v0.1.3-rc.4正式版

2025-10-06 07:06:51 +08:00 · 2025-01-27 14:03:46 +08:00
parent 76d5b55b5a
commit c58f2697f0
41 changed files with 1956 additions and 964 deletions
--- a/.env.example
+++ b/.env.example
@@ -12,11 +12,12 @@ AUTH_TOKEN=
 # 共享的认证令牌，仅Chat端点权限(轮询与AUTH_TOKEN同步)，无其余权限
 SHARED_TOKEN=

-# 启用流式响应检查，关闭则无法响应错误，代价是会对第一个块解析2次
-ENABLE_STREAM_CHECK=true
+# 启用流式响应检查，关闭则无法响应错误，代价是会对第一个块解析2次(已弃用)
+# 新版本已经完成优化
+# ENABLE_STREAM_CHECK=true

-# 流式消息结束后发送包含"finish_reason"为"stop"的空消息块
-INCLUDE_STOP_REASON_STREAM=true
+# 流式消息结束后发送包含"finish_reason"为"stop"的空消息块(已弃用)
+# INCLUDE_STOP_REASON_STREAM=true

 # 令牌文件路径(已弃用)
 # TOKEN_FILE=.token
@@ -88,3 +89,12 @@ REQUEST_LOGS_LIMIT=100

 # Cursor 服务超时(秒)(最大值600)
 SERVICE_TIMEOUT=30
+
+# 包含网络引用
+INCLUDE_WEB_REFERENCES=false
+
+# 持久化日志文件路径
+LOGS_FILE_PATH=logs.bin
+
+# 持久化页面配置文件路径
+PAGES_FILE_PATH=pages.bin
--- a/.gitignore
+++ b/.gitignore
@@ -21,3 +21,5 @@ node_modules
 /logs
 /dev*
 /build*
+/*.bin
+/result.txt
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -17,6 +17,17 @@ version = "2.0.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "512761e0bb2578dd7380c6baaa0f4ce03e84f95e960231d1dec8bf4d7d6e2627"

+[[package]]
+name = "ahash"
+version = "0.7.8"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "891477e0c6a8957309ee5c45a6368af3ae14bb510732d2684ffa19af310920f9"
+dependencies = [
+ "getrandom",
+ "once_cell",
+ "version_check",
+]
+
 [[package]]
 name = "aho-corasick"
 version = "1.1.3"
@@ -175,6 +186,18 @@ version = "2.8.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "8f68f53c83ab957f72c32642f3868eec03eb974d1fb82e453128456482613d36"

+[[package]]
+name = "bitvec"
+version = "1.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1bc2832c24239b0141d5674bb9174f9d68a8b5b3f2753311927c172ca46f7e9c"
+dependencies = [
+ "funty",
+ "radium",
+ "tap",
+ "wyz",
+]
+
 [[package]]
 name = "block-buffer"
 version = "0.10.4"
@@ -211,6 +234,28 @@ version = "3.16.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "79296716171880943b8470b5f8d03aa55eb2e645a4874bdbb28adb49162e012c"

+[[package]]
+name = "bytecheck"
+version = "0.6.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "23cdc57ce23ac53c931e88a43d06d070a6fd142f2617be5855eb75efc9beb1c2"
+dependencies = [
+ "bytecheck_derive",
+ "ptr_meta 0.1.4",
+ "simdutf8",
+]
+
+[[package]]
+name = "bytecheck_derive"
+version = "0.6.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3db406d29fbcd95542e92559bed4d8ad92636d1ca8b3b72ede10b4bcc010e659"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 1.0.109",
+]
+
 [[package]]
 name = "bytemuck"
 version = "1.21.0"
@@ -259,6 +304,7 @@ dependencies = [
 "android-tzdata",
 "iana-time-zone",
 "num-traits",
+ "rkyv 0.7.45",
 "serde",
 "windows-targets",
 ]
@@ -315,7 +361,7 @@ dependencies = [

 [[package]]
 name = "cursor-api"
-version = "0.1.3-rc.4-pre.1"
+version = "0.1.3-rc.4"
 dependencies = [
 "axum",
 "base64",
@@ -327,6 +373,7 @@ dependencies = [
 "gif",
 "hex",
 "image",
+ "memmap2",
 "parking_lot",
 "paste",
 "prost",
@@ -334,13 +381,16 @@ dependencies = [
 "rand",
 "regex",
 "reqwest",
+ "rkyv 0.7.45",
 "serde",
 "serde_json",
 "sha2",
+ "sonic-rs",
 "sysinfo",
 "tokio",
 "tokio-stream",
 "tower-http",
+ "url",
 "uuid",
 ]

@@ -362,7 +412,7 @@ checksum = "97369cbbc041bc366949bc74d34658d6cda5621039731c6310521892a3a20ae0"
 dependencies = [
 "proc-macro2",
 "quote",
- "syn",
+ "syn 2.0.96",
 ]

 [[package]]
@@ -408,6 +458,18 @@ version = "2.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "37909eebbb50d72f9059c3b6d82c0463f2ff062c9e95845c43a6c9c0355411be"

+[[package]]
+name = "faststr"
+version = "0.2.27"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9154486833a83cb5d99de8c4d831314b8ae810dd4ef18d89ceb7a9c7c728dd74"
+dependencies = [
+ "bytes",
+ "rkyv 0.8.10",
+ "serde",
+ "simdutf8",
+]
+
 [[package]]
 name = "fdeflate"
 version = "0.3.7"
@@ -463,6 +525,12 @@ dependencies = [
 "percent-encoding",
 ]

+[[package]]
+name = "funty"
+version = "2.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e6d5a32815ae3f33302d95fdcb2ce17862f8c65363dcfd29360480ba1001fc9c"
+
 [[package]]
 name = "futures"
 version = "0.3.31"
@@ -507,7 +575,7 @@ checksum = "162ee34ebcb7c64a8abebc059ce0fee27c2262618d7b60ed8faf72fef13c3650"
 dependencies = [
 "proc-macro2",
 "quote",
- "syn",
+ "syn 2.0.96",
 ]

 [[package]]
@@ -596,6 +664,15 @@ dependencies = [
 "tracing",
 ]

+[[package]]
+name = "hashbrown"
+version = "0.12.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8a9ee70c43aaf417c914396645a0fa852624801b24ebb7ae78fe8272889ac888"
+dependencies = [
+ "ahash",
+]
+
 [[package]]
 name = "hashbrown"
 version = "0.15.2"
@@ -871,7 +948,7 @@ checksum = "1ec89e9337638ecdc08744df490b221a7399bf8d164eb52a665454e60e075ad6"
 dependencies = [
 "proc-macro2",
 "quote",
- "syn",
+ "syn 2.0.96",
 ]

 [[package]]
@@ -929,7 +1006,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "8c9c992b02b5b4c94ea26e32fe5bccb7aa7d9f390ab5c1221ff895bc7ea8b652"
 dependencies = [
 "equivalent",
- "hashbrown",
+ "hashbrown 0.15.2",
 ]

 [[package]]
@@ -1009,6 +1086,15 @@ version = "2.7.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "78ca9ab1a0babb1e7d5695e3530886289c18cf2f87ec19a575a0abdce112e3a3"

+[[package]]
+name = "memmap2"
+version = "0.9.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fd3f7eed9d3848f8b98834af67102b720745c4ec028fcd0aa0239277e7de374f"
+dependencies = [
+ "libc",
+]
+
 [[package]]
 name = "mime"
 version = "0.3.17"
@@ -1042,6 +1128,26 @@ version = "0.10.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "defc4c55412d89136f966bbb339008b474350e5e6e78d2714439c386b3137a03"

+[[package]]
+name = "munge"
+version = "0.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "64142d38c84badf60abf06ff9bd80ad2174306a5b11bd4706535090a30a419df"
+dependencies = [
+ "munge_macro",
+]
+
+[[package]]
+name = "munge_macro"
+version = "0.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1bb5c1d8184f13f7d0ccbeeca0def2f9a181bce2624302793005f5ca8aa62e5e"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.96",
+]
+
 [[package]]
 name = "native-tls"
 version = "0.2.12"
@@ -1115,7 +1221,7 @@ checksum = "a948666b637a0f465e8564c73e89d4dde00d72d4d473cc972f390fc3dcee7d9c"
 dependencies = [
 "proc-macro2",
 "quote",
- "syn",
+ "syn 2.0.96",
 ]

 [[package]]
@@ -1228,7 +1334,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "6924ced06e1f7dfe3fa48d57b9f74f55d8915f5036121bef647ef4b204895fac"
 dependencies = [
 "proc-macro2",
- "syn",
+ "syn 2.0.96",
 ]

 [[package]]
@@ -1266,7 +1372,7 @@ dependencies = [
 "prost",
 "prost-types",
 "regex",
- "syn",
+ "syn 2.0.96",
 "tempfile",
 ]

@@ -1280,7 +1386,7 @@ dependencies = [
 "itertools",
 "proc-macro2",
 "quote",
- "syn",
+ "syn 2.0.96",
 ]

 [[package]]
@@ -1292,6 +1398,46 @@ dependencies = [
 "prost",
 ]

+[[package]]
+name = "ptr_meta"
+version = "0.1.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0738ccf7ea06b608c10564b31debd4f5bc5e197fc8bfe088f68ae5ce81e7a4f1"
+dependencies = [
+ "ptr_meta_derive 0.1.4",
+]
+
+[[package]]
+name = "ptr_meta"
+version = "0.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fe9e76f66d3f9606f44e45598d155cb13ecf09f4a28199e48daf8c8fc937ea90"
+dependencies = [
+ "ptr_meta_derive 0.3.0",
+]
+
+[[package]]
+name = "ptr_meta_derive"
+version = "0.1.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "16b845dbfca988fa33db069c0e230574d15a3088f147a87b64c7589eb662c9ac"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 1.0.109",
+]
+
+[[package]]
+name = "ptr_meta_derive"
+version = "0.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ca414edb151b4c8d125c12566ab0d74dc9cdba36fb80eb7b848c15f495fd32d1"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.96",
+]
+
 [[package]]
 name = "quick-error"
 version = "2.0.1"
@@ -1307,6 +1453,21 @@ dependencies = [
 "proc-macro2",
 ]

+[[package]]
+name = "radium"
+version = "0.7.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dc33ff2d4973d518d823d61aa239014831e521c75da58e3df4840d3f47749d09"
+
+[[package]]
+name = "rancor"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "caf5f7161924b9d1cea0e4cabc97c372cea92b5f927fc13c6bca67157a0ad947"
+dependencies = [
+ "ptr_meta 0.3.0",
+]
+
 [[package]]
 name = "rand"
 version = "0.8.5"
@@ -1346,6 +1507,26 @@ dependencies = [
 "bitflags 2.8.0",
 ]

+[[package]]
+name = "ref-cast"
+version = "1.0.23"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ccf0a6f84d5f1d581da8b41b47ec8600871962f2a528115b542b362d4b744931"
+dependencies = [
+ "ref-cast-impl",
+]
+
+[[package]]
+name = "ref-cast-impl"
+version = "1.0.23"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bcc303e793d3734489387d205e9b186fac9c6cfacedd98cbb2e8a5943595f3e6"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.96",
+]
+
 [[package]]
 name = "regex"
 version = "1.11.1"
@@ -1375,6 +1556,21 @@ version = "0.8.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "2b15c43186be67a4fd63bee50d0303afffcef381492ebe2c5d87f324e1b8815c"

+[[package]]
+name = "rend"
+version = "0.4.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "71fe3824f5629716b1589be05dacd749f6aa084c87e00e016714a8cdfccc997c"
+dependencies = [
+ "bytecheck",
+]
+
+[[package]]
+name = "rend"
+version = "0.5.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a35e8a6bf28cd121053a66aa2e6a2e3eaffad4a60012179f0e864aa5ffeff215"
+
 [[package]]
 name = "reqwest"
 version = "0.12.12"
@@ -1438,6 +1634,64 @@ dependencies = [
 "windows-sys 0.52.0",
 ]

+[[package]]
+name = "rkyv"
+version = "0.7.45"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9008cd6385b9e161d8229e1f6549dd23c3d022f132a2ea37ac3a10ac4935779b"
+dependencies = [
+ "bitvec",
+ "bytecheck",
+ "bytes",
+ "hashbrown 0.12.3",
+ "ptr_meta 0.1.4",
+ "rend 0.4.2",
+ "rkyv_derive 0.7.45",
+ "seahash",
+ "tinyvec",
+ "uuid",
+]
+
+[[package]]
+name = "rkyv"
+version = "0.8.10"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1e147371c75553e1e2fcdb483944a8540b8438c31426279553b9a8182a9b7b65"
+dependencies = [
+ "bytes",
+ "hashbrown 0.15.2",
+ "indexmap",
+ "munge",
+ "ptr_meta 0.3.0",
+ "rancor",
+ "rend 0.5.2",
+ "rkyv_derive 0.8.10",
+ "tinyvec",
+ "uuid",
+]
+
+[[package]]
+name = "rkyv_derive"
+version = "0.7.45"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "503d1d27590a2b0a3a4ca4c94755aa2875657196ecbf401a42eff41d7de532c0"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 1.0.109",
+]
+
+[[package]]
+name = "rkyv_derive"
+version = "0.8.10"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "246b40ac189af6c675d124b802e8ef6d5246c53e17367ce9501f8f66a81abb7a"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.96",
+]
+
 [[package]]
 name = "rustc-demangle"
 version = "0.1.24"
@@ -1523,6 +1777,12 @@ version = "1.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "94143f37725109f92c262ed2cf5e59bce7498c01bcc1502d7b9afe439a4e9f49"

+[[package]]
+name = "seahash"
+version = "4.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1c107b6f4780854c8b126e228ea8869f4d7b71260f962fefb57b996b8959ba6b"
+
 [[package]]
 name = "security-framework"
 version = "2.11.1"
@@ -1563,7 +1823,7 @@ checksum = "5a9bf7cf98d04a2b28aead066b7496853d4779c9cc183c440dbac457641e19a0"
 dependencies = [
 "proc-macro2",
 "quote",
- "syn",
+ "syn 2.0.96",
 ]

 [[package]]
@@ -1617,12 +1877,27 @@ version = "1.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "0fda2ff0d084019ba4d7c6f371c95d8fd75ce3524c3cb8fb653a3023f6323e64"

+[[package]]
+name = "signal-hook-registry"
+version = "1.4.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a9e9e0b4211b72e7b8b6e85c807d36c212bdb33ea8587f7569562a84df5465b1"
+dependencies = [
+ "libc",
+]
+
 [[package]]
 name = "simd-adler32"
 version = "0.3.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d66dc143e6b11c1eddc06d5c423cfc97062865baf299914ab64caa38182078fe"

+[[package]]
+name = "simdutf8"
+version = "0.1.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e3a9fe34e3e7a50316060351f37187a3f546bce95496156754b601a5fa71b76e"
+
 [[package]]
 name = "slab"
 version = "0.4.9"
@@ -1648,6 +1923,44 @@ dependencies = [
 "windows-sys 0.52.0",
 ]

+[[package]]
+name = "sonic-number"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a8a74044c092f4f43ca7a6cfd62854cf9fb5ac8502b131347c990bf22bef1dfe"
+dependencies = [
+ "cfg-if",
+]
+
+[[package]]
+name = "sonic-rs"
+version = "0.3.17"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0275f9f2f07d47556fe60c2759da8bc4be6083b047b491b2d476aa0bfa558eb1"
+dependencies = [
+ "bumpalo",
+ "bytes",
+ "cfg-if",
+ "faststr",
+ "itoa",
+ "ref-cast",
+ "ryu",
+ "serde",
+ "simdutf8",
+ "sonic-number",
+ "sonic-simd",
+ "thiserror 2.0.11",
+]
+
+[[package]]
+name = "sonic-simd"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "940a24e82c9a97483ef66cef06b92160a8fa5cd74042c57c10b24d99d169d2fc"
+dependencies = [
+ "cfg-if",
+]
+
 [[package]]
 name = "spin"
 version = "0.9.8"
@@ -1666,6 +1979,17 @@ version = "2.6.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "13c2bddecc57b384dee18652358fb23172facb8a2c51ccc10d74c157bdea3292"

+[[package]]
+name = "syn"
+version = "1.0.109"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "72b64191b275b66ffe2469e8af2c1cfe3bafa67b529ead792a6d0160888b4237"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "unicode-ident",
+]
+
 [[package]]
 name = "syn"
 version = "2.0.96"
@@ -1694,7 +2018,7 @@ checksum = "c8af7666ab7b6390ab78131fb5b0fce11d6b7a6951602017c35fa82800708971"
 dependencies = [
 "proc-macro2",
 "quote",
- "syn",
+ "syn 2.0.96",
 ]

 [[package]]
@@ -1731,6 +2055,12 @@ dependencies = [
 "libc",
 ]

+[[package]]
+name = "tap"
+version = "1.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "55937e1799185b12863d447f42597ed69d9928686b8d88a1df17376a097d8369"
+
 [[package]]
 name = "tempfile"
 version = "3.15.0"
@@ -1751,7 +2081,16 @@ version = "1.0.69"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b6aaf5339b578ea85b50e080feb250a3e8ae8cfcdff9a461c9ec2904bc923f52"
 dependencies = [
- "thiserror-impl",
+ "thiserror-impl 1.0.69",
+]
+
+[[package]]
+name = "thiserror"
+version = "2.0.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d452f284b73e6d76dd36758a0c8684b1d5be31f92b89d07fd5822175732206fc"
+dependencies = [
+ "thiserror-impl 2.0.11",
 ]

 [[package]]
@@ -1762,7 +2101,18 @@ checksum = "4fee6c4efc90059e10f81e6d42c60a18f76588c3d74cb83a0b242a2b6c7504c1"
 dependencies = [
 "proc-macro2",
 "quote",
- "syn",
+ "syn 2.0.96",
+]
+
+[[package]]
+name = "thiserror-impl"
+version = "2.0.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "26afc1baea8a989337eeb52b6e72a039780ce45c3edfcc9c5b9d112feeb173c2"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.96",
 ]

 [[package]]
@@ -1775,6 +2125,21 @@ dependencies = [
 "zerovec",
 ]

