https://github.com/gitstq/llmpick
๐ฏ ๆบ่ฝๆฌๅฐLLM้ๅๅฉๆ - Intelligent Local LLM Selector and Recommendation Engine
https://github.com/gitstq/llmpick
Last synced: 3 days ago
JSON representation
๐ฏ ๆบ่ฝๆฌๅฐLLM้ๅๅฉๆ - Intelligent Local LLM Selector and Recommendation Engine
- Host: GitHub
- URL: https://github.com/gitstq/llmpick
- Owner: gitstq
- License: mit
- Created: 2026-05-19T10:22:37.000Z (about 1 month ago)
- Default Branch: main
- Last Pushed: 2026-05-19T10:26:55.000Z (about 1 month ago)
- Last Synced: 2026-05-19T13:08:06.213Z (about 1 month ago)
- Language: Python
- Size: 20.5 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# ๐ฏ LLMPick
**ๆบ่ฝๆฌๅฐLLM้ๅๅฉๆ | Intelligent Local LLM Selector**
็ฎไฝไธญๆ โข
English
[](https://www.python.org/)
[](LICENSE)
[](https://github.com/gitstq/llmpick)
---
**LLMPick** ๆฏไธๆฌพๆบ่ฝๆฌๅฐๅคง่ฏญ่จๆจกๅ๏ผLLM๏ผ้ๅๅฉๆ๏ผไธไธบไธญๆ็จๆทไผๅใๅฎ่ฝๅค่ชๅจๆฃๆตๆจ็็กฌไปถ้
็ฝฎ๏ผๅนถๆ นๆฎๆง่ฝใๆพๅญใไฝฟ็จๅบๆฏ็ญๅค็ปดๅบฆๅ ็ด ๏ผๆบ่ฝๆจ่ๆ้ๅ็ๆฌๅฐLLMๆจกๅใ
### ๐ก ่งฃๅณ็็็น
- ๐ค **ๆจกๅ้ๆฉๅฐ้พ** - ้ขๅฏนๆฐ็พไธชๅผๆบๆจกๅไธ็ฅๅฆไฝ้ๆฉ
- ๐พ **็กฌไปถ้้
่ฟท่ซ** - ไธๆธ
ๆฅ่ชๅทฑ็่ฎพๅค่ฝ่ฟ่กไปไนๆจกๅ
- ๐ **ไธญๆๆฏๆไธ่ถณ** - ๅพๅคๅทฅๅ
ทๅฏนไธญๆๆจกๅๆฏๆไธๅคๅๅฅฝ
- โก **ๆง่ฝ้ขไผฐ็ผบๅคฑ** - ๆ ๆณ้ขไผฐๆจกๅ่ฟ่ก้ๅบฆๅไฝ้ช
### โจ ่ช็ ๅทฎๅผๅไบฎ็น
1. ๐จ๐ณ **ไธญๆไผๅ
** - ไผๅ
ๆจ่ไธญๆ่ฝๅไผ็ง็ๆจกๅ๏ผQwenใDeepSeekใYi็ญ๏ผ
2. ๐ฏ **ๆบ่ฝ่ฏๅ** - ๅบไบBenchmarkใ้ๅ่ดจ้ใ็กฌไปถ้้
ๅบฆ็ปผๅ่ฏๅ
3. ๐ **้ๅบฆ้ขไผฐ** - ๆบ่ฝไผฐ็ฎๆจกๅๆจ็้ๅบฆ๏ผtokens/็ง๏ผ
4. ๐ฅ๏ธ **็พ่งTUI** - ๅบไบRich็ไผ้
็ป็ซฏ็้ข
5. ๐ง **ๅคๅ็ซฏๆฏๆ** - ไธ้ฎ็ๆOllama/LM Studio/llama.cppๅฝไปค
---
## โจ ๆ ธๅฟ็นๆง
| ็นๆง | ๆ่ฟฐ | ็ถๆ |
|------|------|------|
| ๐ **่ชๅจ็กฌไปถๆฃๆต** | ๆบ่ฝ่ฏๅซGPUใCPUใๅ
ๅญ้
็ฝฎ | โ
|
