isagellm 0.1.0.6__cp311-none-any.whl
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.
- isagellm-0.1.0.6.dist-info/METADATA +318 -0
- isagellm-0.1.0.6.dist-info/RECORD +11 -0
- isagellm-0.1.0.6.dist-info/WHEEL +5 -0
- isagellm-0.1.0.6.dist-info/entry_points.txt +3 -0
- isagellm-0.1.0.6.dist-info/top_level.txt +1 -0
- sagellm/__init__.py +158 -0
- sagellm/__init__.pyc +0 -0
- sagellm/__pycache__/__init__.cpython-311.pyc +0 -0
- sagellm/__pycache__/cli.cpython-311.pyc +0 -0
- sagellm/cli.pyc +0 -0
- sagellm/py.typed +2 -0
|
@@ -0,0 +1,318 @@
|
|
|
1
|
+
Metadata-Version: 2.1
|
|
2
|
+
Name: isagellm
|
|
3
|
+
Version: 0.1.0.6
|
|
4
|
+
Summary: sageLLM: Modular LLM inference engine for domestic computing power (Huawei Ascend, NVIDIA)
|
|
5
|
+
Author: IntelliStream Team
|
|
6
|
+
License: Proprietary - IntelliStream
|
|
7
|
+
Project-URL: Homepage, https://github.com/IntelliStream/sagellm
|
|
8
|
+
Project-URL: Documentation, https://github.com/IntelliStream/sagellm#readme
|
|
9
|
+
Project-URL: Repository, https://github.com/IntelliStream/sagellm
|
|
10
|
+
Keywords: llm,inference,ascend,huawei,npu,cuda,domestic
|
|
11
|
+
Classifier: Development Status :: 3 - Alpha
|
|
12
|
+
Classifier: Programming Language :: Python :: 3
|
|
13
|
+
Classifier: Programming Language :: Python :: 3.10
|
|
14
|
+
Classifier: Programming Language :: Python :: 3.11
|
|
15
|
+
Classifier: Programming Language :: Python :: 3.12
|
|
16
|
+
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
|
|
17
|
+
Requires-Python: ==3.11.*
|
|
18
|
+
Description-Content-Type: text/markdown
|
|
19
|
+
Requires-Dist: isagellm-protocol<0.2.0,>=0.1.0
|
|
20
|
+
Requires-Dist: isagellm-backend<0.2.0,>=0.1.1.3
|
|
21
|
+
Requires-Dist: isagellm-core<0.2.0,>=0.1.0.2
|
|
22
|
+
Requires-Dist: isagellm-control-plane<0.2.0,>=0.1.0.2
|
|
23
|
+
Requires-Dist: isagellm-gateway<0.2.0,>=0.1.0.2
|
|
24
|
+
Requires-Dist: isagellm-kv-cache<0.2.0,>=0.1.0
|
|
25
|
+
Requires-Dist: isagellm-comm<0.2.0,>=0.1.0
|
|
26
|
+
Requires-Dist: isagellm-compression<0.2.0,>=0.1.0
|
|
27
|
+
Requires-Dist: click>=8.0.0
|
|
28
|
+
Requires-Dist: rich>=13.0.0
|
|
29
|
+
Requires-Dist: pyyaml>=6.0.0
|
|
30
|
+
Provides-Extra: all
|
|
31
|
+
Requires-Dist: isagellm[benchmark,cuda]; extra == "all"
|
|
32
|
+
Provides-Extra: ascend
|
|
33
|
+
Requires-Dist: torch>=2.0.0; extra == "ascend"
|
|
34
|
+
Provides-Extra: benchmark
|
|
35
|
+
Requires-Dist: isagellm-benchmark>=0.1.0; extra == "benchmark"
|
|
36
|
+
Provides-Extra: cuda
|
|
37
|
+
Requires-Dist: torch>=2.0.0; extra == "cuda"
|
|
38
|
+