+[[package]]
+name = "tinyvec"
+version = "1.8.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "022db8904dfa342efe721985167e9fcd16c29b226db4397ed752a761cfce81e8"
+dependencies = [
+ "tinyvec_macros",
+]
+
+[[package]]
+name = "tinyvec_macros"
+version = "0.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1f3ccbac311fea05f86f61904b462b55fb3df8837a366dfc601a0161d0532f20"
+
 [[package]]
 name = "tokio"
 version = "1.43.0"
@@ -1786,6 +2151,7 @@ dependencies = [
 "libc",
 "mio",
 "pin-project-lite",
+ "signal-hook-registry",
 "socket2",
 "tokio-macros",
 "windows-sys 0.52.0",
@@ -1799,7 +2165,7 @@ checksum = "6e06d43f1345a3bcd39f6a56dbb7dcab2ba47e68e8ac134855e7e2bdbaf8cab8"
 dependencies = [
 "proc-macro2",
 "quote",
- "syn",
+ "syn 2.0.96",
 ]

 [[package]]
@@ -1830,7 +2196,7 @@ checksum = "0d4770b8024672c1101b3f6733eab95b18007dbe0847a8afe341fcf79e06043f"
 dependencies = [
 "either",
 "futures-util",
- "thiserror",
+ "thiserror 1.0.69",
 "tokio",
 ]

@@ -2027,7 +2393,7 @@ dependencies = [
 "log",
 "proc-macro2",
 "quote",
- "syn",
+ "syn 2.0.96",
 "wasm-bindgen-shared",
 ]

@@ -2062,7 +2428,7 @@ checksum = "8ae87ea40c9f689fc23f209965b6fb8a99ad69aeeb0231408be24920604395de"
 dependencies = [
 "proc-macro2",
 "quote",
- "syn",
+ "syn 2.0.96",
 "wasm-bindgen-backend",
 "wasm-bindgen-shared",
 ]
@@ -2166,7 +2532,7 @@ checksum = "9107ddc059d5b6fbfbffdfa7a7fe3e22a226def0b2608f72e9d552763d3e1ad7"
 dependencies = [
 "proc-macro2",
 "quote",
- "syn",
+ "syn 2.0.96",
 ]

 [[package]]
@@ -2177,7 +2543,7 @@ checksum = "29bee4b38ea3cde66011baa44dba677c432a78593e202392d1e9070cf2a7fca7"
 dependencies = [
 "proc-macro2",
 "quote",
- "syn",
+ "syn 2.0.96",
 ]

 [[package]]
@@ -2313,6 +2679,15 @@ version = "0.5.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "1e9df38ee2d2c3c5948ea468a8406ff0db0b29ae1ffde1bcf20ef305bcc95c51"

+[[package]]
+name = "wyz"
+version = "0.5.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "05f360fc0b24296329c78fda852a1e9ae82de9cf7b27dae4b7f62f118f77b9ed"
+dependencies = [
+ "tap",
+]
+
 [[package]]
 name = "yoke"
 version = "0.7.5"
@@ -2333,7 +2708,7 @@ checksum = "2380878cad4ac9aac1e2435f3eb4020e8374b5f13c296cb75b4620ff8e229154"
 dependencies = [
 "proc-macro2",
 "quote",
- "syn",
+ "syn 2.0.96",
 "synstructure",
 ]

@@ -2355,7 +2730,7 @@ checksum = "fa4f8080344d4671fb4e831a13ad1e68092748387dfc4f55e356242fae12ce3e"
 dependencies = [
 "proc-macro2",
 "quote",
- "syn",
+ "syn 2.0.96",
 ]

 [[package]]
@@ -2375,7 +2750,7 @@ checksum = "595eed982f7d355beb85837f651fa22e90b3c044842dc7f2c2842c086f295808"
 dependencies = [
 "proc-macro2",
 "quote",
- "syn",
+ "syn 2.0.96",
 "synstructure",
 ]

@@ -2404,7 +2779,7 @@ checksum = "6eafa6dfb17584ea3e2bd6e76e0cc15ad7af12b09abdd1ca55961bed9b1063c6"
 dependencies = [
 "proc-macro2",
 "quote",
- "syn",
+ "syn 2.0.96",
 ]

 [[package]]
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -1,6 +1,6 @@
 [package]
 name = "cursor-api"
-version = "0.1.3-rc.4-pre.1"
+version = "0.1.3-rc.4"
 edition = "2021"
 authors = ["wisdgod <nav@wisdgod.com>"]
 description = "OpenAI format compatibility layer for the Cursor API"
@@ -16,26 +16,30 @@ axum = { version = "0.8.1", features = ["json"] }
 base64 = { version = "0.22.1", default-features = false, features = ["std"] }
 # brotli = { version = "7.0.0", default-features = false, features = ["std"] }
 bytes = "1.9.0"
-chrono = { version = "0.4.39", default-features = false, features = ["std", "clock", "now", "serde"] }
+chrono = { version = "0.4.39", default-features = false, features = ["std", "clock", "now", "serde", "rkyv-64"] }
 dotenvy = "0.15.7"
 flate2 = { version = "1.0.35", default-features = false, features = ["rust_backend"] }
 futures = { version = "0.3.31", default-features = false, features = ["std"] }
 gif = { version = "0.13.1", default-features = false, features = ["std"] }
 hex = { version = "0.4.3", default-features = false, features = ["std"] }
 image = { version = "0.25.5", default-features = false, features = ["jpeg", "png", "gif", "webp"] }
+memmap2 = "0.9.5"
+# openssl = { version = "0.10.68", features = ["vendored"] }
 parking_lot = "0.12.3"
 paste = "1.0.15"
 prost = "0.13.4"
 rand = { version = "0.8.5", default-features = false, features = ["std", "std_rng"] }
 regex = { version = "1.11.1", default-features = false, features = ["std", "perf"] }
 reqwest = { version = "0.12.12", default-features = false, features = ["gzip", "brotli", "json", "stream", "socks", "__tls", "charset", "default-tls", "h2", "http2", "macos-system-configuration"] }
+rkyv = { version = "0.7.45", default-features = false, features = ["alloc", "std", "bytecheck", "size_64", "validation", "std"] }
 serde = { version = "1.0.217", default-features = false, features = ["std", "derive"] }
-serde_json = "1.0.137"
+serde_json = { package = "sonic-rs", version = "0.3.17" }
 sha2 = { version = "0.10.8", default-features = false }
 sysinfo = { version = "0.33.1", default-features = false, features = ["system"] }
-tokio = { version = "1.43.0", features = ["rt-multi-thread", "macros", "net", "sync", "time", "fs"] }
+tokio = { version = "1.43.0", features = ["rt-multi-thread", "macros", "net", "sync", "time", "fs", "signal"] }
 tokio-stream = { version = "0.1.17", features = ["time"] }
 tower-http = { version = "0.6.2", features = ["cors", "limit"] }
+url = { version = "2.5.4", default-features = false }
 uuid = { version = "1.12.1", features = ["v4"] }

 [profile.release]
--- a/Cross.toml
+++ b/Cross.toml
@@ -1,9 +1,5 @@
 [target.x86_64-unknown-linux-gnu]
 dockerfile = "Dockerfile.cross"

-[target.x86_64-unknown-freebsd]
-pre-build = [
-    "pkg update",
-    "pkg install -y node20 www/npm protobuf ca_root_nss bash gmake pkgconf openssl",
-    "export SSL_CERT_FILE=/etc/ssl/cert.pem"
-]
+[target.aarch64-unknown-linux-gnu]
+dockerfile = "Dockerfile.cross.arm64"
--- a/API.md
+++ b/API.md
@@ -0,0 +1,179 @@
+# Cursor API
+
+## 项目说明
+
+### 版本声明
+- 当前版本已进入稳定阶段
+- 以下问题与程序无关，请勿反馈：
+  - 响应缺字漏字
+  - 首字延迟现象
+  - 响应出现乱码
+- 性能优势：
+  - 达到原生客户端响应速度
+  - 部分场景下表现更优
+- 开源协议要求：
+  - Fork 项目禁止以原作者名义进行宣传推广
+  - 禁止发布任何形式的官方声明
+
+![Cursor API 架构示意图](https://via.placeholder.com/800x400.png?text=Cursor+API+Architecture)
+
+## 快速入门
+
+### 密钥获取
+1. 访问 [Cursor 官网](https://www.cursor.com) 完成注册登录
+2. 开启浏览器开发者工具 (F12)
+3. 在 Application → Cookies 中定位 `WorkosCursorSessionToken`
+4. 复制第三个字段值（注意：`%3A%3A` 为 `::` 的 URL 编码形式）
+
+## 配置指南
+
+### 环境变量
+| 变量名 | 类型 | 默认值 | 说明 |
+|--------|------|--------|-----|
+| PORT   | int  | 3000   | 服务端口号 |
+| AUTH_TOKEN | string | 无 | 认证令牌（必需） |
+| ROUTE_PREFIX | string | 无 | 路由前缀 |
+| TOKEN_LIST_FILE | string | .tokens | Token 存储文件 |
+
+完整配置参见 [env-example](/env-example)
+
+### Token 文件规范
+`.tokens` 文件格式：
+```plaintext
+# 注释行将在下次读取时自动删除
+token1,checksum1
+token2,checksum2
+```
+
+文件管理原则：
+- 系统自动维护文件内容
+- 仅以下情况需要手动编辑：
+  - 删除特定 token
+  - 绑定已有 checksum 到指定 token
+
+## 模型支持列表
+```json
+[
+  "claude-3.5-sonnet",
+  "gpt-4",
+  "gpt-4o",
+  "cursor-fast",
+  "gpt-4o-mini",
+  "deepseek-v3"
+]
+```
+*注：模型列表为固定配置，暂不支持自定义扩展*
+
+## API 文档
+
+### 基础对话接口
+**Endpoint**
+`POST /v1/chat/completions`
+
+**认证方式**
+`Bearer Token` 三级认证机制：
+1. 环境变量 `AUTH_TOKEN`
+2. `.token` 文件轮询
+3. 直接 token,checksum 认证（v0.1.3-rc.3+）
+
+**请求示例**
+```json
+{
+  "model": "gpt-4",
+  "messages": [
+    {
+      "role": "user",
+      "content": "解释量子计算的基本原理"
+    }
+  ],
+  "stream": false
+}
+```
+
+**响应示例（非流式）**
+```json
+{
+  "id": "chatcmpl-9Xy...",
+  "object": "chat.completion",
+  "created": 1628063500,
+  "model": "gpt-4",
+  "choices": [{
+    "index": 0,
+    "message": {
+      "role": "assistant",
+      "content": "量子计算基于量子比特..."
+    },
+    "finish_reason": "stop"
+  }]
+}
+```
+
+### Token 管理接口
+| 端点 | 方法 | 功能 |
+|------|------|-----|
+| `/tokens` | GET | Token 信息管理界面 |
+| `/tokens/update` | POST | 批量更新 Token 列表 |
+| `/tokens/add` | POST | 增量添加 Token |
+| `/tokens/delete` | POST | 删除指定 Token |
+
+```mermaid
+sequenceDiagram
+    participant Client
+    participant API
+    Client->>API: POST /tokens/add
+    API->>API: 验证Token有效性
+    API->>File: 写入.tokens
+    API-->>Client: 返回更新结果
+```
+
+## 高级功能
+
+### 动态密钥生成
+**Endpoint**
+`POST /build-key`
+
+**优势对比**
+| 特性 | 传统模式 | 动态密钥 |
+|------|---------|---------|
+| 密钥长度 | 较长 | 优化缩短 |
+| 配置扩展 | 无 | 支持自定义 |
+| 安全等级 | 基础 | 增强编码 |
+| 验证效率 | 预校验耗时 | 即时验证 |
+
+## 系统监控
+
+### 健康检查
+**Endpoint**
+`GET /health`
+
+**响应示例**
+```json
+{
+  "status": "success",
+  "version": "1.2.0",
+  "uptime": 86400,
+  "models": ["gpt-4", "claude-3.5"],
+  "endpoints": ["/v1/chat", "/tokens"]
+}
+```
+
+## 生态工具
+
+### 开发辅助工具
+- [Token 获取工具](https://github.com/wisdgod/cursor-api/tree/main/tools/get-token)
+  支持 Windows/Linux/macOS 系统
+- [遥测数据重置工具](https://github.com/wisdgod/cursor-api/tree/main/tools/reset-telemetry)
+  清除用户使用数据记录
+
+## 致谢声明
+本项目的发展离不开以下开源项目的启发：
+- [zhx47/cursor-api](https://github.com/zhx47/cursor-api) - 基础架构参考
+- [cursorToApi](https://github.com/luolazyandlazy/cursorToApi) - 认证机制优化方案
+
+--- 
+
+> **项目维护说明**
+> 我们欢迎社区贡献，但请注意：
+> 1. 功能请求需附带使用场景说明
+> 2. Bug 报告请提供复现步骤和环境信息
+> 3. 重要变更需通过 CI/CD 测试流程
--- a/Deno.ts
+++ b/Deno.ts
@@ -1,55 +0,0 @@
-// 定义允许的主机和路径
-const ALLOWED_HOSTS = ["api2.cursor.sh", "www.cursor.com"];
-const ALLOWED_PATHS = [
-  "/aiserver.v1.AiService/StreamChat",
-  "/auth/full_stripe_profile",
-  "/api/usage",
-  "/api/auth/me"
-];
-
-// 创建统一的响应处理函数
-const createResponse = (status: number, message: string) => 
-  new Response(message, {
-    status,
-    headers: { "Access-Control-Allow-Origin": "*" }
-  });
-
-// 主处理函数
-Deno.serve(async (request: Request) => {
-  // 验证目标主机
-  const targetHost = request.headers.get("x-co");
-  if (!targetHost) return createResponse(400, "Missing header");
-  if (!ALLOWED_HOSTS.includes(targetHost)) return createResponse(403, "Host denied");
-
-  // 验证请求路径
-  const url = new URL(request.url);
-  if (!ALLOWED_PATHS.includes(url.pathname)) return createResponse(404, "Path invalid");
-
-  // 处理请求头
-  const headers = new Headers(request.headers);
-  headers.delete("x-co");
-  headers.set("Host", targetHost);
-
-  try {
-    // 转发请求
-    const response = await fetch(
-      `https://${targetHost}${url.pathname}${url.search}`,
-      {
-        method: request.method,
-        headers,
-        body: request.body
-      }
-    );
-
-    // 处理响应头
-    const responseHeaders = new Headers(response.headers);
-    responseHeaders.set("Access-Control-Allow-Origin", "*");
-
-    return new Response(response.body, {
-      status: response.status,
-      headers: responseHeaders
-    });
-  } catch (error) {
-    return createResponse(500, "Server error");
-  }
-});
--- a/4
+++ b/4
@@ -6,7 +6,7 @@ RUN apt-get update && \
    build-essential protobuf-compiler pkg-config libssl-dev nodejs npm \
    && rm -rf /var/lib/apt/lists/*
 COPY . .
-ENV RUSTFLAGS="-C link-arg=-s"
+ENV RUSTFLAGS="-C link-arg=-s -C target-cpu=native"
 RUN cargo build --release && \
    cp target/release/cursor-api /app/cursor-api

@@ -18,7 +18,7 @@ RUN apt-get update && \
    build-essential protobuf-compiler pkg-config libssl-dev nodejs npm \
    && rm -rf /var/lib/apt/lists/*
 COPY . .
-ENV RUSTFLAGS="-C link-arg=-s"
+ENV RUSTFLAGS="-C link-arg=-s -C target-cpu=native"
 RUN cargo build --release && \
    cp target/release/cursor-api /app/cursor-api

--- a/Dockerfile.cross
+++ b/Dockerfile.cross
@@ -16,7 +16,7 @@ RUN apt-get update && \
    && rm -rf /var/lib/apt/lists/*

 # 设置环境变量 (如果需要)
-ENV RUSTFLAGS="-C link-arg=-s"
+# ENV RUSTFLAGS="-C link-arg=-s"

 # 设置 PROTOC 环境变量 (因为你的 build.rs 需要)
 ENV PROTOC=/usr/bin/protoc
--- a/Dockerfile.cross.arm64
+++ b/Dockerfile.cross.arm64
@@ -0,0 +1,30 @@
+# Dockerfile.cross
+
+FROM --platform=linux/arm64 rust:1.84.0-slim-bookworm
+
+WORKDIR /app
+
+# 安装必要的软件包
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends \
+    build-essential \
+    pkg-config \
+    libssl-dev \
+    protobuf-compiler \
+    && rm -rf /var/lib/apt/lists/*
+
+# 设置环境变量 (如果需要)
+# ENV RUSTFLAGS="-C link-arg=-s"
+
+# 设置 PROTOC 环境变量 (因为你的 build.rs 需要)
+ENV PROTOC=/usr/bin/protoc
+
+# 安装特定版本的 protoc (如果你需要特定版本，例如 29.3；否则可以删除这部分)
+# ENV PROTOC_VERSION=29.3
+# ENV PROTOC_ZIP=protoc-${PROTOC_VERSION}-linux-x86_64.zip
+# RUN wget https://github.com/protocolbuffers/protobuf/releases/download/v${PROTOC_VERSION}/${PROTOC_ZIP} -O /tmp/${PROTOC_ZIP} && \
+#     unzip /tmp/${PROTOC_ZIP} -d /usr && \
+#     rm /tmp/${PROTOC_ZIP}
+
+# 验证安装
+RUN protoc --version
--- a/README.md
+++ b/README.md
@@ -171,7 +171,43 @@ data: [DONE]
  "tokens": [
    {
      "token": "string",
-      "checksum": "string"
+      "checksum": "string",
+      "profile": { // 可能存在
+        "usage": {
+          "premium": {
+            "requests": number,
+            "requests_total": number,
+            "tokens": number,
+            "max_requests": number,
+            "max_tokens": number
+          },
+          "standard": {
+            "requests": number,
+            "requests_total": number,
+            "tokens": number,
+            "max_requests": number,
+            "max_tokens": number
+          },
+          "unknown": {
+            "requests": number,
+            "requests_total": number,
+            "tokens": number,
+            "max_requests": number,
+            "max_tokens": number
+          }
+        },
+        "user": {
+          "email": "string",
+          "name": "string",
+          "id": "string",
+          "updated_at": "string"
+        },
+        "stripe": {
+          "membership_type": "free" | "free_trial" | "pro" | "enterprise",
+          "payment_id": "string",
+          "days_remaining_on_trial": number
+        }
+      }
    }
  ],
  "tokens_count": number
@@ -282,14 +318,13 @@ data: [DONE]
 ```json
 {
  "auth_token": "string",  // 格式: {token},{checksum}
-  "enable_stream_check": boolean,  // 可选，启用流式响应首块检查
-  "include_stop_stream": boolean,  // 可选，包含停止流
  "disable_vision": boolean,       // 可选，禁用图片处理能力
  "enable_slow_pool": boolean,     // 可选，启用慢速池
  "usage_check_models": {          // 可选，使用量检查模型配置
    "type": "default" | "disabled" | "all" | "custom",
    "model_ids": "string"  // 当type为custom时生效，以逗号分隔的模型ID列表
-  }
+  },
+  "include_web_references": boolean
 }
 ```

