English | 简体中文
A reverse engineered unofficial ChatGPT
proxy (bypass Cloudflare 403 Access Denied)
- API key acquisition
- Email/password account authentication (Google/Microsoft third-party login is not supported for now because the author does not have an account)
Unofficial
/Official
/ChatGPT-to-API
Http API proxy (for third-party client access)- The original ChatGPT WebUI
- Minimal memory usage
Limitations: This cannot bypass OpenAI's outright IP ban
- Linux musl current supports
x86_64-unknown-linux-musl
aarch64-unknown-linux-musl
armv7-unknown-linux-musleabi
armv7-unknown-linux-musleabihf
arm-unknown-linux-musleabi
arm-unknown-linux-musleabihf
armv5te-unknown-linux-musleabi
- Windows current supports
x86_64-pc-windows-msvc
- MacOS current supports
x86_64-apple-darwin
aarch64-apple-darwin
Making Releases has a precompiled deb package, binaries, in Ubuntu, for example:
wget https://github.com/gngpp/opengpt/releases/download/v0.3.9/opengpt-0.3.9-x86_64-unknown-linux-musl.deb
dpkg -i opengpt-0.3.9-x86_64-unknown-linux-musl.deb
opengpt serve run
docker run --rm -it -p 7999:7999 --hostname=opengpt \
-e OPENGPT_WORKERS=1 \
-e OPENGPT_LOG_LEVEL=info \
gngpp/opengpt:latest serve run
docker-compose
version: '3'
services:
opengpt:
image: ghcr.io/gngpp/opengpt:latest
container_name: opengpt
restart: unless-stopped
environment:
- TZ=Asia/Shanghai
- OPENGPT_PROXY=socks5://cloudflare-warp:10000
# - OPENGPT_CONFIG=/serve.toml
# - OPENGPT_PORT=8080
# - OPENGPT_HOST=0.0.0.0
# - OPENGPT_WORKERS=10
# - OPENGPT_TIMEOUT=360
# - OPENGPT_CONNECT_TIMEOUT=60
# - OPENGPT_TCP_KEEPALIVE=60
# - OPENGPT_TLS_CERT=
# - OPENGPT_TLS_KEY=
# - OPENGPT_UI_API_PREFIX=
# - OPENGPT_SIGNATURE=
# - OPENGPT_TB_ENABLE=
# - OPENGPT_TB_STORE_STRATEGY
# - OPENGPT_TB_REDIS_URL=
# - OPENGPT_TB_CAPACITY=
# - OPENGPT_TB_FILL_RATE=
# - OPENGPT_TB_EXPIRED=
# - OPENGPT_CF_SITE_KEY=
# - OPENGPT_CF_SECRET_KEY=
# volumes:
# - ${PWD}/ssl:/etc
# - ${PWD}/serve.toml:/serve.toml
command: serve run
ports:
- "8080:7999"
depends_on:
- cloudflare-warp
cloudflare-warp:
container_name: cloudflare-warp
image: ghcr.io/gngpp/cloudflare-warp:latest
restart: unless-stopped
watchtower:
container_name: watchtower
image: containrrr/watchtower
volumes:
- /var/run/docker.sock:/var/run/docker.sock
command: --interval 3600 --cleanup
restart: unless-stopped
There are pre-compiled ipk files in GitHub Releases, which currently provide versions of aarch64/x86_64 and other architectures. After downloading, use opkg to install, and use nanopi r4s as example:
wget https://github.com/gngpp/opengpt/releases/download/v0.3.9/opengpt_0.3.9_aarch64_generic.ipk
wget https://github.com/gngpp/opengpt/releases/download/v0.3.9/luci-app-opengpt_1.0.2-1_all.ipk
wget https://github.com/gngpp/opengpt/releases/download/v0.3.9/luci-i18n-opengpt-zh-cn_1.0.2-1_all.ipk
opkg install opengpt_0.3.9_aarch64_generic.ipk
opkg install luci-app-opengpt_1.0.2-1_all.ipk
opkg install luci-i18n-opengpt-zh-cn_1.0.2-1_all.ipk
Public API,
*
means anyURL
suffix
- backend-api, https://host:port/backend-api/*
- public-api, https://host:port/public-api/*
- platform-api, https://host:port/v1/*
- dashboard-api, https://host:port/dashboard/*
- chatgpt-to-api, https://host:port/conv/v1/chat/completions
Detailed API documentation
-
Authentic ChatGPT WebUI
-
Expose
unofficial
/official API
proxies -
The
API
prefix is consistent with the official -
ChatGPT
ToAPI
-
Accessible to third-party clients
-
Access to IP proxy pool to improve concurrency
-
API documentation
-
Parameter Description
--level
, environment variableOPENGPT_LOG_LEVEL
, log level: default info--host
, environment variableOPENGPT_HOST
, service listening address: default 0.0.0.0,--port
, environment variableOPENGPT_PORT
, listening port: default 7999--workers
, environment variableOPENGPT_WORKERS
, worker threads: default 1--tls-cert
, environment variableOPENGPT_TLS_CERT
', TLS certificate public key. Supported format: EC/PKCS8/RSA--tls-key
, environment variableOPENGPT_TLS_KEY
, TLS certificate private key--proxies
, environment variableOPENGPT_PROXY
, proxies,support multiple proxy pools, format: protocol://user:pass@ip:port
...
