[Doc] change logo and rearange docs

README.md

@@ -1,7 +1,7 @@
 <p align="center">
   <picture>
     <source media="(prefers-color-scheme: dark)" srcset="docs/source/logos/UCM.png">
-    <img alt="UCM" src="docs/source/logos/UCM.png" width=70%>
+    <img alt="UCM" src="docs/source/logos/UCM-light.png" width=50%>
   </picture>
 </p>
 

docs/source/_static/css/logo.css

@@ -0,0 +1,4 @@
+.navbar-brand img {
+    max-width: 180px;
+    height: auto;
+}

docs/source/conf.py

@@ -39,14 +39,19 @@
 
 html_title = project
 html_theme = "sphinx_book_theme"
-html_logo = "logos/UCM.png"
+html_static_path = ["_static"]
+html_css_files = ["css/logo.css"]
 html_theme_options = {
     "path_to_docs": "docs/source",
     "repository_url": "https://github.com/ModelEngine-Group/unified-cache-management",
     "use_repository_button": True,
     "use_edit_page_button": True,
+    "logo": {
+        "image_light": "logos/UCM-light.png",
+        "image_dark": "logos/UCM-dark.png",
+        "alt_text": "UCM",
+    },
 }
-
 import os
 import shutil
 
@@ -64,6 +69,4 @@ def setup(app):
     app.connect("build-finished", copy_images)
 
 
-# html_static_path = ['_static']
-
 # language = 'zh_CN'

docs/source/developer/contributing.md → docs/source/developer_guide/contributing.md

@@ -1,4 +1,4 @@
-# Contributing
+# How to contribute
 ## Building and testing
 It’s recommended to set up a local development environment to build and test before you submit a PR.
 ### Run lint locally

docs/source/developer_guide/design/index.md

@@ -0,0 +1,12 @@
+# Design
+This section provides a detailed design guide of UCM features. Developers can refer to this section to see how UCM works.
+
+:::{toctree}
+:caption: Design Index
+:maxdepth: 1
+architecture.md
+add_connector.md
+nfs_connector.md
+vllm_institution.md
+:::
+

docs/source/developer/vllm_institution.md → docs/source/developer_guide/design/vllm_institution.md

@@ -3,7 +3,7 @@ This doc shows how the UcmKVStoreBase work with KV connector api in v1 to suppor
 ## How it works
 As you can see in the README part, the KVStoreBase helps decoupling sparse algorithms and external storage, a class that inherits from KVConnectorBase_V1 named UnifiedCacheConnectorV1 facilitates the connection between vLLM v1 and this class, the The figure below shows how it worked:
 
-![uc_connector](../images/ucconn_ucmconn.png)(../images/ucconn_ucmconn.png)
+![uc_connector](../../images/ucconn_ucmconn.png)(../../images/ucconn_ucmconn.png)
 
 The interfaces designed in KVStoreBase are similar to the KV connector API in v1, which are divided into scheduler-side methods and worker-side methods, as follows:
 - scheduler methods

docs/source/developer_guide/performance/index.md

@@ -0,0 +1,8 @@
+# Performance
+This section provides methods to test UCM performance.
+
+:::{toctree}
+:caption: Performance
+:maxdepth: 1
+performance_benchmark
+:::

docs/source/getting-started/installation/index.md

@@ -0,0 +1,10 @@
+# Installation
+UCM supports the following hardware platforms:
+
+:::{toctree}
+:maxdepth: 1
+:caption: Index
+installation.md
+installation_npu.md
+:::
+

docs/source/getting-started/installation.md → docs/source/getting-started/installation/installation.md

@@ -1,4 +1,4 @@
-# Installation
+# GPU
 This document describes how to install unified-cache-management.
 
 ## Requirements

docs/source/getting-started/installation_npu.md → docs/source/getting-started/installation/installation_npu.md

@@ -1,4 +1,4 @@
-# Installation NPU
+# NPU
 This document describes how to install unified-cache-management when using Ascend NPU manually.
 