@@ -357,8 +392,6 @@ data: [DONE]
    "type": "default" | "text" | "html",
    "content": "string"
  },
-  "enable_stream_check": boolean,
-  "include_stop_stream": boolean,
  "vision_ability": "none" | "base64" | "all", // "disabled" | "base64-only" | "base64-http"
  "enable_slow_pool": boolean,
  "enable_all_claude": boolean,
@@ -368,7 +401,8 @@ data: [DONE]
  },
  "enable_dynamic_key": boolean,
  "share_token": "string",
-  "proxies": "" | "system" | "proxy1,proxy2,..."
+  "proxies": "" | "system" | "proxy1,proxy2,...",
+  "include_web_references": boolean
 }
 ```

@@ -383,8 +417,6 @@ data: [DONE]
      "type": "default" | "text" | "html", // 对于js和css后两者是一样的
      "content": "string"
    },
-    "enable_stream_check": boolean,
-    "include_stop_stream": boolean,
    "vision_ability": "none" | "base64" | "all",
    "enable_slow_pool": boolean,
    "enable_all_claude": boolean,
@@ -394,7 +426,8 @@ data: [DONE]
    },
    "enable_dynamic_key": boolean,
    "share_token": "string",
-    "proxies": "" | "system" | "proxy1,proxy2,..."
+    "proxies": "" | "system" | "proxy1,proxy2,...",
+    "include_web_references": boolean
  }
 }
 ```
--- a/serve.ts
+++ b/serve.ts
@@ -0,0 +1 @@
+Deno.serve(async(r:Request)=>{const rs=(s:number,m:string)=>new Response(m,{status:s,headers:{"Access-Control-Allow-Origin":"*"}});const h=r.headers.get("x-co");if(!h)return rs(400,"Missing header");const a=["api2.cursor.sh","www.cursor.com"];if(!a.includes(h))return rs(403,"Host denied");const u=new URL(r.url),p=["/aiserver.v1.AiService/StreamChat","/aiserver.v1.AiService/StreamChatWeb","/auth/full_stripe_profile","/api/usage","/api/auth/me"];if(!p.includes(u.pathname))return rs(404,"Path invalid");const hd=new Headers(r.headers);hd.delete("x-co");hd.set("Host",h);try{const f=await fetch(`https://${h}${u.pathname}${u.search}`,{method:r.method,headers:hd,body:r.body});const fh=new Headers(f.headers);fh.set("Access-Control-Allow-Origin","*");return new Response(f.body,{status:f.status,headers:fh})}catch(e){return rs(500,"Server error")}});
--- a/src/app/config.rs
+++ b/src/app/config.rs
@@ -65,8 +65,6 @@ pub async fn handle_config_update(
            status: ApiStatus::Success,
            data: Some(ConfigData {
                page_content: AppConfig::get_page_content(&request.path),
-                enable_stream_check: AppConfig::get_stream_check(),
-                include_stop_stream: AppConfig::get_stop_stream(),
                vision_ability: AppConfig::get_vision_ability(),
                enable_slow_pool: AppConfig::get_slow_pool(),
                enable_all_claude: AppConfig::get_allow_claude(),
@@ -74,6 +72,7 @@ pub async fn handle_config_update(
                enable_dynamic_key: AppConfig::get_dynamic_key(),
                share_token: AppConfig::get_share_token(),
                proxies: AppConfig::get_proxies(),
+                include_web_references: AppConfig::get_web_refs(),
            }),
            message: None,
        })),
@@ -96,8 +95,6 @@ pub async fn handle_config_update(
            }

            handle_updates!(request,
-                enable_stream_check => AppConfig::update_stream_check,
-                include_stop_stream => AppConfig::update_stop_stream,
                vision_ability => AppConfig::update_vision_ability,
                enable_slow_pool => AppConfig::update_slow_pool,
                enable_all_claude => AppConfig::update_allow_claude,
@@ -105,6 +102,7 @@ pub async fn handle_config_update(
                enable_dynamic_key => AppConfig::update_dynamic_key,
                share_token => AppConfig::update_share_token,
                proxies => AppConfig::update_proxies,
+                include_web_references => AppConfig::update_web_refs,
            );

            Ok(Json(NormalResponse {
@@ -131,8 +129,6 @@ pub async fn handle_config_update(
            }

            handle_resets!(request,
-                enable_stream_check => AppConfig::reset_stream_check,
-                include_stop_stream => AppConfig::reset_stop_stream,
                vision_ability => AppConfig::reset_vision_ability,
                enable_slow_pool => AppConfig::reset_slow_pool,
                enable_all_claude => AppConfig::reset_allow_claude,
@@ -140,6 +136,7 @@ pub async fn handle_config_update(
                enable_dynamic_key => AppConfig::reset_dynamic_key,
                share_token => AppConfig::reset_share_token,
                proxies => AppConfig::reset_proxies,
+                include_web_references => AppConfig::reset_web_refs,
            );

            Ok(Json(NormalResponse {
--- a/src/app/lazy.rs
+++ b/src/app/lazy.rs
@@ -108,6 +108,12 @@ def_cursor_api_url!(
    "/aiserver.v1.AiService/StreamChat"
 );

+def_cursor_api_url!(
+    CURSOR_API2_CHAT_WEB_URL,
+    CURSOR_API2_HOST,
+    "/aiserver.v1.AiService/StreamChatWeb"
+);
+
 def_cursor_api_url!(
    CURSOR_API2_STRIPE_URL,
    CURSOR_API2_HOST,
@@ -118,6 +124,12 @@ def_cursor_api_url!(CURSOR_USAGE_API_URL, CURSOR_HOST, "/api/usage");

 def_cursor_api_url!(CURSOR_USER_API_URL, CURSOR_HOST, "/api/auth/me");

+pub(super) static LOGS_FILE_PATH: LazyLock<String> =
+    LazyLock::new(|| parse_string_from_env("LOGS_FILE_PATH", "logs.bin"));
+
+pub(super) static PAGES_FILE_PATH: LazyLock<String> =
+    LazyLock::new(|| parse_string_from_env("PAGES_FILE_PATH", "pages.bin"));
+
 pub static DEBUG: LazyLock<bool> = LazyLock::new(|| parse_bool_from_env("DEBUG", false));

 // 使用环境变量 "DEBUG_LOG_FILE" 来指定日志文件路径，默认值为 "debug.log"
--- a/src/app/model.rs
+++ b/src/app/model.rs
@@ -12,18 +12,22 @@ use crate::{
    },
 };
 use parking_lot::RwLock;
+use rkyv::{Archive, Deserialize as RkyvDeserialize, Serialize as RkyvSerialize};
 use serde::{Deserialize, Serialize};
 use std::sync::LazyLock;

 mod usage_check;
 pub use usage_check::UsageCheck;
+mod config;
 mod proxies;
 pub use proxies::Proxies;
 mod build_key;
 pub use build_key::*;

+use super::constant::{STATUS_FAILED, STATUS_PENDING, STATUS_SUCCESS};
+
 // 页面内容类型枚举
-#[derive(Clone, Serialize, Deserialize)]
+#[derive(Clone, Serialize, Deserialize, Archive, RkyvDeserialize, RkyvSerialize)]
 #[serde(tag = "type", content = "content")]
 pub enum PageContent {
    #[serde(rename = "default")]
@@ -41,10 +45,8 @@ impl Default for PageContent {
 }

 // 静态配置
-#[derive(Clone)]
+#[derive(Default, Clone)]
 pub struct AppConfig {
-    stream_check: bool,
-    stop_stream: bool,
    vision_ability: VisionAbility,
    slow_pool: bool,
    allow_claude: bool,
@@ -54,9 +56,10 @@ pub struct AppConfig {
    share_token: String,
    is_share: bool,
    proxies: Proxies,
+    web_refs: bool,
 }

-#[derive(Serialize, Deserialize, Clone, PartialEq)]
+#[derive(Serialize, Deserialize, Clone, Copy, PartialEq)]
 pub enum VisionAbility {
    #[serde(rename = "none", alias = "disabled")]
    None,
@@ -87,7 +90,7 @@ impl Default for VisionAbility {
    }
 }

-#[derive(Clone, Default)]
+#[derive(Clone, Default, Archive, RkyvDeserialize, RkyvSerialize)]
 pub struct Pages {
    pub root_content: PageContent,
    pub logs_content: PageContent,
@@ -114,24 +117,6 @@ pub struct AppState {
 pub static APP_CONFIG: LazyLock<RwLock<AppConfig>> =
    LazyLock::new(|| RwLock::new(AppConfig::default()));

-impl Default for AppConfig {
-    fn default() -> Self {
-        Self {
-            stream_check: true,
-            stop_stream: true,
-            vision_ability: VisionAbility::Base64,
-            slow_pool: false,
-            allow_claude: false,
-            pages: Pages::default(),
-            usage_check: UsageCheck::Default,
-            dynamic_key: false,
-            share_token: String::default(),
-            is_share: false,
-            proxies: Proxies::default(),
-        }
-    }
-}
-
 macro_rules! config_methods {
    ($($field:ident: $type:ty, $default:expr;)*) => {
        $(
@@ -207,8 +192,6 @@ macro_rules! config_methods_clone {
 impl AppConfig {
    pub fn init() {
        let mut config = APP_CONFIG.write();
-        config.stream_check = parse_bool_from_env("ENABLE_STREAM_CHECK", true);
-        config.stop_stream = parse_bool_from_env("INCLUDE_STOP_REASON_STREAM", true);
        config.vision_ability =
            VisionAbility::from_str(&parse_string_from_env("VISION_ABILITY", EMPTY_STRING));
        config.slow_pool = parse_bool_from_env("ENABLE_SLOW_POOL", false);
@@ -221,15 +204,15 @@ impl AppConfig {
        config.proxies = match std::env::var("PROXIES") {
            Ok(proxies) => Proxies::from_str(proxies.as_str()),
            Err(_) => Proxies::default(),
-        }
+        };
+        config.web_refs = parse_bool_from_env("INCLUDE_WEB_REFERENCES", false)
    }

    config_methods! {
-        stream_check: bool, true;
-        stop_stream: bool, true;
        slow_pool: bool, false;
        allow_claude: bool, false;
        dynamic_key: bool, false;
+        web_refs: bool, false;
    }

    config_methods_clone! {
@@ -341,11 +324,20 @@ impl AppConfig {

 impl AppState {
    pub fn new(token_infos: Vec<TokenInfo>) -> Self {
+        // 尝试加载保存的日志
+        let request_logs = tokio::task::block_in_place(|| {
+            tokio::runtime::Handle::current()
+                .block_on(async { Self::load_saved_logs().await.unwrap_or_default() })
+        });
+
        Self {
-            total_requests: 0,
+            total_requests: request_logs.len() as u64,
            active_requests: 0,
-            error_requests: 0,
-            request_logs: Vec::new(),
+            error_requests: request_logs
+                .iter()
+                .filter(|log| matches!(log.status, LogStatus::Failed))
+                .count() as u64,
+            request_logs,
            token_infos,
        }
    }
@@ -357,8 +349,43 @@ impl AppState {
    }
 }

+#[derive(Clone, Archive, RkyvDeserialize, RkyvSerialize)]
+pub enum LogStatus {
+    Pending,
+    Success,
+    Failed,
+}
+
+impl Serialize for LogStatus {
+    fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
+    where
+        S: serde::Serializer,
+    {
+        serializer.serialize_str(self.as_str_name())
+    }
+}
+
+impl LogStatus {
+    pub fn as_str_name(&self) -> &'static str {
+        match self {
+            Self::Pending => STATUS_PENDING,
+            Self::Success => STATUS_SUCCESS,
+            Self::Failed => STATUS_FAILED,
+        }
+    }
+
+    pub fn from_str_name(s: &str) -> Option<Self> {
+        match s {
+            STATUS_PENDING => Some(Self::Pending),
+            STATUS_SUCCESS => Some(Self::Success),
+            STATUS_FAILED => Some(Self::Failed),
+            _ => None,
+        }
+    }
+}
+
 // 请求日志
-#[derive(Serialize, Clone)]
+#[derive(Serialize, Clone, Archive, RkyvDeserialize, RkyvSerialize)]
 pub struct RequestLog {
    pub id: u64,
    pub timestamp: chrono::DateTime<chrono::Local>,
@@ -368,12 +395,12 @@ pub struct RequestLog {
    pub prompt: Option<String>,
    pub timing: TimingInfo,
    pub stream: bool,
-    pub status: &'static str,
+    pub status: LogStatus,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub error: Option<String>,
 }

-#[derive(Serialize, Clone)]
+#[derive(Serialize, Clone, Archive, RkyvDeserialize, RkyvSerialize)]
 pub struct TimingInfo {
    pub total: f64, // 总用时(秒)
    #[serde(skip_serializing_if = "Option::is_none")]
@@ -390,7 +417,7 @@ pub struct ChatRequest {
 }

 // 用于存储 token 信息
-#[derive(Serialize, Clone)]
+#[derive(Serialize, Clone, Archive, RkyvDeserialize, RkyvSerialize)]
 pub struct TokenInfo {
    pub token: String,
    pub checksum: String,
--- a/src/app/model/build_key.rs
+++ b/src/app/model/build_key.rs
@@ -4,23 +4,15 @@ use crate::{app::constant::COMMA, chat::constant::AVAILABLE_MODELS};

 #[derive(Deserialize)]
 pub struct BuildKeyRequest {
-    // 认证令牌(必需)
    pub auth_token: String,
-    // 流第一个块检查
-    #[serde(default)]
-    pub enable_stream_check: Option<bool>,
-    // 包含停止流
-    #[serde(default)]
-    pub include_stop_stream: Option<bool>,
-    // 是否禁用图片处理能力
    #[serde(default)]
    pub disable_vision: Option<bool>,
-    // 慢速池
    #[serde(default)]
    pub enable_slow_pool: Option<bool>,
-    // 使用量检查模型规则
    #[serde(default)]
    pub usage_check_models: Option<UsageCheckModelConfig>,
+    #[serde(default)]
+    pub include_web_references: Option<bool>,
 }
 pub struct UsageCheckModelConfig {
    pub model_type: UsageCheckModelType,
--- a/src/app/model/config.rs
+++ b/src/app/model/config.rs
@@ -0,0 +1,114 @@
+use memmap2::{MmapMut, MmapOptions};
+use rkyv::{archived_root, Deserialize as _};
+use std::fs::OpenOptions;
+
+use crate::app::lazy::{LOGS_FILE_PATH, PAGES_FILE_PATH};
+
+use super::{AppConfig, AppState, Pages, RequestLog, APP_CONFIG};
+
+impl AppState {
+    // 保存日志的方法
+    pub(crate) async fn save_logs(&self) -> Result<(), Box<dyn std::error::Error>> {
+        // 序列化日志
+        let bytes = rkyv::to_bytes::<_, 256>(&self.request_logs)?;
+
+        // 创建或打开文件
+        let file = OpenOptions::new()
+            .read(true)
+            .write(true)
+            .create(true)
+            .open(LOGS_FILE_PATH.as_str())?;
+
+        // 添加大小检查
+        if bytes.len() > usize::MAX / 2 {
+            return Err("日志数据过大".into());
+        }
+
+        // 设置文件大小
+        file.set_len(bytes.len() as u64)?;
+
+        // 创建可写入的内存映射
+        let mut mmap = unsafe { MmapMut::map_mut(&file)? };
+
+        // 写入数据
+        mmap.copy_from_slice(&bytes);
+
+        // 同步到磁盘
+        mmap.flush()?;
+
+        Ok(())
+    }
+
+    // 加载日志的方法
+    pub(super) async fn load_saved_logs() -> Result<Vec<RequestLog>, Box<dyn std::error::Error>> {
+        let file = match OpenOptions::new().read(true).open(LOGS_FILE_PATH.as_str()) {
+            Ok(file) => file,
+            Err(e) if e.kind() == std::io::ErrorKind::NotFound => {
+                return Ok(Vec::new());
+            }
+            Err(e) => return Err(Box::new(e)),
+        };
+
+        // 添加文件大小检查
+        if file.metadata()?.len() > usize::MAX as u64 {
+            return Err("日志文件过大".into());
+        }
+
+        // 创建只读内存映射
+        let mmap = unsafe { MmapOptions::new().map(&file)? };
+
+        // 验证并反序列化数据
+        let archived = unsafe { archived_root::<Vec<RequestLog>>(&mmap) };
+        Ok(archived.deserialize(&mut rkyv::Infallible)?)
+    }
+}
+
+impl AppConfig {
+    pub fn save_config() -> Result<(), Box<dyn std::error::Error>> {
+        let pages = APP_CONFIG.read().pages.clone();
+        let bytes = rkyv::to_bytes::<_, 256>(&pages)?;
+
+        let file = OpenOptions::new()
+            .read(true)
+            .write(true)
+            .create(true)
+            .open(PAGES_FILE_PATH.as_str())?;
+
+        // 添加大小检查
+        if bytes.len() > usize::MAX / 2 {
+            return Err("配置数据过大".into());
+        }
+
+        file.set_len(bytes.len() as u64)?;
+
+        let mut mmap = unsafe { MmapMut::map_mut(&file)? };
+        mmap.copy_from_slice(&bytes);
+        mmap.flush()?;
+
+        Ok(())
+    }
+
+    pub fn load_saved_config() -> Result<(), Box<dyn std::error::Error>> {
+        let file = match OpenOptions::new().read(true).open(PAGES_FILE_PATH.as_str()) {
+            Ok(file) => file,
+            Err(e) if e.kind() == std::io::ErrorKind::NotFound => {
+                return Ok(());
+            }
+            Err(e) => return Err(Box::new(e)),
+        };
+
+        // 添加文件大小检查
+        if file.metadata()?.len() > usize::MAX as u64 {
+            return Err("配置文件过大".into());
+        }
+
+        let mmap = unsafe { MmapOptions::new().map(&file)? };
+
+        let archived = unsafe { archived_root::<Pages>(&mmap) };
+        let pages = archived.deserialize(&mut rkyv::Infallible)?;
+        let mut config = APP_CONFIG.write();
+        config.pages = pages;
+
+        Ok(())
+    }
+}
--- a/src/app/model/proxies.rs
+++ b/src/app/model/proxies.rs
@@ -1,6 +1,7 @@
 use reqwest::{Client, Proxy};
 use serde::{Serialize, Serializer};
 use serde::{Deserialize, Deserializer};
+// use rkyv::{Archive, Deserialize as RkyvDeserialize, Serialize as RkyvSerialize};

 use crate::app::constant::COMMA_STRING;