Provides-Extra: dev
|
|
39
|
+
Requires-Dist: pytest>=7.0.0; extra == "dev"
|
|
40
|
+
Requires-Dist: pytest-cov>=4.0.0; extra == "dev"
|
|
41
|
+
Requires-Dist: ruff>=0.8.0; extra == "dev"
|
|
42
|
+
Requires-Dist: mypy>=1.0.0; extra == "dev"
|
|
43
|
+
|
|
44
|
+
# sageLLM
|
|
45
|
+
|
|
46
|
+
<p align="center">
|
|
47
|
+
<strong>🚀 Modular LLM Inference Engine for Domestic Computing Power</strong>
|
|
48
|
+
</p>
|
|
49
|
+
|
|
50
|
+
<p align="center">
|
|
51
|
+
Ollama-like experience for Chinese hardware ecosystems (Huawei Ascend, NVIDIA)
|
|
52
|
+
</p>
|
|
53
|
+
|
|
54
|
+
---
|
|
55
|
+
|
|
56
|
+
## ✨ Features
|
|
57
|
+
|
|
58
|
+
- 🎯 **One-Click Install** - `pip install isagellm` gets you started immediately
|
|
59
|
+
- 🔌 **Mock-First** - Test without GPU, perfect for CI/CD
|
|
60
|
+
- 🇨🇳 **Domestic Hardware** - First-class support for Huawei Ascend NPU
|
|
61
|
+
- 📊 **Observable** - Built-in metrics (TTFT, TBT, throughput, KV usage)
|
|
62
|
+
- 🧩 **Plugin System** - Extend with custom backends and engines
|
|
63
|
+
|
|
64
|
+
## 📦 Quick Install
|
|
65
|
+
|
|
66
|
+
```bash
|
|
67
|
+
# Install sageLLM (includes mock backend, no GPU required)
|
|
68
|
+
pip install isagellm
|
|
69
|
+
|
|
70
|
+
# With Control Plane (request routing & scheduling)
|
|
71
|
+
pip install 'isagellm[control-plane]'
|
|
72
|
+
|
|
73
|
+
# With API Gateway (OpenAI-compatible REST API)
|
|
74
|
+
pip install 'isagellm[gateway]'
|
|
75
|
+
|
|
76
|
+
# Full server (Control Plane + Gateway)
|
|
77
|
+
pip install 'isagellm[server]'
|
|
78
|
+
|
|
79
|
+
# With CUDA support
|
|
80
|
+
pip install 'isagellm[cuda]'
|
|
81
|
+
|
|
82
|
+
# All features
|
|
83
|
+
pip install 'isagellm[all]'
|
|
84
|
+
```
|
|
85
|
+
|
|
86
|
+
## 🚀 Quick Start
|
|
87
|
+
|
|
88
|
+
### 启动模式选择
|
|
89
|
+
|
|
90
|
+
sageLLM 支持两种启动模式，满足不同场景需求：
|
|
91
|
+
|
|
92
|
+
| 模式 | 使用场景 | 依赖 | 命令示例 |
|
|
93
|
+
|------|---------|------|---------|
|
|
94
|
+
| **Mock** | CI/测试/本地开发 | 无需 GPU | `sage-llm serve --mock` |
|
|
95
|
+
| **生产** | 真实推理服务 | GPU/CPU 后端 | `sage-llm serve --control-plane` |
|
|
96
|
+
|
|
97
|
+
**⚠️ Fail-Fast 保证**：非 mock 模式下，若依赖缺失或配置错误，系统将**立即报错退出**，不会静默回退到 mock 模式。
|
|
98
|
+
|
|
99
|
+
### CLI (like ollama)
|
|
100
|
+
|
|
101
|
+
```bash
|
|
102
|
+
# Show system info
|
|
103
|
+
sage-llm info
|
|
104
|
+
|
|
105
|
+
# Mock 模式（无 GPU 依赖）
|
|
106
|
+
sage-llm serve --mock
|
|
107
|
+
sage-llm run -p "What is LLM inference?" --mock
|
|
108
|
+
sage-llm demo --workload year1 --mock
|
|
109
|
+
|
|
110
|
+
# 生产模式（需要安装 control-plane）
|
|
111
|
+