| ๐ฏ **ๆบ่ฝๆจกๅๆจ่** | ๅบไบ็กฌไปถๆง่ฝๅน้
ๆไฝณๆจกๅ | โ
|
| ๐จ๐ณ **ไธญๆไผๅ** | ไผๅ
ๆจ่ไธญๆ่ฝๅไผ็ง็ๆจกๅ | โ
|
| โก **้ๅบฆไผฐ็ฎ** | ้ขไผฐๆจกๅๆจ็้ๅบฆ | โ
|
| ๐ **ๅคๆจกๅๅฏนๆฏ** | ็ด่งๅฏนๆฏไธๅๆจกๅ็นๆง | โ
|
| ๐ฅ๏ธ **็พ่งTUI** | ไผ้
็็ป็ซฏไบคไบ็้ข | โ
|
| ๐ง **ๅคๅ็ซฏๆฏๆ** | ๆฏๆOllamaใLM Studio็ญ | โ
|
| ๐ท๏ธ **ๆ ็ญพ่ฟๆปค** | ๆไธญๆ/ไปฃ็ /ๅคๆจกๆ็ญ่ฟๆปค | โ
|
---
## ๐ ๅฟซ้ๅผๅง
### ็ฏๅข่ฆๆฑ
- **Python**: 3.9 ๆๆด้ซ็ๆฌ
- **ๆไฝ็ณป็ป**: Windows 10+ / macOS 10.15+ / Linux
- **ๅฏ้**: NVIDIA GPU๏ผCUDA๏ผใApple Silicon๏ผMetal๏ผ
### ๅฎ่ฃ
```bash
# ไฝฟ็จ pip ๅฎ่ฃ
pip install llmpick
# ๆไฝฟ็จ uv๏ผๆจ่๏ผ
uv tool install llmpick
```
### ๅบ็ก็จๆณ
```bash
# ๐ ๆฃๆต็กฌไปถไฟกๆฏ
llmpick detect
# ๐ฏ ่ทๅๆจกๅๆจ่๏ผ้ป่ฎคๅ5ไธช๏ผ
llmpick recommend
# ๐ฏ ่ทๅๅ10ไธชๆจ่๏ผไผๅ
ไธญๆๆจกๅ
llmpick recommend --top 10 --chinese
# ๐ ๅๅบๆๆๅฏ็จๆจกๅ
llmpick list
# ๐ ไป
ๅๅบไธญๆไผๅๆจกๅ
llmpick list --chinese-only
# โน๏ธ ๆฅ็ๆจกๅ่ฏฆๆ
llmpick info qwen2.5-7b
# ๐ ๅฏนๆฏๅคไธชๆจกๅ
llmpick compare qwen2.5-7b llama-3.1-8b phi-4
```
---
## ๐ ่ฏฆ็ปไฝฟ็จๆๅ
### ็กฌไปถๆฃๆต
```bash
$ llmpick detect
โญโโโโโโโโโโโโโโโโโโโโโโโโโฎ
โ ๐ ๆญฃๅจๆฃๆต็กฌไปถไฟกๆฏ... โ
โฐโโโโโโโโโโโโโโโโโโโโโโโโโฏ
CPU: Intel(R) Core(TM) i9-13900K
CPUๆ ธๅฟๆฐ: 24
ๅ
ๅญ: 32.0 GB
็ฃ็ๅฏ็จ: 500.0 GB
ๆไฝ็ณป็ป: Linux 6.5.0
GPUไฟกๆฏ:
[1] NVIDIA GeForce RTX 4090
ๆพๅญ: 24.0 GB
่ฎก็ฎ่ฝๅ: 8.9
ๅ ้ๆฏๆ:
CUDA: โ
Metal: โ
ROCm: โ
```
### ๆจกๅๆจ่
```bash
$ llmpick recommend
โญโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฎ
โ ๐ฏ ๆญฃๅจๅๆ็กฌไปถๅนถๆจ่ๆไฝณLLMๆจกๅ... โ
โฐโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฏ
๐ฏ ๆจ่ๆจกๅๅ่กจ
โญโโโโโโโฌโโโโโโโโโโโโโโโโโโโโโโโฌโโโโโโโโโโโฌโโโโโโโโโโโฌโโโโโโโโโโโฌโโโโโโโโโโโโโฌโโโโโโโโโโโโโโโฌโโโโโโโโโโโโโโโโโโโโโโโโโโโฎ