@@ -30,7 +31,7 @@ impl<'de> Deserialize<'de> for Proxies {
    where
        D: Deserializer<'de>,
    {
-        let s = String::deserialize(deserializer)?;
+        let s = <String as serde::Deserialize>::deserialize(deserializer)?;
        Ok(Proxies::from_str(&s))
    }
 }
--- a/src/app/model/usage_check.rs
+++ b/src/app/model/usage_check.rs
@@ -3,6 +3,7 @@ use crate::{
    chat::{config::key_config, constant::AVAILABLE_MODELS},
 };
 use serde::{Deserialize, Serialize};
+// use rkyv::{Archive, Deserialize as RkyvDeserialize, Serialize as RkyvSerialize};

 #[derive(Clone, PartialEq)]
 pub enum UsageCheck {
@@ -108,7 +109,7 @@ impl<'de> Deserialize<'de> for UsageCheck {
            Custom(String),
        }

-        let helper = UsageCheckHelper::deserialize(deserializer)?;
+        let helper = <UsageCheckHelper as serde::Deserialize>::deserialize(deserializer)?;
        Ok(match helper {
            UsageCheckHelper::None => UsageCheck::None,
            UsageCheckHelper::Default => UsageCheck::Default,
--- a/src/chat/adapter.rs
+++ b/src/chat/adapter.rs
@@ -15,8 +15,7 @@ use crate::{

 use super::{
    aiserver::v1::{
-        conversation_message, image_proto, AzureState, ConversationMessage, ExplicitContext,
-        GetChatRequest, ImageProto, ModelDetails,
+        conversation_message, image_proto, AzureState, ChatExternalLink, ConversationMessage, ExplicitContext, GetChatRequest, ImageProto, ModelDetails
    },
    constant::{ERR_UNSUPPORTED_GIF, ERR_UNSUPPORTED_IMAGE_FORMAT, LONG_CONTEXT_MODELS},
    model::{Message, MessageContent, Role},
@@ -25,7 +24,7 @@ use super::{
 async fn process_chat_inputs(
    inputs: Vec<Message>,
    disable_vision: bool,
-) -> (String, Vec<ConversationMessage>) {
+) -> (String, Vec<ConversationMessage>, Vec<String>) {
    // 收集 system 指令
    let instructions = inputs
        .iter()
@@ -98,9 +97,25 @@ async fn process_chat_inputs(
                file_diff_trajectories: vec![],
                conversation_summary: None,
            }],
+            vec![],
        );
    }

+    // 处理 WebReferences 开头的 assistant 消息
+    chat_inputs = chat_inputs
+        .into_iter()
+        .map(|mut input| {
+            if let (Role::Assistant, MessageContent::Text(text)) = (&input.role, &input.content) {
+                if text.starts_with("WebReferences:") {
+                    if let Some(pos) = text.find("\n\n") {
+                        input.content = MessageContent::Text(text[pos + 2..].to_owned());
+                    }
+                }
+            }
+            input
+        })
+        .collect();
+
    // 如果第一条是 assistant，插入空的 user 消息
    if chat_inputs
        .first()
@@ -226,7 +241,32 @@ async fn process_chat_inputs(
        });
    }

-    (instructions, messages)
+    let mut urls = Vec::new();
+    if let Some(last_msg) = messages.last() {
+        if last_msg.r#type == conversation_message::MessageType::Human as i32 {
+            let text = &last_msg.text;
+            let mut chars = text.chars().peekable();
+            
+            while let Some(c) = chars.next() {
+                if c == '@' {
+                    let mut url = String::new();
+                    while let Some(&next_char) = chars.peek() {
+                        if next_char.is_whitespace() {
+                            break;
+                        }
+                        url.push(chars.next().unwrap());
+                    }
+                    if let Ok(parsed_url) = url::Url::parse(&url) {
+                        if parsed_url.scheme() == "http" || parsed_url.scheme() == "https" {
+                            urls.push(url);
+                        }
+                    }
+                }
+            }
+        }
+    }
+
+    (instructions, messages, urls)
 }

 async fn fetch_image_data(
@@ -344,6 +384,7 @@ pub async fn encode_chat_message(
    model_name: &str,
    disable_vision: bool,
    enable_slow_pool: bool,
+    is_search: bool,
 ) -> Result<Vec<u8>, Box<dyn std::error::Error + Send + Sync>> {
    // 在进入异步操作前获取并释放锁
    let enable_slow_pool = {
@@ -354,7 +395,7 @@ pub async fn encode_chat_message(
        }
    };

-    let (instructions, messages) = process_chat_inputs(inputs, disable_vision).await;
+    let (instructions, messages, urls) = process_chat_inputs(inputs, disable_vision).await;

    let explicit_context = if !instructions.trim().is_empty() {
        Some(ExplicitContext {
@@ -365,6 +406,15 @@ pub async fn encode_chat_message(
        None
    };

+    let base_uuid = rand::random::<u16>();
+    let external_links = urls.into_iter().enumerate().map(|(i, url)| {
+        let uuid = base_uuid.wrapping_add(i as u16);
+        ChatExternalLink {
+            url,
+            uuid: uuid.to_string(),
+        }
+    }).collect();
+
    let chat = GetChatRequest {
        current_file: None,
        conversation: messages,
@@ -394,11 +444,15 @@ pub async fn encode_chat_message(
        is_bash: Some(false),
        conversation_id: Uuid::new_v4().to_string(),
        can_handle_filenames_after_language_ids: Some(true),
-        use_web: None,
+        use_web: if is_search {
+            Some("full_search".to_string())
+        } else {
+            None
+        },
        quotes: vec![],
        debug_info: None,
        workspace_id: None,
-        external_links: vec![],
+        external_links,
        commit_notes: vec![],
        long_context_mode: Some(LONG_CONTEXT_MODELS.contains(&model_name)),
        is_eval: Some(false),
--- a/src/chat/config.rs
+++ b/src/chat/config.rs
@@ -6,21 +6,14 @@ impl KeyConfig {
    pub fn new_with_global() -> Self {
        Self {
            auth_token: None,
-            enable_stream_check: Some(AppConfig::get_stream_check()),
-            include_stop_stream: Some(AppConfig::get_stop_stream()),
            disable_vision: Some(AppConfig::get_vision_ability().is_none()),
            enable_slow_pool: Some(AppConfig::get_slow_pool()),
            usage_check_models: None,
+            include_web_references: Some(AppConfig::get_web_refs()),
        }
    }

    pub fn copy_without_auth_token(&self, config: &mut Self) {
-        if self.enable_stream_check.is_some() {
-            config.enable_stream_check = self.enable_stream_check;
-        }
-        if self.include_stop_stream.is_some() {
-            config.include_stop_stream = self.include_stop_stream;
-        }
        if self.disable_vision.is_some() {
            config.disable_vision = self.disable_vision;
        }
@@ -30,5 +23,8 @@ impl KeyConfig {
        if self.usage_check_models.is_some() {
            config.usage_check_models = self.usage_check_models.clone();
        }
+        if self.include_web_references.is_some() {
+            config.include_web_references = self.include_web_references;
+        }
    }
 }
--- a/src/chat/config/key.proto
+++ b/src/chat/config/key.proto
@@ -17,12 +17,6 @@ message KeyConfig {
  // 认证令牌（必需）
  TokenInfo auth_token = 1;

-  // 是否启用流检查
-  optional bool enable_stream_check = 2;
-
-  // 是否包含停止流
-  optional bool include_stop_stream = 3;
-
  // 是否禁用图片处理能力
  optional bool disable_vision = 4;

@@ -44,6 +38,9 @@ message KeyConfig {
  // 使用量检查模型规则
  optional UsageCheckModel usage_check_models = 6;

+  // 包含网络引用
+  optional bool include_web_references = 7;
+
  // 密码SHA256哈希值
  // bytes secret = 2;
 }
--- a/src/chat/constant.rs
+++ b/src/chat/constant.rs
@@ -6,7 +6,10 @@ macro_rules! def_pub_const {
    };
 }
 def_pub_const!(ERR_UNSUPPORTED_GIF, "不支持动态 GIF");
-def_pub_const!(ERR_UNSUPPORTED_IMAGE_FORMAT, "不支持的图片格式，仅支持 PNG、JPEG、WEBP 和非动态 GIF");
+def_pub_const!(
+    ERR_UNSUPPORTED_IMAGE_FORMAT,
+    "不支持的图片格式，仅支持 PNG、JPEG、WEBP 和非动态 GIF"
+);
 def_pub_const!(ERR_NODATA, "No data");

 const MODEL_OBJECT: &str = "model";
@@ -45,146 +48,137 @@ def_pub_const!(GEMINI_2_0_FLASH_EXP, "gemini-2.0-flash-exp");
 def_pub_const!(DEEPSEEK_V3, "deepseek-v3");
 def_pub_const!(DEEPSEEK_R1, "deepseek-r1");

-pub const AVAILABLE_MODELS: [Model; 23] = [
+#[derive(Clone, PartialEq, rkyv::Archive, rkyv::Deserialize, rkyv::Serialize)]
+pub enum ModelType {
+    Claude35Sonnet,
+    Gpt4,
+    Gpt4o,
+    Claude3Opus,
+    CursorFast,
+    CursorSmall,
+    Gpt35Turbo,
+    Gpt4Turbo202404,
+    Gpt4o128k,
+    Gemini15Flash500k,
+    Claude3Haiku200k,
+    Claude35Sonnet200k,
+    Claude35Sonnet20241022,
+    Gpt4oMini,
+    O1Mini,
+    O1Preview,
+    O1,
+    Claude35Haiku,
+    GeminiExp1206,
+    Gemini20FlashThinkingExp,
+    Gemini20FlashExp,
+    DeepseekV3,
+    DeepseekR1,
+}
+
+macro_rules! create_model {
+    ($($id:expr, $owner:expr),* $(,)?) => {
+        pub const AVAILABLE_MODELS: [Model; count!($( ($id, $owner) )*)] = [
+            $(
                Model {
-        id: CLAUDE_3_5_SONNET,
+                    id: $id,
                    created: CREATED,
                    object: MODEL_OBJECT,
-        owned_by: ANTHROPIC,
+                    owned_by: $owner,
                },
-    Model {
-        id: GPT_4,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: OPENAI,
-    },
-    Model {
-        id: GPT_4O,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: OPENAI,
-    },
-    Model {
-        id: CLAUDE_3_OPUS,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: ANTHROPIC,
-    },
-    Model {
-        id: CURSOR_FAST,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: CURSOR,
-    },
-    Model {
-        id: CURSOR_SMALL,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: CURSOR,
-    },
-    Model {
-        id: GPT_3_5_TURBO,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: OPENAI,
-    },
-    Model {
-        id: GPT_4_TURBO_2024_04_09,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: OPENAI,
-    },
-    Model {
-        id: GPT_4O_128K,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: OPENAI,
-    },
-    Model {
-        id: GEMINI_1_5_FLASH_500K,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: GOOGLE,
-    },
-    Model {
-        id: CLAUDE_3_HAIKU_200K,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: ANTHROPIC,
-    },
-    Model {
-        id: CLAUDE_3_5_SONNET_200K,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: ANTHROPIC,
-    },
-    Model {
-        id: CLAUDE_3_5_SONNET_20241022,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: ANTHROPIC,
-    },
-    Model {
-        id: GPT_4O_MINI,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: OPENAI,
-    },
-    Model {
-        id: O1_MINI,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: OPENAI,
-    },
-    Model {
-        id: O1_PREVIEW,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: OPENAI,
-    },
-    Model {
-        id: O1,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: OPENAI,
-    },
-    Model {
-        id: CLAUDE_3_5_HAIKU,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: ANTHROPIC,
-    },
-    Model {
-        id: GEMINI_EXP_1206,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: GOOGLE,
-    },
-    Model {
-        id: GEMINI_2_0_FLASH_THINKING_EXP,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: GOOGLE,
-    },
-    Model {
-        id: GEMINI_2_0_FLASH_EXP,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: GOOGLE,
-    },
-    Model {
-        id: DEEPSEEK_V3,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: DEEPSEEK,
-    },
-    Model {
-        id: DEEPSEEK_R1,
-        created: CREATED,
-        object: MODEL_OBJECT,
-        owned_by: DEEPSEEK,
-    },
-];
+            )*
+        ];
+    };
+}
+
+macro_rules! count {
+    () => (0);
+    (($id:expr, $owner:expr) $( ($id2:expr, $owner2:expr) )*) => (1 + count!($( ($id2, $owner2) )*));
+}
+
+impl ModelType {
+    pub fn as_str_name(&self) -> &'static str {
+        match self {
+            ModelType::Claude35Sonnet => CLAUDE_3_5_SONNET,
+            ModelType::Gpt4 => GPT_4,
+            ModelType::Gpt4o => GPT_4O,
+            ModelType::Claude3Opus => CLAUDE_3_OPUS,
+            ModelType::CursorFast => CURSOR_FAST,
+            ModelType::CursorSmall => CURSOR_SMALL,
+            ModelType::Gpt35Turbo => GPT_3_5_TURBO,
+            ModelType::Gpt4Turbo202404 => GPT_4_TURBO_2024_04_09,
+            ModelType::Gpt4o128k => GPT_4O_128K,
+            ModelType::Gemini15Flash500k => GEMINI_1_5_FLASH_500K,
+            ModelType::Claude3Haiku200k => CLAUDE_3_HAIKU_200K,
+            ModelType::Claude35Sonnet200k => CLAUDE_3_5_SONNET_200K,
+            ModelType::Claude35Sonnet20241022 => CLAUDE_3_5_SONNET_20241022,
+            ModelType::Gpt4oMini => GPT_4O_MINI,
+            ModelType::O1Mini => O1_MINI,
+            ModelType::O1Preview => O1_PREVIEW,
+            ModelType::O1 => O1,
+            ModelType::Claude35Haiku => CLAUDE_3_5_HAIKU,
+            ModelType::GeminiExp1206 => GEMINI_EXP_1206,
+            ModelType::Gemini20FlashThinkingExp => GEMINI_2_0_FLASH_THINKING_EXP,
+            ModelType::Gemini20FlashExp => GEMINI_2_0_FLASH_EXP,
+            ModelType::DeepseekV3 => DEEPSEEK_V3,
+            ModelType::DeepseekR1 => DEEPSEEK_R1,
+        }
+    }
+
+    pub fn from_str_name(id :&str) -> Option<ModelType> {
+        match id {
+            CLAUDE_3_5_SONNET => Some(ModelType::Claude35Sonnet),
+            GPT_4 => Some(ModelType::Gpt4),
+            GPT_4O => Some(ModelType::Gpt4o),
+            CLAUDE_3_OPUS => Some(ModelType::Claude3Opus),
+            CURSOR_FAST => Some(ModelType::CursorFast),
+            CURSOR_SMALL => Some(ModelType::CursorSmall),
+            GPT_3_5_TURBO => Some(ModelType::Gpt35Turbo),
+            GPT_4_TURBO_2024_04_09 => Some(ModelType::Gpt4Turbo202404),
+            GPT_4O_128K => Some(ModelType::Gpt4o128k),
+            GEMINI_1_5_FLASH_500K => Some(ModelType::Gemini15Flash500k),
+            CLAUDE_3_HAIKU_200K => Some(ModelType::Claude3Haiku200k),
+            CLAUDE_3_5_SONNET_200K => Some(ModelType::Claude35Sonnet200k),
+            CLAUDE_3_5_SONNET_20241022 => Some(ModelType::Claude35Sonnet20241022),
+            GPT_4O_MINI => Some(ModelType::Gpt4oMini),
+            O1_MINI => Some(ModelType::O1Mini),
+            O1_PREVIEW => Some(ModelType::O1Preview),
+            O1 => Some(ModelType::O1),
+            CLAUDE_3_5_HAIKU => Some(ModelType::Claude35Haiku),
+            GEMINI_EXP_1206 => Some(ModelType::GeminiExp1206),
+            GEMINI_2_0_FLASH_THINKING_EXP => Some(ModelType::Gemini20FlashThinkingExp),
+            GEMINI_2_0_FLASH_EXP => Some(ModelType::Gemini20FlashExp),
+            DEEPSEEK_V3 => Some(ModelType::DeepseekV3),
+            DEEPSEEK_R1 => Some(ModelType::DeepseekR1),
+            _ => None,
+        }
+    }
+}
+
+create_model!(
+    CLAUDE_3_5_SONNET, ANTHROPIC,
+    GPT_4, OPENAI,
+    GPT_4O, OPENAI,
+    CLAUDE_3_OPUS, ANTHROPIC,
+    CURSOR_FAST, CURSOR,
+    CURSOR_SMALL, CURSOR,
+    GPT_3_5_TURBO, OPENAI,
+    GPT_4_TURBO_2024_04_09, OPENAI,
+    GPT_4O_128K, OPENAI,
+    GEMINI_1_5_FLASH_500K, GOOGLE,
+    CLAUDE_3_HAIKU_200K, ANTHROPIC,
+    CLAUDE_3_5_SONNET_200K, ANTHROPIC,
+    CLAUDE_3_5_SONNET_20241022, ANTHROPIC,
+    GPT_4O_MINI, OPENAI,
+    O1_MINI, OPENAI,
+    O1_PREVIEW, OPENAI,
+    O1, OPENAI,
+    CLAUDE_3_5_HAIKU, ANTHROPIC,
+    GEMINI_EXP_1206, GOOGLE,
+    GEMINI_2_0_FLASH_THINKING_EXP, GOOGLE,
+    GEMINI_2_0_FLASH_EXP, GOOGLE,
+    DEEPSEEK_V3, DEEPSEEK,
+    DEEPSEEK_R1, DEEPSEEK,
+);

 pub const USAGE_CHECK_MODELS: [&str; 11] = [
    CLAUDE_3_5_SONNET_20241022,
--- a/src/chat/error.rs
+++ b/src/chat/error.rs
@@ -125,7 +125,6 @@ impl ErrorResponse {
 pub enum StreamError {
    ChatError(ChatError),
    DataLengthLessThan5,
-    EmptyMessage,
 }

 impl std::fmt::Display for StreamError {
@@ -133,7 +132,6 @@ impl std::fmt::Display for StreamError {
        match self {
            StreamError::ChatError(error) => write!(f, "{}", error.error.code),
            StreamError::DataLengthLessThan5 => write!(f, "data length less than 5"),
-            StreamError::EmptyMessage => write!(f, "empty message"),
        }
    }
 }
--- a/src/chat/model.rs
+++ b/src/chat/model.rs
@@ -86,8 +86,8 @@ pub struct Model {
    pub owned_by: &'static str,
 }