$ opengpt serve --help
Start the http server
Usage: opengpt serve run [OPTIONS]
Options:
-C, --config <CONFIG>
Configuration file path (toml format file) [env: OPENGPT_CONFIG=]
-H, --host <HOST>
Server Listen host [env: OPENGPT_HOST=] [default: 0.0.0.0]
-L, --level <LEVEL>
Log level (info/debug/warn/trace/error) [env: OPENGPT_LOG_LEVEL=] [default: info]
-P, --port <PORT>
Server Listen port [env: OPENGPT_PORT=] [default: 7999]
-W, --workers <WORKERS>
Server worker-pool size (Recommended number of CPU cores) [env: OPENGPT_WORKERS=] [default: 1]
--concurrent-limit <CONCURRENT_LIMIT>
Concurrent limit (Enforces a limit on the concurrent number of requests the underlying) [env: OPENGPT_CONCUURENT_LIMIT=] [default: 65535]
--proxies <PROXIES>
Server proxies pool, Example: protocol://user:pass@ip:port [env: OPENGPT_PROXY=]
--timeout <TIMEOUT>
Client timeout (seconds) [env: OPENGPT_TIMEOUT=] [default: 600]
--connect-timeout <CONNECT_TIMEOUT>
Client connect timeout (seconds) [env: OPENGPT_CONNECT_TIMEOUT=] [default: 60]
--tcp-keepalive <TCP_KEEPALIVE>
TCP keepalive (seconds) [env: OPENGPT_TCP_KEEPALIVE=] [default: 60]
--tls-cert <TLS_CERT>
TLS certificate file path [env: OPENGPT_TLS_CERT=]
--tls-key <TLS_KEY>
TLS private key file path (EC/PKCS8/RSA) [env: OPENGPT_TLS_KEY=]
--puid <PUID>
PUID cookie value of Plus account [env: OPENGPT_PUID=]
--puid-user <PUID_USER>
Obtain the PUID of the Plus account user. Supports split: ':', '-', '--' ... , Example: `user:pass` or `user:pass:mfa` [env: OPENGPT_PUID_USER=]
--api-prefix <API_PREFIX>
Web UI api prefix [env: OPENGPT_UI_API_PREFIX=]
--arkose-endpoint <ARKOSE_ENDPOINT>
Arkose endpoint, Example: https://client-api.arkoselabs.com [env: OPENGPT_ARKOSE_ENDPOINT=]
-A, --arkose-token-endpoint <ARKOSE_TOKEN_ENDPOINT>
Get arkose-token endpoint [env: OPENGPT_ARKOSE_TOKEN_ENDPOINT=]
--arkose-yescaptcha-key <ARKOSE_YESCAPTCHA_KEY>
yescaptcha client key [env: OPENGPT_ARKOSE_YESCAPTCHA_KEY=]
-S, --sign-secret-key <SIGN_SECRET_KEY>
Enable url signature (signature secret key) [env: OPENGPT_SIGNATURE=]
-T, --tb-enable
Enable token bucket flow limitation [env: OPENGPT_TB_ENABLE=]
--tb-store-strategy <TB_STORE_STRATEGY>
Token bucket store strategy (mem/redis) [env: OPENGPT_TB_STORE_STRATEGY=] [default: mem]
--tb-redis-url <TB_REDIS_URL>
Token bucket redis url, Example: redis://user:pass@ip:port [env: OPENGPT_TB_REDIS_URL=] [default: redis://127.0.0.1:6379]
--tb-capacity <TB_CAPACITY>
Token bucket capacity [env: OPENGPT_TB_CAPACITY=] [default: 60]
--tb-fill-rate <TB_FILL_RATE>
Token bucket fill rate [env: OPENGPT_TB_FILL_RATE=] [default: 1]
--tb-expired <TB_EXPIRED>
Token bucket expired (seconds) [env: OPENGPT_TB_EXPIRED=] [default: 86400]
--cf-site-key <CF_SITE_KEY>
Cloudflare turnstile captcha site key [env: OPENGPT_CF_SITE_KEY=]
--cf-secret-key <CF_SECRET_KEY>
Cloudflare turnstile captcha secret key [env: OPENGPT_CF_SECRET_KEY=]
-D, --disable-webui
Disable WebUI [env: OPENGPT_DISABLE_WEBUI=]
-h, --help
Print help
- Linux compile, Ubuntu machine for example:
sudo apt update -y && sudo apt install rename
# Native compilation
git clone https://github.com/gngpp/opengpt.git && cd opengpt
./build.sh
# Cross-platform compilation, relying on docker (if you can solve cross-platform compilation dependencies on your own)
# Default using docker build linux/windows platform
./build_cross.sh
# The MacOS platform is built on MacOS by default
os=macos ./build_cross.sh
# Compile a single platform binary, take aarch64-unknown-linux-musl as an example:
docker run --rm -it --user=$UID:$(id -g $USER) \
-v $(pwd):/home/rust/src \
-v $HOME/.cargo/registry:/root/.cargo/registry \
-v $HOME/.cargo/git:/root/.cargo/git \
ghcr.io/gngpp/opengpt-builder:x86_64-unknown-linux-musl \
cargo build --release
- OpenWrt compile
cd package
svn co https://github.com/gngpp/opengpt/trunk/openwrt
cd -
make menuconfig # choose LUCI->Applications->luci-app-opengpt
make V=s