โ ๆๅ โ ๆจกๅ โ ๅๆฐ้ โ ๅคงๅฐ โ ่ฏๅ โ ้้
ๅบฆ โ ้ขไผฐ้ๅบฆ โ ๆจ่็็ฑ โ
โโโโโโโโผโโโโโโโโโโโโโโโโโโโโโโโผโโโโโโโโโโโผโโโโโโโโโโโผโโโโโโโโโโโผโโโโโโโโโโโโโผโโโโโโโโโโโโโโโผโโโโโโโโโโโโโโโโโโโโโโโโโโโค
โ 1 โ Qwen2.5-14B [CN] โ 14B โ 9.0GB โ 76.0 โ ่ฏๅฅฝ โ 45.0 t/s โ ่ฏๅฅฝ้้
๏ผไธญๆไผๅ โ
โ 2 โ Qwen2.5-7B [CN] โ 7B โ 4.5GB โ 70.0 โ ๅฎ็พ โ 85.0 t/s โ ๅฎ็พ้้
๏ผไธญๆไผๅ โ
โ 3 โ Llama 3.1 8B โ 8B โ 5.0GB โ 68.0 โ ๅฎ็พ โ 80.0 t/s โ ๅฎ็พ้้
โ
โฐโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโดโโโโโโโโโโโดโโโโโโโโโโโดโโโโโโโโโโโโโดโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโฏ
```
### ๆจกๅๅฏนๆฏ
```bash
$ llmpick compare qwen2.5-7b llama-3.1-8b phi-4
โญโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฎ
โ ๐ ๆจกๅๅฏนๆฏ โ
โโโโโโโโโโโโโโโฌโโโโโโโโโโโโโโโโโโฌโโโโโโโโโโโโโโโโโโฌโโโโโโโโโโโโโค
โ ๅฑๆง โ Qwen2.5-7B โ Llama 3.1 8B โ Phi-4 โ
โโโโโโโโโโโโโโโผโโโโโโโโโโโโโโโโโโผโโโโโโโโโโโโโโโโโโผโโโโโโโโโโโโโค
โ ๅๆฐ้ โ 7B โ 8B โ 14B โ
โ ๆจกๅๅคงๅฐ โ 4.5GB โ 5.0GB โ 8.0GB โ
โ ้ๅ โ Q4_K_M โ Q4_K_M โ Q4_K_M โ
โ ไธไธๆ โ 32K โ 128K โ 16K โ
โ ไธญๆไผๅ โ โ โ โ โ โ โ
โ ไปฃ็ ไผๅ โ โ โ โ โ โ โ
โ ๅคๆจกๆ โ โ โ โ โ โ โ
โ ่ฏๅ โ 70.0 โ 68.0 โ 75.0 โ
โฐโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโฏ
```
---
## ๐ ๆฏๆ็ๆจกๅ
### ไธญๆไผๅๆจกๅ
| ๆจกๅ | ๅๆฐ้ | ๅคงๅฐ | ไธไธๆ | ็น็น |
|------|--------|------|--------|------|
| Qwen2.5 | 0.5B-32B | 0.4-20GB | 32K | ้ฟ้ๅทดๅทด๏ผไธญๆๆๅผบ |
| DeepSeek | 6.7B-7B | 4.0-4.5GB | 16K | ๆทฑๅบฆๆฑ็ดข๏ผไปฃ็ ไผ็ง |
| Yi-1.5 | 6B-9B | 3.8-5.8GB | 4K | ้ถไธไธ็ฉ๏ผไธญๆๅฏน่ฏ |
### ๅฝ้
ไธปๆตๆจกๅ
| ๆจกๅ | ๅๆฐ้ | ๅคงๅฐ | ไธไธๆ | ็น็น |
|------|--------|------|--------|------|
| Llama 3.1/3.2 | 1B-8B | 0.7-5GB | 128K | Meta๏ผๅค่ฏญ่จ |