# pip install 'isagellm[server]'
|
|
112
|
+
sage-llm serve --control-plane
|
|
113
|
+
sage-llm gateway --control-plane --port 8080
|
|
114
|
+
|
|
115
|
+
# 如果缺少依赖，将看到：
|
|
116
|
+
# ❌ Error: Control Plane required but not installed
|
|
117
|
+
# Install: pip install 'isagellm[control-plane]'
|
|
118
|
+
```
|
|
119
|
+
|
|
120
|
+
### Python API
|
|
121
|
+
|
|
122
|
+
```python
|
|
123
|
+
from sagellm import Request, MockEngine
|
|
124
|
+
|
|
125
|
+
# Create mock engine (no GPU needed)
|
|
126
|
+
engine = MockEngine()
|
|
127
|
+
|
|
128
|
+
# Run inference
|
|
129
|
+
request = Request(
|
|
130
|
+
request_id="demo-001",
|
|
131
|
+
prompt="Hello, world!",
|
|
132
|
+
max_tokens=128,
|
|
133
|
+
)
|
|
134
|
+
response = engine.generate(request)
|
|
135
|
+
|
|
136
|
+
print(f"Response: {response.text}")
|
|
137
|
+
print(f"TTFT: {response.metrics.ttft_ms:.2f} ms")
|
|
138
|
+
print(f"Throughput: {response.metrics.throughput_tps:.2f} tokens/s")
|
|
139
|
+
```
|
|
140
|
+
|
|
141
|
+
### Configuration
|
|
142
|
+
|
|
143
|
+
```yaml
|
|
144
|
+
# ~/.sage-llm/config.yaml
|
|
145
|
+
backend:
|
|
146
|
+
kind: mock # Options: mock, cpu, cuda, ascend
|
|
147
|
+
|
|
148
|
+
# Fail-fast 配置：如果指定了非 mock 后端但不可用，将报错退出
|
|
149
|
+
strict_mode: true # 默认为 true，符合申报书要求
|
|
150
|
+
|
|
151
|
+
engine:
|
|
152
|
+
kind: mock
|
|
153
|
+
model: Qwen/Qwen2-7B
|
|
154
|
+
|
|
155
|
+
# Mock 模式配置
|
|
156
|
+
mock:
|
|
157
|
+
enabled: false # true 时强制使用 mock，无论其他配置
|
|
158
|
+
deterministic: true # mock 输出是否固定（用于回归测试）
|
|
159
|
+
|
|
160
|
+
# 生产模式最低要求
|
|
161
|
+
production:
|
|
162
|
+
control_plane:
|
|
163
|
+
required: true # true 时缺少 control-plane 将报错（非 mock 模式）
|
|
164
|
+
endpoint: "localhost:8080"
|
|
165
|
+
backend:
|
|
166
|
+
required: true # true 时缺少真实后端将报错
|
|
167
|
+
fallback_to_mock: false # 禁止自动降级到 mock（fail-fast）
|
|
168
|
+
|
|
169
|
+
workload:
|
|
170
|
+
segments:
|
|
171
|
+
- short # 128 in → 128 out
|
|
172
|
+
- long # 2048 in → 512 out
|
|
173
|
+
- stress # concurrent requests
|
|
174
|
+
```
|
|
175
|
+
|
|
176
|
+
## 📊 Year 1 Demo Contract
|
|
177
|
+
|
|
178
|
+
sageLLM must produce these metrics for validation:
|
|
179
|
+
|
|
180
|
+
```json
|
|
181
|
+
{
|
|
182
|
+
"ttft_ms": 45.2,
|
|
183
|
+
"tbt_ms": 12.5,
|
|
184
|
+
"throughput_tps": 80.0,
|
|
185
|
+
"peak_mem_mb": 24576,
|
|
186
|
+
"kv_used_tokens": 4096,
|
|
187
|
+
"prefix_hit_rate": 0.85,
|
|
188
|
+
"evict_count": 3
|
|
189
|
+
}
|
|
190
|
+
```
|
|
191
|
+
|
|
192
|
+
Run validation:
|
|
193
|
+
```bash
|
|
194
|
+
sage-llm demo --workload year1 --output metrics.json