-use crate::app::model::{AppConfig, UsageCheck};
 use super::constant::USAGE_CHECK_MODELS;
+use crate::app::model::{AppConfig, UsageCheck};

 impl Model {
    pub fn is_usage_check(&self, usage_check: Option<UsageCheck>) -> bool {
--- a/src/chat/route/config.rs
+++ b/src/chat/route/config.rs
@@ -164,11 +164,10 @@ pub async fn handle_build_key(
    // 构建 proto 消息
    let mut key_config = KeyConfig {
        auth_token: Some(token_info),
-        enable_stream_check: request.enable_stream_check,
-        include_stop_stream: request.include_stop_stream,
        disable_vision: request.disable_vision,
        enable_slow_pool: request.enable_slow_pool,
        usage_check_models: None,
+        include_web_references: request.include_web_references,
    };

    if let Some(usage_check_models) = request.usage_check_models {
--- a/src/chat/service.rs
+++ b/src/chat/service.rs
@@ -2,12 +2,13 @@ use crate::{
    app::{
        constant::{
            AUTHORIZATION_BEARER_PREFIX, FINISH_REASON_STOP, OBJECT_CHAT_COMPLETION,
-            OBJECT_CHAT_COMPLETION_CHUNK, STATUS_FAILED, STATUS_PENDING, STATUS_SUCCESS,
+            OBJECT_CHAT_COMPLETION_CHUNK,
        },
-        lazy::{
-            AUTH_TOKEN, KEY_PREFIX, KEY_PREFIX_LEN, REQUEST_LOGS_LIMIT, SERVICE_TIMEOUT,
+        lazy::{AUTH_TOKEN, KEY_PREFIX, KEY_PREFIX_LEN, REQUEST_LOGS_LIMIT, SERVICE_TIMEOUT},
+        model::{
+            AppConfig, AppState, ChatRequest, LogStatus, RequestLog, TimingInfo, TokenInfo,
+            UsageCheck,
        },
-        model::{AppConfig, AppState, ChatRequest, RequestLog, TimingInfo, TokenInfo, UsageCheck},
    },
    chat::{
        config::KeyConfig,
@@ -16,11 +17,11 @@ use crate::{
        model::{
            ChatResponse, Choice, Delta, Message, MessageContent, ModelsResponse, Role, Usage,
        },
-        stream::{parse_stream_data, StreamMessage},
+        stream::{StreamDecoder, StreamMessage},
    },
    common::{
        client::build_client,
-        model::{error::ChatError, userinfo::MembershipType, ErrorResponse},
+        model::{error::ChatError, userinfo::MembershipType, ApiStatus, ErrorResponse},
        utils::{
            format_time_ms, from_base64, get_token_profile, tokeninfo_to_token,
            validate_token_and_checksum,
@@ -38,16 +39,13 @@ use axum::{
    Json,
 };
 use bytes::Bytes;
-use futures::{Stream, StreamExt};
+use futures::StreamExt;
 use prost::Message as _;
+use std::sync::atomic::{AtomicUsize, Ordering};
 use std::{
    convert::Infallible,
    sync::{atomic::AtomicBool, Arc},
 };
-use std::{
-    pin::Pin,
-    sync::atomic::{AtomicUsize, Ordering},
-};
 use tokio::sync::Mutex;
 use uuid::Uuid;

@@ -66,8 +64,16 @@ pub async fn handle_chat(
    Json(request): Json<ChatRequest>,
 ) -> Result<Response<Body>, (StatusCode, Json<ErrorResponse>)> {
    let allow_claude = AppConfig::get_allow_claude();
+
+    let is_search = request.model.ends_with("-online");
+    let model_name = if is_search {
+        request.model[..request.model.len() - 7].to_string()
+    } else {
+        request.model.clone()
+    };
+
    // 验证模型是否支持并获取模型信息
-    let model = AVAILABLE_MODELS.iter().find(|m| m.id == request.model);
+    let model = AVAILABLE_MODELS.iter().find(|m| m.id == model_name);
    let model_supported = model.is_some();

    if !(model_supported || allow_claude && request.model.starts_with("claude")) {
@@ -168,7 +174,7 @@ pub async fn handle_chat(
                    return false;
                }

-                let is_premium = USAGE_CHECK_MODELS.contains(&request.model.as_str());
+                let is_premium = USAGE_CHECK_MODELS.contains(&model_name.as_str());
                let standard = &profile.usage.standard;
                let premium = &profile.usage.premium;

@@ -213,14 +219,28 @@ pub async fn handle_chat(
            tokio::spawn(async move {
                let profile = get_token_profile(&auth_token_clone).await;
                let mut state = state_clone.lock().await;
-                // 根据id查找对应的日志
-                if let Some(log) = state
-                    .request_logs
-                    .iter_mut()
-                    .rev()
-                    .find(|log| log.id == log_id)
-                {
-                    log.token_info.profile = profile;
+
+                // 先找到所有需要更新的位置的索引
+                let token_info_idx = state
+                    .token_infos
+                    .iter()
+                    .position(|info| info.token == auth_token_clone);
+
+                let log_idx = state.request_logs.iter().rposition(|log| log.id == log_id);
+
+                // 根据索引更新
+                match (token_info_idx, log_idx) {
+                    (Some(t_idx), Some(l_idx)) => {
+                        state.token_infos[t_idx].profile = profile.clone();
+                        state.request_logs[l_idx].token_info.profile = profile;
+                    }
+                    (Some(t_idx), None) => {
+                        state.token_infos[t_idx].profile = profile;
+                    }
+                    (None, Some(l_idx)) => {
+                        state.request_logs[l_idx].token_info.profile = profile;
+                    }
+                    (None, None) => {}
                }
            });
        }
@@ -240,7 +260,7 @@ pub async fn handle_chat(
                first: None,
            },
            stream: request.stream,
-            status: STATUS_PENDING,
+            status: LogStatus::Pending,
            error: None,
        });

@@ -252,9 +272,10 @@ pub async fn handle_chat(
    // 将消息转换为hex格式
    let hex_data = match super::adapter::encode_chat_message(
        request.messages,
-        &request.model,
+        &model_name,
        current_config.disable_vision(),
        current_config.enable_slow_pool(),
+        is_search,
    )
    .await
    {
@@ -267,7 +288,7 @@ pub async fn handle_chat(
                .rev()
                .find(|log| log.id == current_id)
            {
-                log.status = STATUS_FAILED;
+                log.status = LogStatus::Failed;
                log.error = Some(e.to_string());
            }
            state.active_requests -= 1;
@@ -282,7 +303,7 @@ pub async fn handle_chat(
    };

    // 构建请求客户端
-    let client = build_client(&auth_token, &checksum);
+    let client = build_client(&auth_token, &checksum, is_search);
    // 添加超时设置
    let response = tokio::time::timeout(
        std::time::Duration::from_secs(*SERVICE_TIMEOUT),
@@ -303,7 +324,7 @@ pub async fn handle_chat(
                        .rev()
                        .find(|log| log.id == current_id)
                    {
-                        log.status = STATUS_SUCCESS;
+                        log.status = LogStatus::Success;
                    }
                }
                resp
@@ -318,7 +339,7 @@ pub async fn handle_chat(
                        .rev()
                        .find(|log| log.id == current_id)
                    {
-                        log.status = STATUS_FAILED;
+                        log.status = LogStatus::Failed;
                        log.error = Some(e.to_string());
                    }
                    state.active_requests -= 1;
@@ -340,7 +361,7 @@ pub async fn handle_chat(
                    .rev()
                    .find(|log| log.id == current_id)
                {
-                    log.status = STATUS_FAILED;
+                    log.status = LogStatus::Failed;
                    log.error = Some("Request timeout".to_string());
                }
                state.active_requests -= 1;
@@ -359,149 +380,60 @@ pub async fn handle_chat(
        state.active_requests -= 1;
    }

+    let convert_web_ref = current_config.include_web_references();
+
    if request.stream {
        let response_id = format!("chatcmpl-{}", Uuid::new_v4().simple());
-        let full_text = Arc::new(Mutex::new(String::with_capacity(1024)));
        let is_start = Arc::new(AtomicBool::new(true));
        let start_time = std::time::Instant::now();
-        let first_chunk_time = Arc::new(Mutex::new(None));
+        let first_chunk_time = Arc::new(Mutex::new(None::<f64>));
+        let decoder = Arc::new(Mutex::new(StreamDecoder::new()));

-        let stream = {
-            // 创建新的 stream
-            let mut stream = response.bytes_stream();
+        // 定义消息处理器的上下文结构体
+        struct MessageProcessContext<'a> {
+            response_id: &'a str,
+            model: &'a str,
+            is_start: &'a AtomicBool,
+            first_chunk_time: &'a Mutex<Option<f64>>,
+            start_time: std::time::Instant,
+            state: &'a Mutex<AppState>,
+            current_id: u64,
+        }

-            if current_config.enable_stream_check() {
-                // 检查第一个 chunk
-                match stream.next().await {
-                    Some(first_chunk) => {
-                        let chunk = first_chunk.map_err(|e| {
-                            let error_message = format!("Failed to read response chunk: {}", e);
-                            // 理论上，若程序正常，必定成功，因为前面判断过了
-                            (
-                                StatusCode::INTERNAL_SERVER_ERROR,
-                                Json(ChatError::RequestFailed(error_message).to_json()),
-                            )
-                        })?;
-
-                        match parse_stream_data(&chunk) {
-                            Err(StreamError::ChatError(error)) => {
-                                let error_respone = error.to_error_response();
-                                // 更新请求日志为失败
-                                {
-                                    let mut state = state.lock().await;
-                                    if let Some(log) = state
-                                        .request_logs
-                                        .iter_mut()
-                                        .rev()
-                                        .find(|log| log.id == current_id)
-                                    {
-                                        log.status = STATUS_FAILED;
-                                        log.error = Some(error_respone.native_code());
-                                        log.timing.total =
-                                            format_time_ms(start_time.elapsed().as_secs_f64());
-                                        state.error_requests += 1;
-                                    }
-                                }
-                                return Err((
-                                    error_respone.status_code(),
-                                    Json(error_respone.to_common()),
-                                ));
-                            }
-                            Ok(_) | Err(_) => {
-                                // 创建一个包含第一个 chunk 的 stream
-                                Box::pin(
-                                    futures::stream::once(async move { Ok(chunk) }).chain(stream),
-                                )
-                                    as Pin<
-                                        Box<
-                                            dyn Stream<Item = Result<Bytes, reqwest::Error>> + Send,
-                                        >,
-                                    >
-                            }
-                        }
-                    }
-                    None => {
-                        // Box::pin(stream)
-                        //     as Pin<Box<dyn Stream<Item = Result<Bytes, reqwest::Error>> + Send>>
-                        // 更新请求日志为失败
-                        {
-                            let mut state = state.lock().await;
-                            if let Some(log) = state
-                                .request_logs
-                                .iter_mut()
-                                .rev()
-                                .find(|log| log.id == current_id)
-                            {
-                                log.status = STATUS_FAILED;
-                                log.error = Some("Empty stream response".to_string());
-                                state.error_requests += 1;
-                            }
-                        }
-                        return Err((
-                            StatusCode::INTERNAL_SERVER_ERROR,
-                            Json(
-                                ChatError::RequestFailed("Empty stream response".to_string())
-                                    .to_json(),
-                            ),
-                        ));
-                    }
-                }
-            } else {
-                Box::pin(stream)
-                    as Pin<Box<dyn Stream<Item = Result<Bytes, reqwest::Error>> + Send>>
-            }
-        }
-        .then({
-            let buffer = Arc::new(Mutex::new(Vec::new()));
-            let first_chunk_time = first_chunk_time.clone();
-            let state = state.clone();
-
-            move |chunk| {
-                let buffer = buffer.clone();
-                let response_id = response_id.clone();
-                let model = request.model.clone();
-                let is_start = is_start.clone();
-                let full_text = full_text.clone();
-                let first_chunk_time = first_chunk_time.clone();
-                let state = state.clone();
-                // 根据配置决定是否发送最后的 finish_reason
-                let include_finish_reason = current_config.include_stop_stream();
-
-                async move {
-                    let chunk = chunk.unwrap_or_default();
-                    let mut buffer_guard = buffer.lock().await;
-                    buffer_guard.extend_from_slice(&chunk);
-
-                    match parse_stream_data(&buffer_guard) {
-                        Ok(StreamMessage::Content(texts)) => {
-                            buffer_guard.clear();
+        // 处理消息并生成响应数据的辅助函数
+        async fn process_messages(
+            messages: Vec<StreamMessage>,
+            ctx: &MessageProcessContext<'_>,
+        ) -> String {
            let mut response_data = String::new();

+            for message in messages {
+                match message {
+                    StreamMessage::Content(text) => {
                        // 记录首字时间(如果还未记录)
-                            if let Ok(mut first_time) = first_chunk_time.try_lock() {
+                        if let Ok(mut first_time) = ctx.first_chunk_time.try_lock() {
                            if first_time.is_none() {
-                                    *first_time =
-                                        Some(format_time_ms(start_time.elapsed().as_secs_f64()));
+                                *first_time = Some(ctx.start_time.elapsed().as_secs_f64());
                            }
                        }

-                            // 处理文本内容
-                            for text in texts {
-                                let mut text_guard = full_text.lock().await;
-                                text_guard.push_str(&text);
-                                let is_first = is_start.load(Ordering::SeqCst);
+                        let is_first = ctx.is_start.load(Ordering::SeqCst);

                        let response = ChatResponse {
-                                    id: response_id.clone(),
+                            id: ctx.response_id.to_string(),
                            object: OBJECT_CHAT_COMPLETION_CHUNK.to_string(),
                            created: chrono::Utc::now().timestamp(),
-                                    model: if is_first { Some(model.clone()) } else { None },
+                            model: if is_first {
+                                Some(ctx.model.to_string())
+                            } else {
+                                None
+                            },
                            choices: vec![Choice {
                                index: 0,
                                message: None,
                                delta: Some(Delta {
                                    role: if is_first {
-                                                is_start.store(false, Ordering::SeqCst);
+                                        ctx.is_start.store(false, Ordering::SeqCst);
                                        Some(Role::Assistant)
                                    } else {
                                        None
@@ -518,60 +450,26 @@ pub async fn handle_chat(
                            serde_json::to_string(&response).unwrap()
                        ));
                    }
-
-                            Ok::<_, Infallible>(Bytes::from(response_data))
-                        }
-                        Ok(StreamMessage::StreamStart) => {
-                            buffer_guard.clear();
-                            // 发送初始响应，包含模型信息
-                            let response = ChatResponse {
-                                id: response_id.clone(),
-                                object: OBJECT_CHAT_COMPLETION_CHUNK.to_string(),
-                                created: chrono::Utc::now().timestamp(),
-                                model: {
-                                    is_start.store(true, Ordering::SeqCst);
-                                    Some(model.clone())
-                                },
-                                choices: vec![Choice {
-                                    index: 0,
-                                    message: None,
-                                    delta: Some(Delta {
-                                        role: Some(Role::Assistant),
-                                        content: Some(String::new()),
-                                    }),
-                                    finish_reason: None,
-                                }],
-                                usage: None,
-                            };
-
-                            Ok(Bytes::from(format!(
-                                "data: {}\n\n",
-                                serde_json::to_string(&response).unwrap()
-                            )))
-                        }
-                        Ok(StreamMessage::StreamEnd) => {
-                            buffer_guard.clear();
-
+                    StreamMessage::StreamEnd => {
                        // 计算总时间和首次片段时间
-                            let total_time = format_time_ms(start_time.elapsed().as_secs_f64());
-                            let first_time = first_chunk_time.lock().await.unwrap_or(total_time);
+                        let total_time = ctx.start_time.elapsed().as_secs_f64();
+                        let first_time = ctx.first_chunk_time.lock().await.unwrap_or(total_time);

                        {
-                                let mut state = state.lock().await;
+                            let mut state = ctx.state.lock().await;
                            if let Some(log) = state
                                .request_logs
                                .iter_mut()
                                .rev()
-                                    .find(|log| log.id == current_id)
+                                .find(|log| log.id == ctx.current_id)
                            {
-                                    log.timing.total = total_time;
-                                    log.timing.first = Some(first_time);
+                                log.timing.total = format_time_ms(total_time);
+                                log.timing.first = Some(format_time_ms(first_time));
                            }
                        }