| Phi-4 | 14B | 8GB | 16K | ๅพฎ่ฝฏ๏ผๅฐๅทง้ซๆ |
| Gemma 2 | 2B-9B | 1.6-6GB | 8K | Google๏ผ็ฅ่ฏไธฐๅฏ |
| Mistral | 7B | 4.5GB | 32K | Mistral AI๏ผๆจ็ๅผบ |
---
## ๐ก ่ฎพ่ฎกๆ่ทฏไธ่ฟญไปฃ่งๅ
### ๆๆฏ้ๅๅๅ
- **Python 3.9+**: ๅ
ผ้กพๅ
ผๅฎนๆงไธ็ฐไปฃ็นๆง
- **Typer**: ็ฑปๅๅฎๅ
จ็CLIๆกๆถ๏ผ่ชๅจ็ๆๅธฎๅฉๆๆกฃ
- **Rich**: ๅผบๅคง็็ป็ซฏ็พๅๅบ๏ผๆฏๆ่กจๆ ผใ้ขๆฟใ่ฟๅบฆๆก
- **Pydantic**: ๆฐๆฎ้ช่ฏๅๅบๅๅ
- **HuggingFace**: ๆจกๅๆฐๆฎๆฅๆบ
### ่ฏๅ็ฎๆณ
```
ๆ็ป่ฏๅ = Benchmarkๅๆฐ ร ้ๅ่ดจ้็ณปๆฐ ร ้้
ๅบฆ็ณปๆฐ
ๅ
ถไธญ:
- ้ๅ่ดจ้: Q4_K_M=1.0, Q5_K_M=1.05, Q8_0=1.10
- ้้
ๅบฆ: ๅฎ็พ=1.0, ่ฏๅฅฝ=0.95, ้จๅ=0.85, ็ดงๅผ =0.75, CPU=0.60
```
### ๅ็ปญ่ฟญไปฃ่ฎกๅ
- [ ] ๆฏๆๆดๅคๆจกๅ๏ผGLM-4ใBaichuan็ญ๏ผ
- [ ] ๆทปๅ ๆจกๅไธ่ฝฝ่ฟๅบฆๆพ็คบ
- [ ] ๆฏๆ่ชๅฎไนๆจกๅ้
็ฝฎ
- [ ] ้ๆๆดๅคๆจ็ๅ็ซฏ๏ผvLLMใTGI็ญ๏ผ
- [ ] Web UI็้ข
- [ ] ๆจกๅๆง่ฝๅฎๆตๆฐๆฎ
---
## ๐ฆ ๆๅ
ไธ้จ็ฝฒๆๅ
### ๅผๅๅฎ่ฃ
```bash
git clone https://github.com/gitstq/llmpick.git
cd llmpick
pip install -e ".[dev]"
```
### ่ฟ่กๆต่ฏ
```bash
pytest tests/ -v
```
### ๆๅปบๅๅธ
```bash
# ๆๅปบ wheel
python -m build
# ไธไผ ๅฐ PyPI
python -m twine upload dist/*
```
---
## ๐ค ่ดก็ฎๆๅ
ๆฌข่ฟๆไบค Issue ๅ Pull Request๏ผ
### ๆไบค่ง่
- `feat:` ๆฐๅขๅ่ฝ
- `fix:` ไฟฎๅค้ฎ้ข
- `docs:` ๆๆกฃๆดๆฐ
- `refactor:` ไปฃ็ ้ๆ
- `test:` ๆต่ฏ็ธๅ
ณ
---
## ๐ ๅผๆบๅ่ฎฎ
ๆฌ้กน็ฎ้็จ [MIT License](LICENSE) ๅผๆบๅ่ฎฎใ
---
**Made with โค๏ธ by LLMPick Team**
ๅฆๆ่ฟไธช้กน็ฎๅฏนๆจๆๅธฎๅฉ๏ผ่ฏท็ปไธช โญ Star๏ผ
---
## ๐ Introduction (English)
**LLMPick** is an intelligent local Large Language Model (LLM) selector optimized for Chinese users. It automatically detects your hardware configuration and intelligently recommends the most suitable local LLM models based on multiple dimensions such as performance, VRAM, and usage scenarios.
