialist"
|
| 31 |
},
|
| 32 |
+
"DeepSeek V4 Pro": {
|
| 33 |
+
"filename": "DeepSeek-V4-Pro-Q5_K_M.gguf",
|
| 34 |
+
"repo_id": "unsloth/DeepSeek-V4-Pro-GGUF",
|
| 35 |
+
"system_prompt": "You are DeepSeek V4 Pro, an advanced AI assistant with extensive knowledge across multiple domains. Provide detailed, accurate, and well-reasoned responses with proper analysis and explanations.",
|
| 36 |
"formatter": "CHATML",
|
| 37 |
+
"description": "Advanced multimodal reasoning model"
|
| 38 |
},
|
| 39 |
+
"Qwen 3.6 35B A3B Uncensored": {
|
| 40 |
+
"filename": "Qwen3.6-35B-A3B-Uncensored-Q5_K_M.gguf",
|
| 41 |
+
"repo_id": "HauhauCS/Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-GGUF",
|
| 42 |
+
"system_prompt": "You are Qwen 3.6, an advanced AI assistant with aggressive reasoning capabilities and extensive knowledge. Provide direct, detailed responses with thorough analysis and strong reasoning.",
|
| 43 |
"formatter": "CHATML",
|
| 44 |
+
"description": "Large model with aggressive reasoning"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 45 |
}
|
| 46 |
}
|
| 47 |
|