-                            if include_finish_reason {
                        let response = ChatResponse {
-                                    id: response_id.clone(),
+                            id: ctx.response_id.to_string(),
                            object: OBJECT_CHAT_COMPLETION_CHUNK.to_string(),
                            created: chrono::Utc::now().timestamp(),
                            model: None,
@@ -586,36 +484,150 @@ pub async fn handle_chat(
                            }],
                            usage: None,
                        };
-                                Ok(Bytes::from(format!(
+                        response_data.push_str(&format!(
                            "data: {}\n\ndata: [DONE]\n\n",
                            serde_json::to_string(&response).unwrap()
-                                )))
-                            } else {
-                                Ok(Bytes::from("data: [DONE]\n\n"))
+                        ));
                    }
-                        }
-                        Ok(StreamMessage::Incomplete) => {
-                            // 保持buffer中的数据以待下一个chunk
-                            Ok(Bytes::new())
-                        }
-                        Ok(StreamMessage::Debug(debug_prompt)) => {
-                            buffer_guard.clear();
-                            if let Ok(mut state) = state.try_lock() {
+                    StreamMessage::Debug(debug_prompt) => {
+                        if let Ok(mut state) = ctx.state.try_lock() {
                            if let Some(last_log) = state.request_logs.last_mut() {
-                                    last_log.prompt = Some(debug_prompt.clone());
+                                last_log.prompt = Some(debug_prompt);
                            }
                        }
-                            Ok(Bytes::new())
                    }
+                    _ => {} // 忽略其他消息类型
+                }
+            }
+
+            response_data
+        }
+
+        let stream = {
+            let mut stream = response.bytes_stream();
+
+            // 处理第一个chunk并获取first_result
+            while decoder.lock().await.has_no_first_result() {
+                match stream.next().await {
+                    Some(first_chunk) => {
+                        let chunk = first_chunk.map_err(|e| {
+                            let error_message = format!("Failed to read response chunk: {}", e);
+                            (
+                                StatusCode::INTERNAL_SERVER_ERROR,
+                                Json(ChatError::RequestFailed(error_message).to_json()),
+                            )
+                        })?;
+
+                        if let Err(StreamError::ChatError(error)) =
+                            decoder.lock().await.decode(&chunk, convert_web_ref)
+                        {
+                            let error_response = error.to_error_response();
+                            // 更新请求日志为失败
+                            {
+                                let mut state = state.lock().await;
+                                if let Some(log) = state
+                                    .request_logs
+                                    .iter_mut()
+                                    .rev()
+                                    .find(|log| log.id == current_id)
+                                {
+                                    log.status = LogStatus::Failed;
+                                    log.error = Some(error_response.native_code());
+                                    log.timing.total =
+                                        format_time_ms(start_time.elapsed().as_secs_f64());
+                                    state.error_requests += 1;
+                                }
+                            }
+                            return Err((
+                                error_response.status_code(),
+                                Json(error_response.to_common()),
+                            ));
+                        }
+                    }
+                    None => {
+                        // 更新请求日志为失败
+                        {
+                            let mut state = state.lock().await;
+                            if let Some(log) = state
+                                .request_logs
+                                .iter_mut()
+                                .rev()
+                                .find(|log| log.id == current_id)
+                            {
+                                log.status = LogStatus::Failed;
+                                log.error = Some("Empty stream response".to_string());
+                                state.error_requests += 1;
+                            }
+                        }
+                        return Err((
+                            StatusCode::INTERNAL_SERVER_ERROR,
+                            Json(
+                                ChatError::RequestFailed("Empty stream response".to_string())
+                                    .to_json(),
+                            ),
+                        ));
+                    }
+                }
+            }
+
+            // 处理后续的stream
+            stream.then({
+                let decoder = decoder.clone();
+                let response_id = response_id.clone();
+                let model = request.model.clone();
+                let is_start = is_start.clone();
+                let first_chunk_time = first_chunk_time.clone();
+                let state = state.clone();
+
+                move |chunk| {
+                    let decoder = decoder.clone();
+                    let response_id = response_id.clone();
+                    let model = model.clone();
+                    let is_start = is_start.clone();
+                    let first_chunk_time = first_chunk_time.clone();
+                    let state = state.clone();
+
+                    async move {
+                        let chunk = chunk.unwrap_or_default();
+
+                        let ctx = MessageProcessContext {
+                            response_id: &response_id,
+                            model: &model,
+                            is_start: &is_start,
+                            first_chunk_time: &first_chunk_time,
+                            start_time,
+                            state: &state,
+                            current_id,
+                        };
+
+                        // 使用decoder处理chunk
+                        let messages = match decoder.lock().await.decode(&chunk, convert_web_ref) {
+                            Ok(msgs) => msgs,
                            Err(e) => {
-                            buffer_guard.clear();
                                eprintln!("[警告] Stream error: {}", e);
-                            Ok(Bytes::new())
+                                return Ok::<_, Infallible>(Bytes::new());
+                            }
+                        };
+
+                        let mut response_data = String::new();
+
+                        if let Some(first_msg) = decoder.lock().await.take_first_result() {
+                            let first_response = process_messages(first_msg, &ctx).await;
+                            if !first_response.is_empty() {
+                                response_data.push_str(&first_response);
                            }
                        }
+
+                        let current_response = process_messages(messages, &ctx).await;
+                        if !current_response.is_empty() {
+                            response_data.push_str(&current_response);
+                        }
+
+                        Ok(Bytes::from(response_data))
                    }
                }
-        });
+            })
+        };

        Ok(Response::builder()
            .header("Cache-Control", "no-cache")
@@ -626,81 +638,62 @@ pub async fn handle_chat(
    } else {
        // 非流式响应
        let start_time = std::time::Instant::now();
-        let mut first_chunk_received = false;
-        let mut first_chunk_time = 0.0;
+        let mut first_chunk_time = None::<f64>;
+        let mut decoder = StreamDecoder::new();
        let mut full_text = String::with_capacity(1024);
        let mut stream = response.bytes_stream();
-        let mut prompt = None;
+        let mut all_chunks = Vec::new();

-        let mut buffer = Vec::new();
+        // 收集所有的chunks
        while let Some(chunk) = stream.next().await {
            let chunk = chunk.map_err(|e| {
-                // 更新请求日志为失败
-                if let Ok(mut state) = state.try_lock() {
-                    if let Some(log) = state
-                        .request_logs
-                        .iter_mut()
-                        .rev()
-                        .find(|log| log.id == current_id)
-                    {
-                        log.status = STATUS_FAILED;
-                        log.error = Some(format!("Failed to read response chunk: {}", e));
-                        state.error_requests += 1;
-                    }
-                }
+                let error_message = format!("Failed to read response chunk: {}", e);
                (
                    StatusCode::INTERNAL_SERVER_ERROR,
-                    Json(
-                        ChatError::RequestFailed(format!("Failed to read response chunk: {}", e))
-                            .to_json(),
-                    ),
+                    Json(ChatError::RequestFailed(error_message).to_json()),
                )
            })?;
-
-            buffer.extend_from_slice(&chunk);
-
-            match parse_stream_data(&buffer) {
-                Ok(StreamMessage::Content(texts)) => {
-                    if !first_chunk_received {
-                        first_chunk_time = format_time_ms(start_time.elapsed().as_secs_f64());
-                        first_chunk_received = true;
+            all_chunks.extend(chunk);
+        }
+
+        // 一次性解码所有数据
+        let messages = match decoder.decode(&all_chunks, convert_web_ref) {
+            Ok(msgs) => msgs,
+            Err(StreamError::ChatError(error)) => {
+                let error_response = error.to_error_response();
+                return Err((
+                    error_response.status_code(),
+                    Json(error_response.to_common()),
+                ));
+            }
+            Err(e) => {
+                let error_response = ErrorResponse {
+                    status: ApiStatus::Error,
+                    code: Some(500),
+                    error: Some(e.to_string()),
+                    message: None,
+                };
+                return Err((StatusCode::INTERNAL_SERVER_ERROR, Json(error_response)));
+            }
+        };
+
+        // 处理所有消息
+        for message in messages {
+            match message {
+                StreamMessage::Content(text) => {
+                    if first_chunk_time.is_none() {
+                        first_chunk_time = Some(start_time.elapsed().as_secs_f64());
                    }
-                    for text in texts {
                    full_text.push_str(&text);
                }
-                    buffer.clear();
-                }
-                Ok(StreamMessage::Incomplete) => continue,
-                Ok(StreamMessage::Debug(debug_prompt)) => {
-                    prompt = Some(debug_prompt);
-                    buffer.clear();
-                }
-                Ok(StreamMessage::StreamStart) | Ok(StreamMessage::StreamEnd) => {
-                    buffer.clear();
-                }
-                Err(StreamError::ChatError(error)) => {
-                    let error = error.to_error_response();
-                    // 更新请求日志为失败
-                    {
-                        let mut state = state.lock().await;
-                        if let Some(log) = state
-                            .request_logs
-                            .iter_mut()
-                            .rev()
-                            .find(|log| log.id == current_id)
-                        {
-                            log.status = STATUS_FAILED;
-                            log.error = Some(error.native_code());
-                            log.timing.total = format_time_ms(start_time.elapsed().as_secs_f64());
-                            state.error_requests += 1;
+                StreamMessage::Debug(debug_prompt) => {
+                    if let Ok(mut state) = state.try_lock() {
+                        if let Some(last_log) = state.request_logs.last_mut() {
+                            last_log.prompt = Some(debug_prompt);
                        }
                    }
-                    return Err((error.status_code(), Json(error.to_common())));
-                }
-                Err(_) => {
-                    buffer.clear();
-                    continue;
                }
+                _ => {}
            }
        }

@@ -715,11 +708,8 @@ pub async fn handle_chat(
                    .rev()
                    .find(|log| log.id == current_id)
                {
-                    log.status = STATUS_FAILED;
+                    log.status = LogStatus::Failed;
                    log.error = Some("Empty response received".to_string());
-                    if let Some(p) = prompt {
-                        log.prompt = Some(p);
-                    }
                    state.error_requests += 1;
                }
            }
@@ -761,9 +751,8 @@ pub async fn handle_chat(
                .find(|log| log.id == current_id)
            {
                log.timing.total = total_time;
-                log.timing.first = Some(first_chunk_time);
-                log.prompt = prompt;
-                log.status = STATUS_SUCCESS;
+                log.timing.first = first_chunk_time;
+                log.status = LogStatus::Success;
            }
        }