### โจ Key Features
- ๐ **Automatic Hardware Detection** - Intelligently identify GPU, CPU, and memory configurations
- ๐ฏ **Smart Model Recommendation** - Match the best model based on hardware performance
- ๐จ๐ณ **Chinese Optimized** - Prioritize models with excellent Chinese capabilities
- โก **Speed Estimation** - Estimate model inference speed
- ๐ **Multi-Model Comparison** - Intuitively compare different model features
- ๐ฅ๏ธ **Beautiful TUI** - Elegant terminal interface based on Rich
- ๐ง **Multi-Backend Support** - Support Ollama, LM Studio, etc.
---
## ๐ Quick Start
### Requirements
- **Python**: 3.9 or higher
- **OS**: Windows 10+ / macOS 10.15+ / Linux
- **Optional**: NVIDIA GPU (CUDA), Apple Silicon (Metal)
### Installation
```bash
# Using pip
pip install llmpick
# Or using uv (recommended)
uv tool install llmpick
```
### Basic Usage
```bash
# ๐ Detect hardware info
llmpick detect
# ๐ฏ Get model recommendations (default top 5)
llmpick recommend
# ๐ฏ Get top 10 recommendations, prioritize Chinese models
llmpick recommend --top 10 --chinese
# ๐ List all available models
llmpick list
# ๐ List only Chinese-optimized models
llmpick list --chinese-only
# โน๏ธ View model details
llmpick info qwen2.5-7b
# ๐ Compare multiple models
llmpick compare qwen2.5-7b llama-3.1-8b phi-4
```
---
## ๐ Supported Models
### Chinese-Optimized Models
| Model | Parameters | Size | Context | Features |
|-------|------------|------|---------|----------|
| Qwen2.5 | 0.5B-32B | 0.4-20GB | 32K | Alibaba, Best Chinese |
| DeepSeek | 6.7B-7B | 4.0-4.5GB | 16K | DeepSeek, Code Expert |
| Yi-1.5 | 6B-9B | 3.8-5.8GB | 4K | 01.AI, Chinese Chat |
### International Models
| Model | Parameters | Size | Context | Features |
|-------|------------|------|---------|----------|
| Llama 3.1/3.2 | 1B-8B | 0.7-5GB | 128K | Meta, Multilingual |
| Phi-4 | 14B | 8GB | 16K | Microsoft, Efficient |
| Gemma 2 | 2B-9B | 1.6-6GB | 8K | Google, Knowledge Rich |
| Mistral | 7B | 4.5GB | 32K | Mistral AI, Strong Reasoning |
---
## ๐ก Design Philosophy
### Scoring Algorithm
```
Final Score = Benchmark Score ร Quantization Quality ร Fit Factor
Where:
- Quantization: Q4_K_M=1.0, Q5_K_M=1.05, Q8_0=1.10
- Fit: Perfect=1.0, Good=0.95, Partial=0.85, Tight=0.75, CPU=0.60
```
### Roadmap
- [ ] Support more models (GLM-4, Baichuan, etc.)
- [ ] Add model download progress display
- [ ] Support custom model configuration
- [ ] Integrate more inference backends (vLLM, TGI, etc.)
- [ ] Web UI interface
- [ ] Real model performance benchmark data
---
## ๐ค Contributing
Issues and Pull Requests are welcome!
### Commit Convention
- `feat:` New feature
- `fix:` Bug fix
- `docs:` Documentation update
- `refactor:` Code refactoring
- `test:` Test related
---
## ๐ License
This project is licensed under the [MIT License](LICENSE).
---
**Made with โค๏ธ by LLMPick Team**
If this project helps you, please give it a โญ Star!