|
|
195
|
+
```
|
|
196
|
+
|
|
197
|
+
## 🏗️ Architecture
|
|
198
|
+
|
|
199
|
+
```
|
|
200
|
+
isagellm (umbrella package)
|
|
201
|
+
├── isagellm-protocol # Protocol v0.1 types
|
|
202
|
+
│ └── Request, Response, Metrics, Error, StreamEvent
|
|
203
|
+
├── isagellm-core # Runtime & Demo Runner
|
|
204
|
+
│ └── Config, Engine, Factory, DemoRunner
|
|
205
|
+
├── isagellm-backend # Hardware abstraction
|
|
206
|
+
│ └── BackendProvider, MockBackend, (CUDABackend, AscendBackend)
|
|
207
|
+
├── isagellm-control-plane # Request routing & scheduling (optional)
|
|
208
|
+
│ └── ControlPlaneManager, Router, Policies, Lifecycle
|
|
209
|
+
└── isagellm-gateway # OpenAI-compatible REST API (optional)
|
|
210
|
+
└── FastAPI server, /v1/chat/completions, Session management
|
|
211
|
+
```
|
|
212
|
+
|
|
213
|
+
## 🔧 Development
|
|
214
|
+
|
|
215
|
+
### Quick Setup (Development Mode)
|
|
216
|
+
|
|
217
|
+
```bash
|
|
218
|
+
# Clone all repositories
|
|
219
|
+
./scripts/clone-all-repos.sh
|
|
220
|
+
|
|
221
|
+
# Install all packages in editable mode
|
|
222
|
+
./quickstart.sh
|
|
223
|
+
|
|
224
|
+
# Open all repos in VS Code Multi-root Workspace
|
|
225
|
+
code sagellm.code-workspace
|
|
226
|
+
```
|
|
227
|
+
|
|
228
|
+
**📖 See [WORKSPACE_GUIDE.md](WORKSPACE_GUIDE.md) for Multi-root Workspace usage.**
|
|
229
|
+
|
|
230
|
+
### Testing
|
|
231
|
+
|
|
232
|
+
```bash
|
|
233
|
+
# Clone and setup
|
|
234
|
+
git clone https://github.com/IntelliStream/sagellm.git
|
|
235
|
+
cd sagellm
|
|
236
|
+
pip install -e ".[dev]"
|
|
237
|
+
|
|
238
|
+
# Run tests
|
|
239
|
+
pytest -v
|
|
240
|
+
|
|
241
|
+
# Format & lint
|
|
242
|
+
ruff format .
|
|
243
|
+
ruff check . --fix
|
|
244
|
+
|
|
245
|
+
# Type check
|
|
246
|
+
mypy src/sagellm/
|
|
247
|
+
|
|
248
|
+
# Verify dependency hierarchy
|
|
249
|
+
python scripts/verify_dependencies.py
|
|
250
|
+
```
|
|
251
|
+
|
|
252
|
+
### 📖 Development Resources
|
|
253
|
+
|
|
254
|
+
- **[DEPLOYMENT_GUIDE.md](docs/DEPLOYMENT_GUIDE.md)** - 完整部署与配置指南
|
|
255
|
+
- **[TROUBLESHOOTING.md](docs/TROUBLESHOOTING.md)** - 故障排查快速参考
|
|
256
|
+
- **[ENVIRONMENT_VARIABLES.md](docs/ENVIRONMENT_VARIABLES.md)** - 环境变量完整参考
|
|
257
|
+
- **[DEVELOPER_GUIDE.md](docs/DEVELOPER_GUIDE.md)** - 开发者指南
|
|
258
|
+
- **[WORKSPACE_GUIDE.md](docs/WORKSPACE_GUIDE.md)** - Multi-root Workspace 使用
|
|
259
|
+
- **[INFERENCE_FLOW.md](docs/INFERENCE_FLOW.md)** - 推理流程详解
|
|
260
|
+
- **[PR_CHECKLIST.md](docs/PR_CHECKLIST.md)** - Pull Request 检查清单
|
|
261
|
+
|
|
262
|
+
---
|
|
263
|
+
|
|