--- a/src/chat/stream.rs
+++ b/src/chat/stream.rs
@@ -1,230 +1,2 @@
-use super::aiserver::v1::StreamChatResponse;
-use flate2::read::GzDecoder;
-use prost::Message;
-use std::io::Read;
-
-use super::error::{ChatError, StreamError};
-
-// 解压gzip数据
-fn decompress_gzip(data: &[u8]) -> Option<Vec<u8>> {
-    let mut decoder = GzDecoder::new(data);
-    let mut decompressed = Vec::new();
-
-    match decoder.read_to_end(&mut decompressed) {
-        Ok(_) => Some(decompressed),
-        Err(_) => {
-            // println!("gzip解压失败: {}", e);
-            None
-        }
-    }
-}
-
-pub enum StreamMessage {
-    // 未完成
-    Incomplete,
-    // 调试
-    Debug(String),
-    // 流开始标志 b"\0\0\0\0\0"
-    StreamStart,
-    // 消息内容
-    Content(Vec<String>),
-    // 流结束标志 b"\x02\0\0\0\x02{}"
-    StreamEnd,
-}
-
-pub fn parse_stream_data(data: &[u8]) -> Result<StreamMessage, StreamError> {
-    if data.len() < 5 {
-        return Err(StreamError::DataLengthLessThan5);
-    }
-
-    // 检查是否为流开始标志
-    // if data == b"\0\0\0\0\0" {
-    //     return Ok(StreamMessage::StreamStart);
-    // }
-
-    // 检查是否为流结束标志
-    // if data == b"\x02\0\0\0\x02{}" {
-    //     return Ok(StreamMessage::StreamEnd);
-    // }
-
-    let mut messages = Vec::new();
-    let mut offset = 0;
-
-    while offset + 5 <= data.len() {
-        // 获取消息类型和长度
-        let msg_type = data[offset];
-        let msg_len = u32::from_be_bytes([
-            data[offset + 1],
-            data[offset + 2],
-            data[offset + 3],
-            data[offset + 4],
-        ]) as usize;
-
-        // 流开始
-        if msg_type == 0 && msg_len == 0 {
-            return Ok(StreamMessage::StreamStart);
-        }
-
-        // 检查剩余数据长度是否足够
-        if offset + 5 + msg_len > data.len() {
-            return Ok(StreamMessage::Incomplete);
-        }
-
-        let msg_data = &data[offset + 5..offset + 5 + msg_len];
-
-        match msg_type {
-            // 文本消息
-            0 => {
-                if let Ok(response) = StreamChatResponse::decode(msg_data) {
-                    // crate::debug_println!("[text] StreamChatResponse: {:?}", response);
-                    if !response.text.is_empty() {
-                        messages.push(response.text);
-                    } else {
-                        // println!("[text] StreamChatResponse: {:?}", response);
-                        return Ok(StreamMessage::Debug(
-                            response.filled_prompt.unwrap_or_default(),
-                            // response.is_using_slow_request,
-                        ));
-                    }
-                }
-            }
-            // gzip压缩消息
-            1 => {
-                if let Some(text) = decompress_gzip(msg_data) {
-                    if let Ok(response) = StreamChatResponse::decode(&text[..]) {
-                        // crate::debug_println!("[gzip] StreamChatResponse: {:?}", response);
-                        if !response.text.is_empty() {
-                            messages.push(response.text);
-                        } else {
-                            // println!("[gzip] StreamChatResponse: {:?}", response);
-                            return Ok(StreamMessage::Debug(
-                                response.filled_prompt.unwrap_or_default(),
-                                // response.is_using_slow_request,
-                            ));
-                        }
-                    }
-                }
-            }
-            // JSON字符串
-            2 => {
-                if msg_len == 2 {
-                    return Ok(StreamMessage::StreamEnd);
-                }
-                if let Ok(text) = String::from_utf8(msg_data.to_vec()) {
-                    // println!("JSON消息: {}", text);
-                    if let Ok(error) = serde_json::from_str::<ChatError>(&text) {
-                        return Err(StreamError::ChatError(error));
-                    }
-                    // 未预计
-                    // messages.push(text);
-                }
-            }
-            // gzip压缩消息
-            3 => {
-                if let Some(text) = decompress_gzip(msg_data) {
-                    if text.len() == 2 {
-                        return Ok(StreamMessage::StreamEnd);
-                    }
-                    if let Ok(text) = String::from_utf8(text) {
-                        // println!("JSON消息: {}", text);
-                        if let Ok(error) = serde_json::from_str::<ChatError>(&text) {
-                            return Err(StreamError::ChatError(error));
-                        }
-                        // 未预计
-                        // messages.push(text);
-                    }
-                }
-            }
-            // 其他类型暂不处理
-            t => {
-                eprintln!("收到未知消息类型: {}，请尝试联系开发者以获取支持", t);
-                crate::debug_println!("消息类型: {}，消息内容: {}", t, hex::encode(msg_data));
-            }
-        }
-
-        offset += 5 + msg_len;
-    }
-
-    if messages.is_empty() {
-        Err(StreamError::EmptyMessage)
-    } else {
-        Ok(StreamMessage::Content(messages))
-    }
-}
-
-#[test]
-fn test_parse_stream_data() {
-    // 使用include_str!加载测试数据文件
-    let stream_data = include_str!("../../tests/data/stream_data.txt");
-
-    // 将整个字符串按每两个字符分割成字节
-    let bytes: Vec<u8> = stream_data
-        .as_bytes()
-        .chunks(2)
-        .map(|chunk| {
-            let hex_str = std::str::from_utf8(chunk).unwrap();
-            u8::from_str_radix(hex_str, 16).unwrap()
-        })
-        .collect();
-
-    // 辅助函数：找到下一个消息边界
-    fn find_next_message_boundary(bytes: &[u8]) -> usize {
-        if bytes.len() < 5 {
-            return bytes.len();
-        }
-        let msg_len = u32::from_be_bytes([bytes[1], bytes[2], bytes[3], bytes[4]]) as usize;
-        5 + msg_len
-    }
-
-    // 辅助函数：将字节转换为hex字符串
-    fn bytes_to_hex(bytes: &[u8]) -> String {
-        bytes
-            .iter()
-            .map(|b| format!("{:02X}", b))
-            .collect::<Vec<String>>()
-            .join("")
-    }
-
-    // 多次解析数据
-    let mut offset = 0;
-    while offset < bytes.len() {
-        let remaining_bytes = &bytes[offset..];
-        let msg_boundary = find_next_message_boundary(remaining_bytes);
-        let current_msg_bytes = &remaining_bytes[..msg_boundary];
-        let hex_str = bytes_to_hex(current_msg_bytes);
-
-        match parse_stream_data(current_msg_bytes) {
-            Ok(message) => {
-                match message {
-                    StreamMessage::Content(messages) => {
-                        print!("消息内容 [hex: {}]:", hex_str);
-                        for msg in messages {
-                            println!(" {}", msg);
-                        }
-                        offset += msg_boundary;
-                    }
-                    StreamMessage::Debug(_) => {
-                        // println!("调试信息 [hex: {}]: {}", hex_str, prompt);
-                        offset += msg_boundary;
-                    }
-                    StreamMessage::StreamEnd => {
-                        println!("流结束 [hex: {}]", hex_str);
-                        break;
-                    }
-                    StreamMessage::StreamStart => {
-                        println!("流开始 [hex: {}]", hex_str);
-                        offset += msg_boundary;
-                    }
-                    StreamMessage::Incomplete => {
-                        println!("数据不完整 [hex: {}]", hex_str);
-                        break;
-                    }
-                }
-            }
-            Err(e) => {
-                println!("解析错误 [hex: {}]: {}", hex_str, e);
-                break;
-            }
-        }
-    }
-}
+mod decoder;
+pub use decoder::*;
--- a/src/chat/stream/decoder.rs
+++ b/src/chat/stream/decoder.rs
@@ -0,0 +1,398 @@
+use crate::chat::{
+    aiserver::v1::StreamChatResponse,
+    error::{ChatError, StreamError},
+};
+use flate2::read::GzDecoder;
+use prost::Message;
+use std::{collections::BTreeMap, io::Read};
+
+// 解压gzip数据
+fn decompress_gzip(data: &[u8]) -> Option<Vec<u8>> {
+    let mut decoder = GzDecoder::new(data);
+    let mut decompressed = Vec::new();
+
+    match decoder.read_to_end(&mut decompressed) {
+        Ok(_) => Some(decompressed),
+        Err(_) => {
+            // println!("gzip解压失败: {}", e);
+            None
+        }
+    }
+}
+
+pub trait ToMarkdown {
+    fn to_markdown(&self) -> String;
+}
+
+impl ToMarkdown for BTreeMap<String, String> {
+    fn to_markdown(&self) -> String {
+        if self.is_empty() {
+            return String::new();
+        }
+
+        let mut result = String::from("WebReferences:\n");
+        for (i, (url, title)) in self.iter().enumerate() {
+            result.push_str(&format!("{}. [{}]({})\n", i + 1, title, url));
+        }
+        result.push_str("\n");
+        result
+    }
+}
+
+#[derive(PartialEq, Clone)]
+pub enum StreamMessage {
+    // 调试
+    Debug(String),
+    // 网络引用
+    WebReference(BTreeMap<String, String>),
+    // 内容开始标志
+    ContentStart,
+    // 消息内容
+    Content(String),
+    // 流结束标志
+    StreamEnd,
+}
+
+impl StreamMessage {
+    fn convert_web_ref_to_content(self) -> Self {
+        match self {
+            StreamMessage::WebReference(refs) => StreamMessage::Content(refs.to_markdown()),
+            other => other,
+        }
+    }
+}
+
+pub struct StreamDecoder {
+    buffer: Vec<u8>,
+    first_result: Option<Vec<StreamMessage>>,
+    first_result_taken: bool,
+}
+
+impl StreamDecoder {
+    pub fn new() -> Self {
+        Self {
+            buffer: Vec::new(),
+            first_result: None,
+            first_result_taken: false,
+        }
+    }
+
+    // 获取第一个结果的引用
+    pub fn take_first_result(&mut self) -> Option<Vec<StreamMessage>> {
+        if self.is_incomplete() {
+            return None;
+        }
+        if self.first_result.is_some() {
+            self.first_result_taken = true;
+        }
+        self.first_result.take()
+    }
+
+    fn is_incomplete(&self) -> bool {
+        !self.buffer.is_empty()
+    }
+
+    pub fn has_no_first_result(&self) -> bool {
+        self.first_result.is_none()
+    }
+
+    pub fn decode(&mut self, data: &[u8], convert_web_ref: bool) -> Result<Vec<StreamMessage>, StreamError> {
+        self.buffer.extend_from_slice(data);
+
+        if self.buffer.len() < 5 {
+            crate::debug_println!("数据长度小于5字节，当前数据: {}", hex::encode(&self.buffer));
+            return Err(StreamError::DataLengthLessThan5);
+        }
+
+        let mut messages = Vec::new();
+        let mut offset = 0;
+
+        while offset + 5 <= self.buffer.len() {
+            let msg_type = self.buffer[offset];
+            let msg_len = u32::from_be_bytes([
+                self.buffer[offset + 1],
+                self.buffer[offset + 2],
+                self.buffer[offset + 3],
+                self.buffer[offset + 4],
+            ]) as usize;
+
+            if msg_len == 0 {
+                offset += 5;
+                messages.push(StreamMessage::ContentStart);
+                continue;
+            }
+
+            if offset + 5 + msg_len > self.buffer.len() {
+                break;
+            }
+
+            let msg_data = &self.buffer[offset + 5..offset + 5 + msg_len];
+
+            match self.process_message(msg_type, msg_data)? {
+                Some(msg) => {
+                    if convert_web_ref {
+                        messages.push(msg.convert_web_ref_to_content());
+                    } else {
+                        messages.push(msg);
+                    }
+                }
+                _ => {}
+            }
+
+            offset += 5 + msg_len;
+        }
+
+        self.buffer.drain(..offset);
+
+        if !self.first_result_taken && !messages.is_empty() {
+            if self.first_result.is_none() {
+                self.first_result = Some(messages.clone());
+            } else {
+                self.first_result.as_mut().unwrap().extend(messages.clone());
+            }
+        }
+        Ok(messages)
+    }
+
+    fn process_message(
+        &self,
+        msg_type: u8,
+        msg_data: &[u8],
+    ) -> Result<Option<StreamMessage>, StreamError> {
+        match msg_type {
+            0 => self.handle_text_message(msg_data),
+            1 => self.handle_gzip_message(msg_data),
+            2 => self.handle_json_message(msg_data),
+            3 => self.handle_gzip_json_message(msg_data),
+            t => {
+                eprintln!("收到未知消息类型: {}，请尝试联系开发者以获取支持", t);
+                crate::debug_println!("消息类型: {}，消息内容: {}", t, hex::encode(msg_data));
+                Ok(None)
+            }
+        }
+    }
+
+    fn handle_text_message(&self, msg_data: &[u8]) -> Result<Option<StreamMessage>, StreamError> {
+        if let Ok(response) = StreamChatResponse::decode(msg_data) {
+            // crate::debug_println!("[text] StreamChatResponse [hex: {}]: {:?}", hex::encode(msg_data), response);
+            if !response.text.is_empty() {
+                Ok(Some(StreamMessage::Content(response.text)))
+            } else if let Some(filled_prompt) = response.filled_prompt {
+                Ok(Some(StreamMessage::Debug(filled_prompt)))
+            } else if let Some(web_citation) = response.web_citation {
+                let mut refs = BTreeMap::new();
+                for reference in web_citation.references {
+                    refs.insert(reference.url, reference.title);
+                }
+                Ok(Some(StreamMessage::WebReference(refs)))
+            } else {
+                Ok(None)
+            }
+        } else {
+            Ok(None)
+        }
+    }
+
+    fn handle_gzip_message(&self, msg_data: &[u8]) -> Result<Option<StreamMessage>, StreamError> {
+        if let Some(text) = decompress_gzip(msg_data) {
+            if let Ok(response) = StreamChatResponse::decode(&text[..]) {
+                // crate::debug_println!("[gzip] StreamChatResponse [hex: {}]: {:?}", hex::encode(msg_data), response);
+                if !response.text.is_empty() {
+                    Ok(Some(StreamMessage::Content(response.text)))
+                } else if let Some(filled_prompt) = response.filled_prompt {
+                    Ok(Some(StreamMessage::Debug(filled_prompt)))
+                } else if let Some(web_citation) = response.web_citation {
+                    let mut refs = BTreeMap::new();
+                    for reference in web_citation.references {
+                        refs.insert(reference.url, reference.title);
+                    }
+                    Ok(Some(StreamMessage::WebReference(refs)))
+                } else {
+                    Ok(None)
+                }
+            } else {
+                Ok(None)
+            }
+        } else {
+            Ok(None)
+        }
+    }
+
+    fn handle_json_message(&self, msg_data: &[u8]) -> Result<Option<StreamMessage>, StreamError> {
+        if msg_data.len() == 2 {
+            return Ok(Some(StreamMessage::StreamEnd));
+        }
+        if let Ok(text) = String::from_utf8(msg_data.to_vec()) {
+            // println!("JSON消息: {}", text);
+            if let Ok(error) = serde_json::from_str::<ChatError>(&text) {
+                return Err(StreamError::ChatError(error));
+            }
+        }
+        Ok(None)
+    }
+
+    fn handle_gzip_json_message(
+        &self,
+        msg_data: &[u8],
+    ) -> Result<Option<StreamMessage>, StreamError> {
+        if let Some(text) = decompress_gzip(msg_data) {
+            if text.len() == 2 {
+                return Ok(Some(StreamMessage::StreamEnd));
+            }
+            if let Ok(text) = String::from_utf8(text) {
+                // println!("JSON消息: {}", text);
+                if let Ok(error) = serde_json::from_str::<ChatError>(&text) {
+                    return Err(StreamError::ChatError(error));
+                }
+            }
+        }
+        Ok(None)
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_single_chunk() {
+        // 使用include_str!加载测试数据文件
+        let stream_data = include_str!("../../../tests/data/stream_data.txt");
+
+        // 将整个字符串按每两个字符分割成字节
+        let bytes: Vec<u8> = stream_data
+            .as_bytes()
+            .chunks(2)
+            .map(|chunk| {
+                let hex_str = std::str::from_utf8(chunk).unwrap();
+                u8::from_str_radix(hex_str, 16).unwrap()
+            })
+            .collect();
+
+        // 创建解码器
+        let mut decoder = StreamDecoder::new();
+
+        match decoder.decode(&bytes, false) {
+            Ok(messages) => {
+                for message in messages {
+                    match message {
+                        StreamMessage::StreamEnd => {
+                            println!("流结束");
+                            break;
+                        }
+                        StreamMessage::Content(msg) => {
+                            println!("消息内容: {}", msg);
+                        }
+                        StreamMessage::WebReference(refs) => {
+                            println!("网页引用:");
+                            for (i, (url, title)) in refs.iter().enumerate() {
+                                println!("{}. {} - {}", i, url, title);
+                            }
+                        }
+                        StreamMessage::Debug(prompt) => {
+                            println!("调试信息: {}", prompt);
+                        }
+                        StreamMessage::ContentStart => {
+                            println!("流开始");
+                        }
+                    }
+                }
+            }
+            Err(e) => {
+                println!("解析错误: {}", e);
+            }
+        }
+        if decoder.is_incomplete() {
+            println!("数据不完整");
+        }
+    }
+
+    #[test]
+    fn test_multiple_chunks() {
+        // 使用include_str!加载测试数据文件
+        let stream_data = include_str!("../../../tests/data/stream_data.txt");
+
+        // 将整个字符串按每两个字符分割成字节
+        let bytes: Vec<u8> = stream_data
+            .as_bytes()
+            .chunks(2)
+            .map(|chunk| {
+                let hex_str = std::str::from_utf8(chunk).unwrap();
+                u8::from_str_radix(hex_str, 16).unwrap()
+            })
+            .collect();
+
+        // 创建解码器
+        let mut decoder = StreamDecoder::new();
+
+        // 辅助函数：找到下一个消息边界
+        fn find_next_message_boundary(bytes: &[u8]) -> usize {
+            if bytes.len() < 5 {
+                return bytes.len();
+            }
+            let msg_len = u32::from_be_bytes([bytes[1], bytes[2], bytes[3], bytes[4]]) as usize;
+            5 + msg_len
+        }
+
+        // 辅助函数：将字节转换为hex字符串
+        fn bytes_to_hex(bytes: &[u8]) -> String {
+            bytes
+                .iter()
+                .map(|b| format!("{:02X}", b))
+                .collect::<Vec<String>>()
+                .join("")
+        }
+
+        // 多次解析数据
+        let mut offset = 0;
+        let mut should_break = false;
+
+        while offset < bytes.len() {
+            let remaining_bytes = &bytes[offset..];
+            let msg_boundary = find_next_message_boundary(remaining_bytes);
+            let current_msg_bytes = &remaining_bytes[..msg_boundary];
+            let hex_str = bytes_to_hex(current_msg_bytes);
+
+            match decoder.decode(current_msg_bytes, false) {
+                Ok(messages) => {
+                    for message in messages {
+                        match message {
+                            StreamMessage::StreamEnd => {
+                                println!("流结束 [hex: {}]", hex_str);
+                                should_break = true;
+                                break;
+                            }
+                            StreamMessage::Content(msg) => {
+                                println!("消息内容 [hex: {}]: {}", hex_str, msg);
+                            }
+                            StreamMessage::WebReference(refs) => {
+                                println!("网页引用 [hex: {}]:", hex_str);
+                                for (i, (url, title)) in refs.iter().enumerate() {
+                                    println!("{}. {} - {}", i, url, title);
+                                }
+                            }
+                            StreamMessage::Debug(prompt) => {
+                                println!("调试信息 [hex: {}]: {}", hex_str, prompt);
+                            }
+                            StreamMessage::ContentStart => {
+                                println!("流开始 [hex: {}]", hex_str);
+                            }
+                        }
+                    }
+                    if should_break {
+                        break;
+                    }
+                    if decoder.is_incomplete() {
+                        println!("数据不完整 [hex: {}]", hex_str);
+                        break;
+                    }
+                    offset += msg_boundary;
+                }
+                Err(e) => {
+                    println!("解析错误 [hex: {}]: {}", hex_str, e);
+                    break;
+                }
+            }
+        }
+    }
+}
--- a/src/common/client.rs
+++ b/src/common/client.rs
@@ -5,8 +5,7 @@ use crate::{app::{
        HEADER_NAME_GHOST_MODE, TRUE,
    },
    lazy::{
-        CURSOR_API2_CHAT_URL, CURSOR_API2_STRIPE_URL, CURSOR_USAGE_API_URL, CURSOR_USER_API_URL,
-        REVERSE_PROXY_HOST, USE_REVERSE_PROXY,
+        CURSOR_API2_CHAT_URL, CURSOR_API2_CHAT_WEB_URL, CURSOR_API2_STRIPE_URL, CURSOR_USAGE_API_URL, CURSOR_USER_API_URL, REVERSE_PROXY_HOST, USE_REVERSE_PROXY
    },
 }, AppConfig};
 use reqwest::header::{
@@ -67,19 +66,24 @@ pub fn rebuild_http_client() {
 /// # 返回
 ///
 /// * `reqwest::RequestBuilder` - 配置好的请求构建器
-pub fn build_client(auth_token: &str, checksum: &str) -> RequestBuilder {
+pub fn build_client(auth_token: &str, checksum: &str, is_search: bool) -> RequestBuilder {
    let trace_id = Uuid::new_v4().to_string();
+    let url = if is_search {
+        &*CURSOR_API2_CHAT_WEB_URL
+    } else {
+        &*CURSOR_API2_CHAT_URL
+    };

    let client = if *USE_REVERSE_PROXY {
        HTTP_CLIENT
            .read()
-            .post(&*CURSOR_API2_CHAT_URL)
+            .post(url)
            .header(HOST, &*REVERSE_PROXY_HOST)
            .header(PROXY_HOST, CURSOR_API2_HOST)
    } else {
        HTTP_CLIENT
            .read()
-            .post(&*CURSOR_API2_CHAT_URL)
+            .post(url)
            .header(HOST, CURSOR_API2_HOST)
    };

--- a/src/common/model/config.rs
+++ b/src/common/model/config.rs
@@ -5,8 +5,6 @@ use crate::app::model::{PageContent, UsageCheck, VisionAbility, Proxies};
 #[derive(Serialize)]
 pub struct ConfigData {
    pub page_content: Option<PageContent>,
-    pub enable_stream_check: bool,
-    pub include_stop_stream: bool,
    pub vision_ability: VisionAbility,
    pub enable_slow_pool: bool,
    pub enable_all_claude: bool,
@@ -15,6 +13,7 @@ pub struct ConfigData {
    #[serde(skip_serializing_if = "String::is_empty")]
    pub share_token: String,
    pub proxies: Proxies,
+    pub include_web_references: bool,
 }

 #[derive(Deserialize, Default)]
@@ -23,8 +22,6 @@ pub struct ConfigUpdateRequest {
    pub action: String, // "get", "update", "reset"
    pub path: String,
    pub content: Option<PageContent>, // "default", "text", "html"
-    pub enable_stream_check: Option<bool>,
-    pub include_stop_stream: Option<bool>,
    pub vision_ability: Option<VisionAbility>,
    pub enable_slow_pool: Option<bool>,
    pub enable_all_claude: Option<bool>,
@@ -32,4 +29,5 @@ pub struct ConfigUpdateRequest {
    pub enable_dynamic_key: Option<bool>,
    pub share_token: Option<String>,
    pub proxies: Option<Proxies>,
+    pub include_web_references: Option<bool>,
 }
--- a/src/common/model/userinfo.rs
+++ b/src/common/model/userinfo.rs
@@ -1,5 +1,6 @@
 use chrono::{DateTime, Local};
 use serde::{Deserialize, Serialize};
+use rkyv::{Archive, Deserialize as RkyvDeserialize, Serialize as RkyvSerialize};

 #[derive(Serialize)]
 #[serde(untagged)]
@@ -8,14 +9,14 @@ pub enum GetUserInfo {
    Error { error: String },
 }

-#[derive(Serialize, Clone)]
+#[derive(Serialize, Clone, Archive, RkyvDeserialize, RkyvSerialize)]
 pub struct TokenProfile {
    pub usage: UsageProfile,
    pub user: UserProfile,
    pub stripe: StripeProfile,
 }

-#[derive(Deserialize, Serialize, PartialEq, Clone)]
+#[derive(Deserialize, Serialize, PartialEq, Clone, Archive, RkyvDeserialize, RkyvSerialize)]
 pub enum MembershipType {
    #[serde(rename = "free")]
    Free,
@@ -27,7 +28,7 @@ pub enum MembershipType {
    Enterprise,
 }

-#[derive(Deserialize, Serialize, Clone)]
+#[derive(Deserialize, Serialize, Clone, Archive, RkyvDeserialize, RkyvSerialize)]
 pub struct StripeProfile {
    #[serde(rename(deserialize = "membershipType"))]
    pub membership_type: MembershipType,
@@ -41,7 +42,7 @@ pub struct StripeProfile {
    pub days_remaining_on_trial: u32,
 }

-#[derive(Deserialize, Serialize, Clone)]
+#[derive(Deserialize, Serialize, Clone, Archive, RkyvDeserialize, RkyvSerialize)]
 pub struct ModelUsage {
    #[serde(rename(deserialize = "numRequests", serialize = "requests"))]
    pub num_requests: u32,
@@ -65,7 +66,7 @@ pub struct ModelUsage {
    pub max_tokens: Option<u32>,
 }

-#[derive(Deserialize, Serialize, Clone)]
+#[derive(Deserialize, Serialize, Clone, Archive, RkyvDeserialize, RkyvSerialize)]
 pub struct UsageProfile {
    #[serde(rename(deserialize = "gpt-4"))]
    pub premium: ModelUsage,
@@ -75,7 +76,7 @@ pub struct UsageProfile {
    pub unknown: ModelUsage,
 }

-#[derive(Deserialize, Serialize, Clone)]
+#[derive(Deserialize, Serialize, Clone, Archive, RkyvDeserialize, RkyvSerialize)]
 pub struct UserProfile {
    pub email: String,
    // pub email_verified: bool,
--- a/src/main.rs
+++ b/src/main.rs
@@ -32,6 +32,7 @@ use chat::{
 };
 use common::utils::{load_tokens, parse_string_from_env, parse_usize_from_env};
 use std::sync::Arc;
+use tokio::signal;
 use tokio::sync::Mutex;
 use tower_http::{cors::CorsLayer, limit::RequestBodyLimitLayer};

@@ -63,6 +64,11 @@ async fn main() {
    // 初始化应用状态
    let state = Arc::new(Mutex::new(AppState::new(token_infos)));

+    // 尝试加载保存的配置
+    if let Err(e) = AppConfig::load_saved_config() {
+        eprintln!("加载保存的配置失败: {}", e);
+    }
+
    // 创建一个克隆用于后台任务
    let state_for_reload = state.clone();

@@ -84,10 +90,55 @@ async fn main() {

            let mut app_state = state_for_reload.lock().await;
            app_state.update_checksum();
-            debug_println!("checksum 自动刷新: {}", next_reload);
+            // debug_println!("checksum 自动刷新: {}", next_reload);
        }
    });