 ## Requirements

docs/source/getting-started/quick_start.md

@@ -0,0 +1 @@
+# Quickstart

docs/source/index.md

@@ -1,10 +1,10 @@
 # Welcome to Unified Cache Manager
 
-:::{figure} ./logos/UCM.png
+:::{figure} ./logos/UCM-light.png
 :align: center
 :alt: UCM
 :class: no-scaled-link
-:width: 70%
+:width: 50%
 :::
 
 :::{raw} html
@@ -23,11 +23,34 @@
 
 Make KVCache Great Again!
 
+## Documentation
+
+:::{toctree}
+:caption: Getting Started
+:maxdepth: 1
+getting-started/quick_start
+getting-started/installation/index
+:::
+
 :::{toctree}
-:maxdepth: 3
-getting-started/index.md
-feature/index.md
-developer/index.md
-about.md
+:caption: User Guide
+:maxdepth: 1
+user_guide/support_matrix/index
+user_guide/features/index
+user_guide/connector_guide/index
+user_guide/engine_guide/index
 :::
 
+:::{toctree}
+:caption: Developer Guide
+:maxdepth: 1
+developer_guide/design/index
+developer_guide/contributing
+developer_guide/performance/index
+:::
+
+:::{toctree}
+:caption: About Us
+:maxdepth: 1
+about
+:::

docs/source/user_guide/connector_guide/index.md

@@ -0,0 +1,2 @@
+# Connector Guide
+This section provides a guide to connectors currently supported by UCM.

docs/source/user_guide/engine_guide/index.md

@@ -0,0 +1,2 @@
+# Engine Guide
+This section provides a guide to the serving engines currently supported by UCM. Users can refer to this guide to use serving engines with UCM.

docs/source/getting-started/example/disaggregated_prefill/1p1d.md → docs/source/user_guide/examples/disaggregated_prefill/1p1d.md

@@ -69,7 +69,7 @@ CUDA_VISIBLE_DEVICES=1 vllm serve /home/models/Qwen2.5-7B-Instruct \
 Make sure prefill nodes and decode nodes can connect to each other.
 ```bash
 cd vllm-workspace/unified-cache-management/test/
-python3 toy_proxy_server.py --host localhost --port 7802 --prefiller-host <prefill-node-ip> --prefiller-port 7800 --decoder-host <prefill-node-ip> --decoder-port 7801
+python3 toy_proxy_server.py --host localhost --port 7802 --prefiller-host <prefill-node-ip> --prefiller-port 7800 --decoder-host <decode-node-ip> --decoder-port 7801
 ```
 
 ## Testing and Benchmarking
@@ -80,7 +80,7 @@ curl http://localhost:7802/v1/completions \
     -H "Content-Type: application/json" \
     -d '{
         "model": "/home/models/Qwen2.5-7B-Instruct",
-        "prompt": "content": "What date is today?",
+        "prompt": "What date is today?",
         "max_tokens": 20,
         "temperature": 0
     }'

docs/source/getting-started/example/disaggregated_prefill/xpyd.md → docs/source/user_guide/examples/disaggregated_prefill/xpyd.md

@@ -133,7 +133,7 @@ curl http://localhost:7805/v1/completions \
     -H "Content-Type: application/json" \
     -d '{
         "model": "/home/models/Qwen2.5-7B-Instruct",
-        "prompt": "content": "What date is today?",
+        "prompt": "What date is today?",
         "max_tokens": 20,
         "temperature": 0
     }'

docs/source/user_guide/features/index.md

@@ -0,0 +1,13 @@
+# Feature Guide
+This section provides a detailed usage guide of UCM features.
+
+:::{toctree}
+:maxdepth: 1
+:caption: Feature Index
+prefix_cache.md
+pd.md
+rag_cache.md
+sparse/index.md
+store/index.md
+:::
+

docs/source/feature/sparse_attn.md → docs/source/user_guide/features/sparse/base.md

@@ -3,12 +3,12 @@
 Attention mechanisms, especially in LLMs, are often the bottleneck in terms of latency during inference due to their computational complexity. Despite their importance in capturing contextual relationships, traditional attention requires processing all token interactions, leading to significant delays.
 