264
|
+
## 📚 Documentation Index
|
|
265
|
+
|
|
266
|
+
### 用户文档
|
|
267
|
+
- [快速开始](README.md#-quick-start) - 5 分钟上手
|
|
268
|
+
- [部署指南](docs/DEPLOYMENT_GUIDE.md) - 生产环境部署
|
|
269
|
+
- [配置参考](docs/DEPLOYMENT_GUIDE.md#配置文件说明) - 完整配置选项
|
|
270
|
+
- [环境变量](docs/ENVIRONMENT_VARIABLES.md) - 环境变量参考
|
|
271
|
+
- [故障排查](docs/TROUBLESHOOTING.md) - 常见问题解决
|
|
272
|
+
|
|
273
|
+
### 开发者文档
|
|
274
|
+
- [开发指南](docs/DEVELOPER_GUIDE.md) - 贡献代码
|
|
275
|
+
- [架构设计](README.md#-architecture) - 系统架构
|
|
276
|
+
- [Workspace 使用](docs/WORKSPACE_GUIDE.md) - Multi-root 工作区
|
|
277
|
+
- [PR 检查清单](docs/PR_CHECKLIST.md) - 提交前检查
|
|
278
|
+
|
|
279
|
+
### API 文档
|
|
280
|
+
- OpenAI 兼容 API - 参见 [sagellm-gateway](https://github.com/intellistream/sagellm-gateway)
|
|
281
|
+
- Python API - 参见 [API_REFERENCE.md](docs/API_REFERENCE.md)（待补充）
|
|
282
|
+
|
|
283
|
+
### 子包文档
|
|
284
|
+
- [sagellm-protocol](https://github.com/intellistream/sagellm-protocol) - 协议定义
|
|
285
|
+
- [sagellm-backend](https://github.com/intellistream/sagellm-backend) - 后端抽象
|
|
286
|
+
- [sagellm-core](https://github.com/intellistream/sagellm-core) - 引擎核心
|
|
287
|
+
- [sagellm-control-plane](https://github.com/intellistream/sagellm-control-plane) - 控制面
|
|
288
|
+
- [sagellm-gateway](https://github.com/intellistream/sagellm-gateway) - API 网关
|
|
289
|
+
- [sagellm-benchmark](https://github.com/intellistream/sagellm-benchmark) - 基准测试
|
|
290
|
+
|
|
291
|
+
- [**DEVELOPER_GUIDE.md**](DEVELOPER_GUIDE.md) - 架构规范与开发指南
|
|
292
|
+
- [**PR_CHECKLIST.md**](PR_CHECKLIST.md) - Pull Request 审查清单
|
|
293
|
+
- [**scripts/verify_dependencies.py**](scripts/verify_dependencies.py) - 依赖层次验证
|
|
294
|
+
|
|
295
|
+
## 📚 Package Details
|
|
296
|
+
|
|
297
|
+
| Package | PyPI Name | Import Name | Description |
|
|
298
|
+
|---------|-----------|-------------|-------------|
|
|
299
|
+
| sagellm | `isagellm` | `sagellm` | Umbrella package (install this) |
|
|
300
|
+
| sagellm-protocol | `isagellm-protocol` | `sagellm_protocol` | Protocol v0.1 types |
|
|
301
|
+
| sagellm-core | `isagellm-core` | `sagellm_core` | Runtime & config |
|
|
302
|
+
| sagellm-backend | `isagellm-backend` | `sagellm_backend` | Hardware abstraction |
|
|
303
|
+
|
|
304
|
+
## 🎯 Roadmap
|
|
305
|
+
|
|
306
|
+
- **Year 1**: Core inference with KV cache, prefix sharing, basic eviction
|
|
307
|
+
- **Year 2**: Multi-node inference, advanced scheduling
|
|
308
|
+
- **Year 3**: Full production-ready deployment
|
|
309
|
+
|
|
310
|
+
## 📄 License
|
|
311
|
+
|
|
312
|
+
Proprietary - IntelliStream. Internal use only.