+    // 创建一个克隆用于信号处理
+    let state_for_shutdown = state.clone();
+
+    // 设置关闭信号处理
+    let shutdown_signal = async move {
+        let ctrl_c = async {
+            signal::ctrl_c()
+                .await
+                .expect("failed to install Ctrl+C handler");
+        };
+
+        #[cfg(unix)]
+        let terminate = async {
+            signal::unix::signal(signal::unix::SignalKind::terminate())
+                .expect("failed to install signal handler")
+                .recv()
+                .await;
+        };
+
+        #[cfg(not(unix))]
+        let terminate = std::future::pending::<()>();
+
+        tokio::select! {
+            _ = ctrl_c => {},
+            _ = terminate => {},
+        }
+
+        println!("正在关闭服务器...");
+
+        // 保存配置
+        if let Err(e) = AppConfig::save_config() {
+            eprintln!("保存配置失败: {}", e);
+        } else {
+            println!("配置已保存");
+        }
+
+        // 保存日志
+        let state = state_for_shutdown.lock().await;
+        if let Err(e) = state.save_logs().await {
+            eprintln!("保存日志失败: {}", e);
+        } else {
+            println!("日志已保存");
+        }
+    };
+
    // 设置路由
    let app = Router::new()
        .route(ROUTE_ROOT_PATH, get(handle_root))
@@ -128,9 +179,19 @@ async fn main() {
    println!("服务器运行在端口 {}", port);
    println!("当前版本: v{}", PKG_VERSION);
    // if PKG_VERSION.contains("pre") {
-    println!("当前是测试版，有问题及时反馈哦~");
+    // println!("当前是测试版，有问题及时反馈哦~");
    // }

    let listener = tokio::net::TcpListener::bind(addr).await.unwrap();
-    axum::serve(listener, app).await.unwrap();
+    let server = axum::serve(listener, app);
+    tokio::select! {
+        result = server => {
+            if let Err(e) = result {
+                eprintln!("服务器错误: {}", e);
+            }
+        }
+        _ = shutdown_signal => {
+            println!("服务器已关闭");
+        }
+    }
 }
--- a/static/api.html
+++ b/static/api.html
@@ -118,7 +118,7 @@
    <div class="button-group">
      <button onclick="calibrateToken()">校准 Token</button>
      <button onclick="getUserInfo()">获取用户信息</button>
-      <button onclick="getModels()">获取模型列表</button>
+      <button onclick="getModels(true)">获取模型列表</button>
    </div>

    <div class="form-group model-input-container">
@@ -130,7 +130,7 @@
    <div class="form-group custom-suffix">
      <input type="checkbox" id="customSuffix" onchange="toggleCustomSuffix()">
      <label for="customSuffix">添加自定义后缀</label>
-      <input type="text" id="suffixInput" placeholder="@OpenAI" style="display: none;">
+      <input type="text" id="suffixInput" placeholder="-online@OpenAI" style="display: none;">
    </div>
  </div>

@@ -197,14 +197,16 @@
    }

    // 获取模型列表
-    async function getModels() {
+    async function getModels(showMessage = false) {
      try {
        const modelList = document.getElementById('modelList');
        const suffix = document.getElementById('customSuffix').checked ?
          document.getElementById('suffixInput').value : '';

        modelList.value = globalModels.map(model => model + suffix).join(',');
+        if (showMessage) {
          showGlobalMessage('模型列表已更新');
+        }
      } catch (error) {
        showGlobalMessage('获取模型列表失败', true);
      }
@@ -223,7 +225,7 @@
      const suffixInput = document.getElementById('suffixInput');
      suffixInput.style.display = document.getElementById('customSuffix').checked ? 'block' : 'none';
      if (document.getElementById('customSuffix').checked) {
-        getModels();
+        getModels(false);
      }
    }

@@ -363,8 +365,13 @@
        await startStatusCheck();
        showGlobalMessage('系统初始化完成');

-        // 监听后缀输入变化
-        document.getElementById('suffixInput').addEventListener('input', getModels);
+        // 使用防抖处理后缀输入事件
+        const suffixInput = document.getElementById('suffixInput');
+        let debounceTimer;
+        suffixInput.addEventListener('input', () => {
+          clearTimeout(debounceTimer);
+          debounceTimer = setTimeout(() => getModels(false), 300);
+        });
      } catch (error) {
        showGlobalMessage('系统初始化失败', true);
      }
--- a/static/build_key.html
+++ b/static/build_key.html
@@ -75,24 +75,6 @@
      <input type="password" id="dataToken" placeholder="输入数据认证令牌">
    </div>

-    <div class="form-group">
-      <label>流第一个块检查:</label>
-      <select id="enableStreamCheck">
-        <option value="">跟随全局</option>
-        <option value="true">启用</option>
-        <option value="false">禁用</option>
-      </select>
-    </div>
-
-    <div class="form-group">
-      <label>包含停止流:</label>
-      <select id="includeStopStream">
-        <option value="">跟随全局</option>
-        <option value="true">启用</option>
-        <option value="false">禁用</option>
-      </select>
-    </div>
-
    <div class="form-group">
      <label>图片处理能力:</label>
      <select id="disableVision">
@@ -125,6 +107,15 @@
      </div>
    </div>

+    <div class="form-group">
+      <label>包含网络引用:</label>
+      <select id="includeWebReferences">
+        <option value="">跟随全局</option>
+        <option value="true">启用</option>
+        <option value="false">禁用</option>
+      </select>
+    </div>
+
    <div class="button-group">
      <button onclick="buildKey()">构建 Key</button>
      <button onclick="clearForm()" class="secondary">清空表单</button>
@@ -187,14 +178,13 @@

      const data = {
        auth_token: dataToken,
-        enable_stream_check: parseBooleanFromString(document.getElementById('enableStreamCheck').value, undefined),
-        include_stop_stream: parseBooleanFromString(document.getElementById('includeStopStream').value, undefined),
        disable_vision: parseBooleanFromString(document.getElementById('disableVision').value, undefined),
        enable_slow_pool: parseBooleanFromString(document.getElementById('enableSlowPool').value, undefined),
        usage_check_models: type ? {
          type: type,
          model_ids: type === 'custom' ? modelIds : undefined
-        } : undefined
+        } : undefined,
+        include_web_references: parseBooleanFromString(document.getElementById('includeWebReferences').value, undefined)
      };

      try {
@@ -227,11 +217,10 @@
    function clearForm() {
      document.getElementById('authToken').value = '';
      document.getElementById('dataToken').value = '';
-      document.getElementById('enableStreamCheck').value = '';
-      document.getElementById('includeStopStream').value = '';
      document.getElementById('disableVision').value = '';
      document.getElementById('enableSlowPool').value = '';
      document.getElementById('usageCheckType').value = 'default';
+      document.getElementById('includeWebReferences').value = '';
      document.getElementById('modelListContainer').style.display = 'none';
      document.getElementById('keyResult').style.display = 'none';
      showGlobalMessage('表单已清空');
--- a/static/config.html
+++ b/static/config.html
@@ -45,24 +45,6 @@
      <textarea id="content"></textarea>
    </div>

-    <div class="form-group">
-      <label>流第一个块检查:</label>
-      <select id="enable_stream_check">
-        <option value="">保持不变</option>
-        <option value="true">启用</option>
-        <option value="false">禁用</option>
-      </select>
-    </div>
-
-    <div class="form-group">
-      <label>包含停止流:</label>
-      <select id="include_stop_stream">
-        <option value="">保持不变</option>
-        <option value="true">启用</option>
-        <option value="false">禁用</option>
-      </select>
-    </div>
-
    <div class="form-group">
      <label>图片处理能力:</label>
      <select id="vision_ability">
@@ -182,10 +164,6 @@
              visionValue = 'base64-http';
              break;
          }
-          document.getElementById('enable_stream_check').value =
-            parseStringFromBoolean(data.data.enable_stream_check, '');
-          document.getElementById('include_stop_stream').value =
-            parseStringFromBoolean(data.data.include_stop_stream, '');
          document.getElementById('vision_ability').value = visionValue;
          document.getElementById('enable_slow_pool').value =
            parseStringFromBoolean(data.data.enable_slow_pool, '');
@@ -243,18 +221,10 @@
          };
        }

-        const shareToken = document.getElementById('shareToken').value.trim();
-
        const data = {
          action,
          path: document.getElementById('path').value,
          ...(contentObj && { content: contentObj }),
-          ...(document.getElementById('enable_stream_check').value && {
-            enable_stream_check: parseBooleanFromString(document.getElementById('enable_stream_check').value)
-          }),
-          ...(document.getElementById('include_stop_stream').value && {
-            include_stop_stream: parseBooleanFromString(document.getElementById('include_stop_stream').value)
-          }),
          ...(document.getElementById('vision_ability').value && {
            vision_ability: document.getElementById('vision_ability').value
          }),
@@ -290,9 +260,7 @@
              }
            })()
          }),
-          ...(shareToken && {
-            share_token: shareToken
-          }),
+          share_token: document.getElementById('shareToken').value.trim(),
        };

        const result = await makeAuthenticatedRequest('/config', {
--- a/static/logs.html
+++ b/static/logs.html
@@ -551,21 +551,6 @@
      return 'high';
    }

-    function formatMembershipType(type) {
-      if (!type) return '-';
-
-      switch (type) {
-        case 'free_trial': return 'Pro Trial';
-        case 'pro': return 'Pro';
-        case 'free': return 'Free';
-        case 'enterprise': return 'Business';
-        default: return type
-          .split('_')
-          .map(word => word.charAt(0).toUpperCase() + word.slice(1))
-          .join(' ');
-      }
-    }
-
    function showTokenModal(tokenInfo) {
      const modal = document.getElementById('tokenModal');
      const deleteBtn = document.getElementById('deleteTokenBtn');
--- a/static/shared.js
+++ b/static/shared.js
@@ -1,10 +1,19 @@
 // Token 管理功能
+/**
+ * 保存认证令牌到本地存储
+ * @param {string} token - 要保存的认证令牌
+ * @returns {void}
+ */
 function saveAuthToken(token) {
  const expiryTime = new Date().getTime() + (24 * 60 * 60 * 1000); // 24小时后过期
  localStorage.setItem('authToken', token);
  localStorage.setItem('authTokenExpiry', expiryTime);
 }

+/**
+ * 获取存储的认证令牌
+ * @returns {string|null} 如果令牌有效则返回令牌，否则返回 null
+ */
 function getAuthToken() {
  const token = localStorage.getItem('authToken');
  const expiry = localStorage.getItem('authTokenExpiry');
@@ -23,6 +32,13 @@ function getAuthToken() {
 }

 // 消息显示功能
+/**
+ * 在指定元素中显示消息
+ * @param {string} elementId - 目标元素的 ID
+ * @param {string} text - 要显示的消息文本
+ * @param {boolean} [isError=false] - 是否为错误消息
+ * @returns {void}
+ */
 function showMessage(elementId, text, isError = false) {
  let msg = document.getElementById(elementId);

@@ -38,6 +54,10 @@ function showMessage(elementId, text, isError = false) {
 }

 // 确保消息容器存在
+/**
+ * 确保消息容器存在于 DOM 中
+ * @returns {HTMLElement} 消息容器元素
+ */
 function ensureMessageContainer() {
  let container = document.querySelector('.message-container');
  if (!container) {
@@ -48,6 +68,13 @@ function ensureMessageContainer() {
  return container;
 }

+/**
+ * 显示全局消息提示
+ * @param {string} text - 要显示的消息文本
+ * @param {boolean} [isError=false] - 是否为错误消息
+ * @param {number} [timeout=3000] - 消息显示时长（毫秒）
+ * @returns {void}
+ */
 function showGlobalMessage(text, isError = false, timeout = 3000) {
  const container = ensureMessageContainer();

@@ -270,3 +297,27 @@ function showPromptModal(promptStr) {
    console.error('原始prompt:', promptStr);
  }
 }
+
+/**
+ * 将会员类型代码转换为显示名称
+ * @param {string|null} type - 会员类型代码,如 'free_trial', 'pro', 'free', 'enterprise' 等
+ * @returns {string} 格式化后的会员类型显示名称
+ * @example
+ * formatMembershipType('free_trial') // 返回 'Pro Trial'
+ * formatMembershipType('pro') // 返回 'Pro'
+ * formatMembershipType(null) // 返回 '-'
+ * formatMembershipType('custom_type') // 返回 'Custom Type'
+ */
+function formatMembershipType(type) {
+  if (!type) return '-';
+  switch (type) {
+    case 'free_trial': return 'Pro Trial';
+    case 'pro': return 'Pro';
+    case 'free': return 'Free';
+    case 'enterprise': return 'Business';
+    default: return type
+      .split('_')
+      .map(word => word.charAt(0).toUpperCase() + word.slice(1))
+      .join(' ');
+  }
+}
--- a/static/tokens.html
+++ b/static/tokens.html
@@ -274,6 +274,7 @@
      <h3>Token 管理</h3>
      <div class="button-group">
        <button onclick="getTokenInfo()">获取当前配置</button>
+        <button onclick="reloadTokens()" class="secondary">重载Token</button>
        <button onclick="addTokens()" class="secondary">添加Token</button>
        <button onclick="deleteTokens()" class="danger">删除Token</button>
      </div>
@@ -295,6 +296,10 @@
            <tr>
              <th>Token</th>
              <th>Checksum</th>
+              <th>邮箱</th>
+              <th>会员类型</th>
+              <th>Premium用量</th>
+              <th>试用剩余</th>
              <th class="action-cell">操作</th>
            </tr>
          </thead>
@@ -317,22 +322,6 @@
    <div class="modal-header">
      <h3>生成动态Key</h3>
    </div>
-    <div class="form-group">
-      <label>流第一个块检查:</label>
-      <select id="enableStreamCheck">
-        <option value="">跟随全局</option>
-        <option value="true">启用</option>
-        <option value="false">禁用</option>
-      </select>
-    </div>
-    <div class="form-group">
-      <label>包含停止流:</label>
-      <select id="includeStopStream">
-        <option value="">跟随全局</option>
-        <option value="true">启用</option>
-        <option value="false">禁用</option>
-      </select>
-    </div>
    <div class="form-group">
      <label>图片处理能力:</label>
      <select id="disableVision">
@@ -362,6 +351,14 @@
        <!-- 模型列表将通过 JavaScript 动态填充 -->
      </div>
    </div>
+    <div class="form-group">
+      <label>包含网络引用:</label>
+      <select id="includeWebReferences">
+        <option value="">跟随全局</option>
+        <option value="true">启用</option>
+        <option value="false">禁用</option>
+      </select>
+    </div>
    <div class="key-result" id="keyResult" style="display: none;" onclick="copyGeneratedKey()">
      <div class="key-content" id="keyContent"></div>
    </div>
@@ -376,7 +373,15 @@
      const data = await makeAuthenticatedRequest('/tokens/get');
      if (data) {
        const tableBody = document.getElementById('tokenTableBody');
-        tableBody.innerHTML = data.tokens.map(t => `<tr><td title="${t.token}">${t.token}</td><td title="${t.checksum}">${t.checksum}</td><td class="action-cell"><button onclick="showKeyModal('${t.token}','${t.checksum}')" class="secondary">生成Key</button></td></tr>`).join('');
+        tableBody.innerHTML = data.tokens.map(t => {
+          const profile = t.profile || {};
+          const user = profile.user || {};
+          const stripe = profile.stripe || {};
+          const usage = profile.usage || {};
+          const premium = usage.premium || {};
+
+          return `<tr><td title="${t.token}">${t.token}</td><td title="${t.checksum}">${t.checksum}</td><td>${user.email || '-'}</td><td>${formatMembershipType(stripe.membership_type)}</td><td>${premium.requests || 0}/${premium.max_requests || '∞'}</td><td>${stripe.days_remaining_on_trial > 0 ? `${stripe.days_remaining_on_trial}天` : '-'}</td><td class="action-cell"><button onclick="showKeyModal('${t.token}','${t.checksum}')" class="secondary">生成Key</button></td></tr>`;
+        }).join('');
        showGlobalMessage('配置获取成功');
      }
    }
@@ -397,6 +402,14 @@
      });
    }

+    async function reloadTokens() {
+      const data = await makeAuthenticatedRequest('/tokens/reload');
+      if (data) {
+        showGlobalMessage(`Token重载成功: ${data.message}`);
+        getTokenInfo(); // 刷新当前配置
+      }
+    }
+
    async function addTokens() {
      const tokensInput = document.getElementById('tokenInput').value;

@@ -508,12 +521,11 @@
      });

      // 重置所有选项
-      document.getElementById('enableStreamCheck').value = '';
-      document.getElementById('includeStopStream').value = '';
      document.getElementById('disableVision').value = '';
      document.getElementById('enableSlowPool').value = '';
      document.getElementById('usageCheckType').value = '';
      document.getElementById('modelListContainer').style.display = 'none';
+      document.getElementById('includeWebReferences').value = '';
    }

    function closeKeyModal() {
@@ -537,14 +549,13 @@

      const payload = {
        auth_token: `${currentToken},${currentChecksum}`,
-        enable_stream_check: parseBooleanFromString(document.getElementById('enableStreamCheck').value, undefined),
-        include_stop_stream: parseBooleanFromString(document.getElementById('includeStopStream').value, undefined),
        disable_vision: parseBooleanFromString(document.getElementById('disableVision').value, undefined),
        enable_slow_pool: parseBooleanFromString(document.getElementById('enableSlowPool').value, undefined),
        usage_check_models: type ? {
          type: type,
          model_ids: type === 'custom' ? modelIds : undefined
-        } : undefined
+        } : undefined,
+        include_web_references: parseBooleanFromString(document.getElementById('includeWebReferences').value, undefined)
      };

      const data = await makeAuthenticatedRequest('/build-key', {
--- a/tests/data/stream_data.txt
+++ b/tests/data/stream_data.txt
--- a/worker.js
+++ b/worker.js
@@ -11,6 +11,7 @@ async function handleRequest(request) {
    const allowedHosts = ["api2.cursor.sh", "www.cursor.com"];
    const allowedPaths = [
      "/aiserver.v1.AiService/StreamChat",
+      "/aiserver.v1.AiService/StreamChatWeb",
      "/auth/full_stripe_profile",
      "/api/usage",
      "/api/auth/me"
				`@@ -0,0 +1 @@`
				Deno.serve(async(r:Request)=>{const rs=(s:number,m:string)=>new Response(m,{status:s,headers:{"Access-Control-Allow-Origin":""}});const h=r.headers.get("x-co");if(!h)return rs(400,"Missing header");const a=["api2.cursor.sh","www.cursor.com"];if(!a.includes(h))return rs(403,"Host denied");const u=new URL(r.url),p=["/aiserver.v1.AiService/StreamChat","/aiserver.v1.AiService/StreamChatWeb","/auth/full_stripe_profile","/api/usage","/api/auth/me"];if(!p.includes(u.pathname))return rs(404,"Path invalid");const hd=new Headers(r.headers);hd.delete("x-co");hd.set("Host",h);try{const f=await fetch(`https://${h}${u.pathname}${u.search}`,{method:r.method,headers:hd,body:r.body});const fh=new Headers(f.headers);fh.set("Access-Control-Allow-Origin","");return new Response(f.body,{status:f.status,headers:fh})}catch(e){return rs(500,"Server error")}});