 <p align="center">
-  <img alt="UCM" src="../images/attention_overhead.png" width="80%">
+  <img alt="UCM" src="../../../images/attention_overhead.png" width="80%">
 </p>
 
 Researchers have found that attention in LLM is highly dispersed:
 <p align="center">
-  <img alt="UCM" src="../images/attention_sparsity.png" width="80%">
+  <img alt="UCM" src="../../../images/attention_sparsity.png" width="80%">
 </p>
 
 This movitates them actively developing sparse attention algorithms to address the latency issue. These algorithms aim to reduce the number of token interactions by focusing only on the most relevant parts of the input, thereby lowering the computation and memory requirements.
@@ -24,7 +24,7 @@ By utilizing UCM, researchers can efficiently implement rapid prototyping and te
 ### Overview
 The core concept of our UCMSparse attention framework is to offload the complete Key-Value (KV) cache to a dedicated KV cache storage. We then identify the crucial KV pairs relevant to the current context, as determined by our sparse attention algorithms, and selectively load only the necessary portions of the KV cache from storage into High Bandwidth Memory (HBM). This design significantly reduces the HBM footprint while accelerating generation speed.
 <p align="center">
-  <img alt="UCM" src="../images/sparse_attn_arch.png" width="80%">
+  <img alt="UCM" src="../../../images/sparse_attn_arch.png" width="80%">
 </p>
 
 

docs/source/user_guide/features/sparse/cacheblend.md

@@ -0,0 +1 @@
+# Cache Blend

docs/source/user_guide/features/sparse/esa.md

@@ -0,0 +1 @@
+# Esa

docs/source/user_guide/features/sparse/gsa.md

@@ -0,0 +1 @@
+# Gsa

docs/source/user_guide/features/sparse/index.md

@@ -0,0 +1,13 @@
+# Sparse
+
+:::{toctree}
+:maxdepth: 1
+:caption: Index
+Base <base>
+esa
+gsa
+kvcomp
+kvstar
+prefill_offload
+cacheblend
+:::

docs/source/user_guide/features/sparse/kvcomp.md

@@ -0,0 +1 @@
+# KV Comp

docs/source/user_guide/features/sparse/kvstar.md

@@ -0,0 +1 @@
+# KVStar

docs/source/user_guide/features/sparse/prefill_offload.md

@@ -0,0 +1 @@
+# Prefill Offload

docs/source/user_guide/features/store/3fs_store.md

@@ -0,0 +1 @@
+# 3fs Store

docs/source/user_guide/features/store/base.md

@@ -0,0 +1 @@
+# Base

docs/source/user_guide/features/store/dram_store.md

@@ -0,0 +1 @@
+# Dram Store

docs/source/user_guide/features/store/index.md

@@ -0,0 +1,13 @@
+# Store
+
+:::{toctree}
+:maxdepth: 1
+:caption: Index
+base
+3fs_store
+dram_store
+vfs_store
+nfs_store
+nds_store
+mooncake_store
+:::

docs/source/user_guide/features/store/mooncake_store.md

@@ -0,0 +1 @@
+# Mooncake Store

docs/source/user_guide/features/store/nds_store.md

@@ -0,0 +1 @@
+# Nds Store

docs/source/user_guide/features/store/nfs_store.md

@@ -0,0 +1 @@
+# Nfs Store

docs/source/user_guide/features/store/vfs_store.md

@@ -0,0 +1 @@
+# Vfs Store

docs/source/user_guide/support_matrix/index.md

@@ -0,0 +1,8 @@
+# Support Matrix
+This section provides support matrices of UCM.
+
+:::{toctree}
+:maxdepth: 1
+:caption: Support Index
+support
+:::