|
|
313
|
+
|
|
314
|
+
---
|
|
315
|
+
|
|
316
|
+
<p align="center">
|
|
317
|
+
<sub>Built with ❤️ by IntelliStream Team for domestic AI infrastructure</sub>
|
|
318
|
+
</p>
|
|
@@ -0,0 +1,11 @@
|
|
|
1
|
+
sagellm/__init__.py,sha256=fvvYT8rqWc77Q6mHyEZp5o4BiTwa1fpmxHjzXypz5xo,5379
|
|
2
|
+
sagellm/__init__.pyc,sha256=GwhdZnZYMqAatjKR1T-2j1l7gDvAmRgGLg8QFseZcgw,4591
|
|
3
|
+
sagellm/cli.pyc,sha256=pUhwr6R1ChduZB-Z8x_-kgRXy0OtDFqnM_r3XSSlp_s,52742
|
|
4
|
+
sagellm/py.typed,sha256=ixa8YukDZ3kLo0WsFJRGohLMyHzbMur1ALmmASML2cs,64
|
|
5
|
+
sagellm/__pycache__/__init__.cpython-311.pyc,sha256=xGiRA2rVPAyoIUzq5phb-y3bV3VRpjYU3eIJbmGhID0,4616
|
|
6
|
+
sagellm/__pycache__/cli.cpython-311.pyc,sha256=n-HafVk9XrfnPpl56eonKAeIF2XCoAaUvdkJ0unLJBY,52767
|
|
7
|
+
isagellm-0.1.0.6.dist-info/METADATA,sha256=9VO23GmlyIdekx0baW9SS3H5nhXHMbhWQIdpkjSzgP8,10008
|
|
8
|
+
isagellm-0.1.0.6.dist-info/WHEEL,sha256=ZJeWpR6hcCRGwxVKXlDk-HsGwijNyTq4fszaDj4Ycyo,93
|
|
9
|
+
isagellm-0.1.0.6.dist-info/entry_points.txt,sha256=NqSiD9EEbziWs94BYtKFUzrnKTyCFG2MuZcvrRryhtg,73
|
|
10
|
+
isagellm-0.1.0.6.dist-info/top_level.txt,sha256=q-O8RUHV2YT7pQv12AYgFiK7PNvB9cHVg_7s5Tp08xI,8
|
|
11
|
+
isagellm-0.1.0.6.dist-info/RECORD,,
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
sagellm
|
sagellm/__init__.py
ADDED
|
@@ -0,0 +1,158 @@
|
|
|
1
|
+
"""sageLLM: Modular LLM inference engine for domestic computing power.
|
|
2
|
+
|
|
3
|
+
Ollama-like experience for Chinese hardware ecosystems (Huawei Ascend, NVIDIA).
|
|
4
|
+
|
|
5
|
+
Quick Start:
|
|
6
|
+
# Install
|
|
7
|
+
pip install isagellm
|
|
8
|
+
|
|
9
|
+
# CLI usage (like ollama)
|
|
10
|
+
sage-llm serve --mock # Start mock server
|
|
11
|
+
sage-llm run -p "Hello world" # Single inference
|
|
12
|
+
sage-llm demo --workload year1 # Run Year1 demo validation
|
|
13
|
+
sage-llm info # Show system info
|
|
14
|
+
|
|
15
|
+
# Python API
|
|
16
|
+
from sagellm import Request, MockEngine, create_engine
|
|
17
|
+
|
|
18
|
+
engine = MockEngine()
|
|
19
|
+
response = engine.generate(Request(prompt="Hello", max_tokens=128))
|
|
20
|
+
print(response.text)
|
|
21
|
+
|
|
22
|
+
Architecture:
|
|
23
|
+
sagellm (umbrella)
|
|
24
|
+
├── sagellm-protocol # Protocol v0.1 types (Request, Response, Metrics, Error)
|
|
25
|
+
├── sagellm-core # Runtime (config, engine factory, demo runner)
|
|
26
|
+
└── sagellm-backend # Hardware abstraction (CUDA, Ascend, Mock)
|
|
27
|
+
"""
|
|
28
|
+
|
|
29
|
+
from __future__ import annotations
|
|
30
|
+
|
|
31
|
+
__version__ = "0.1.0.6"
|
|
32
|
+
|
|
33
|
+
# Lazy imports to handle installation order
|
|
34
|
+
_LAZY_IMPORTS: dict[str, tuple[str, str]] = {
|
|
35
|
+
# Protocol types
|
|
36
|
+
"Request": ("sagellm_protocol", "Request"),
|
|
37
|
+
"Response": ("sagellm_protocol", "Response"),
|
|
38
|
+
"Metrics": ("sagellm_protocol", "Metrics"),
|
|
39
|
+
"Error": ("sagellm_protocol", "Error"),
|
|
40
|
+
"ErrorCode": ("sagellm_protocol", "ErrorCode"),
|
|
41
|
+
"Timestamps": ("sagellm_protocol", "Timestamps"),
|
|
42
|
+
"StreamEvent": ("sagellm_protocol", "StreamEvent"),
|
|
43
|
+
"StreamEventStart": ("sagellm_protocol", "StreamEventStart"),
|
|
44
|
+
"StreamEventDelta": ("sagellm_protocol", "StreamEventDelta"),
|
|
45
|
+
"StreamEventEnd": ("sagellm_protocol", "StreamEventEnd"),
|
|
46
|
+
# KV hooks
|
|
47
|
+
"KVAllocateParams": ("sagellm_protocol", "KVAllocateParams"),
|
|
48
|
+
"KVHandle": ("sagellm_protocol", "KVHandle"),
|
|
49
|
+
"KVMigrateParams": ("sagellm_protocol", "KVMigrateParams"),
|
|
50
|
+
# Backend
|
|
51
|
+
"BackendProvider": ("sagellm_backend", "BackendProvider"),
|
|
52
|
+
"CapabilityDescriptor": ("sagellm_backend", "CapabilityDescriptor"),
|
|
53
|
+
"DType": ("sagellm_backend", "DType"),
|
|
54
|
+
"KernelKind": ("sagellm_backend", "KernelKind"),
|
|
55
|
+
"MockBackendProvider": ("sagellm_backend", "MockBackendProvider"),
|
|
56
|
+
"create_mock_backend": ("sagellm_backend", "create_mock_backend"),
|
|
57
|
+
# Core - Config
|
|
58
|
+
"BackendConfig": ("sagellm_core", "BackendConfig"),
|
|
59
|
+
"EngineConfig": ("sagellm_core", "EngineConfig"),
|
|
60
|
+
"DemoConfig": ("sagellm_core", "DemoConfig"),
|
|
61
|
+
"WorkloadConfig": ("sagellm_core", "WorkloadConfig"),
|
|
62
|
+
"WorkloadSegment": ("sagellm_core", "WorkloadSegment"),
|
|
63
|
+
"load_config": ("sagellm_core", "load_config"),
|
|
64
|
+
# Core - Engine
|
|
65
|
+
"BaseEngine": ("sagellm_core", "BaseEngine"),
|
|
66
|
+
"MockEngine": ("sagellm_core", "MockEngine"),
|
|
67
|
+
"create_engine": ("sagellm_core", "create_engine"),
|
|
68
|
+
"create_backend": ("sagellm_core", "create_backend"),
|
|
69
|
+
# Core - Demo
|
|
70
|
+
"DemoRunner": ("sagellm_core", "DemoRunner"),
|
|
71
|
+
"demo_main": ("sagellm_core", "demo_main"),
|
|
72
|
+
# Control Plane (optional - install with isagellm[control-plane])
|
|
73
|
+
"ControlPlaneManager": ("sagellm_control", "ControlPlaneManager"),
|
|
74
|
+
"MockControlPlane": ("sagellm_control", "MockControlPlane"),
|
|
75
|
+
"EngineInfo": ("sagellm_control", "EngineInfo"),
|
|
76
|
+
"EngineState": ("sagellm_control", "EngineState"),
|
|
77
|
+
"SchedulingDecision": ("sagellm_control", "SchedulingDecision"),
|
|
78
|
+
}
|
|
79
|
+
|
|
80
|
+
|
|
81
|
+
def __getattr__(name: str) -> object:
|
|
82
|
+
"""Lazy import for all exported symbols."""
|
|
83
|
+
if name in _LAZY_IMPORTS:
|
|
84
|
+
module_name, attr_name = _LAZY_IMPORTS[name]
|
|
85
|
+
import importlib
|
|
86
|
+
|
|
87
|
+
try:
|
|
88
|
+
module = importlib.import_module(module_name)
|
|
89
|
+
return getattr(module, attr_name)
|
|
90
|
+
except ImportError as e:
|
|
91
|
+
# Provide helpful error for optional dependencies
|
|
92
|
+
if module_name == "sagellm_control":
|
|
93
|
+
raise ImportError(
|
|
94
|
+
f"{name} requires sagellm-control-plane. "
|
|
95
|
+
"Install with: pip install 'isagellm[control-plane]'"
|
|
96
|
+
) from e
|
|
97
|
+
if module_name == "sagellm_gateway":
|
|
98
|
+
raise ImportError(
|
|
99
|
+
f"{name} requires sagellm-gateway. "
|
|
100
|
+
"Install with: pip install 'isagellm[gateway]'"
|
|
101
|
+
) from e
|
|
102
|
+
raise
|
|
103
|
+
raise AttributeError(f"module 'sagellm' has no attribute {name!r}")
|
|
104
|
+
|
|
105
|
+
|
|
106
|
+
def __dir__() -> list[str]:
|
|
107
|
+
"""Return all public symbols."""
|
|
108
|
+
return list(__all__)
|
|
109
|
+
|
|
110
|
+
|
|
111
|
+
__all__ = [
|
|
112
|
+
# Version
|
|
113
|
+
"__version__",
|
|
114
|
+
# Protocol - Core types
|
|
115
|
+
"Request",
|
|
116
|
+
"Response",
|
|
117
|
+
"Metrics",
|
|
118
|
+
"Error",
|
|
119
|
+
"ErrorCode",
|
|
120
|
+
"Timestamps",
|
|
121
|
+
# Protocol - Streaming
|
|
122
|
+
"StreamEvent",
|
|
123
|
+
"StreamEventStart",
|
|
124
|
+
"StreamEventDelta",
|
|
125
|
+
"StreamEventEnd",
|
|
126
|
+
# Protocol - KV hooks
|
|
127
|
+
"KVAllocateParams",
|
|
128
|
+
"KVHandle",
|
|
129
|
+
"KVMigrateParams",
|
|
130
|
+
# Backend
|
|
131
|
+
"BackendProvider",
|
|
132
|
+
"CapabilityDescriptor",
|
|
133
|
+
"DType",
|
|
134
|
+
"KernelKind",
|
|
135
|
+
"MockBackendProvider",
|
|
136
|
+
"create_mock_backend",
|
|
137
|
+
# Core - Config
|
|
138
|
+
"BackendConfig",
|
|
139
|
+
"EngineConfig",
|
|
140
|
+
"DemoConfig",
|
|
141
|
+
"WorkloadConfig",
|
|
142
|
+
"WorkloadSegment",
|
|
143
|
+
"load_config",
|
|
144
|
+
# Core - Engine
|
|
145
|
+
"BaseEngine",
|
|
146
|
+
"MockEngine",
|
|
147
|
+
"create_engine",
|
|
148
|
+
"create_backend",
|
|
149
|
+
# Core - Demo
|
|
150
|
+
"DemoRunner",
|
|
151
|
+
"demo_main",
|
|
152
|
+
# Control Plane (optional)
|
|
153
|
+
"ControlPlaneManager",
|
|
154
|
+
"MockControlPlane",
|
|
155
|
+
"EngineInfo",
|
|
156
|
+
"EngineState",
|
|
157
|
+
"SchedulingDecision",
|
|
158
|
+
]
|
sagellm/__init__.pyc
ADDED
|
Binary file
|
|
Binary file
|
|
Binary file
|
sagellm/cli.pyc
ADDED
|
Binary file
|
sagellm/py.typed
ADDED