Onllama.LiteGateway

Prevent Ollama from being abused by exposing it directly to the public network. The easiest way to add Apikey authentication to Ollama.

防止 Ollama 直接暴露在公网受到滥用，为 Ollama 添加 Apikey 鉴权的最简单方式。

wget https://github.com/onllama/Onllama.LiteGateway/releases/latest/download/Onllama.LiteGateway.linux-x64 -O /usr/bin/Onllama.LiteGateway
wget https://raw.githubusercontent.com/onllama/Onllama.LiteGateway/refs/heads/main/onllama-litegateway@.service -O /etc/systemd/system/onllama-litegateway@.service
chmod +x /usr/bin/Onllama.LiteGateway 
systemctl enable onllama-litegateway@sk-just-for-example --now

# Please replace 'sk-just-for-example' with the apikey you want to set. / 请替换 'sk-just-for-example' 为你想要设置的 apikey。
# curl http://127.0.0.1:22434

Onllama.LiteGateway - The simplest Ollama authentication way.
Copyright (c) 2025 Milkey Tan. Code released under the MIT License

Usage: Onllama.LiteGateway [options] <keys>

Arguments:
  keys                                     Setting up the API key

Options:
  -?|-he|--help                            Show help information.
  -l|--listen <IPEndPoint>                 Set server listening address and port
  -t|--target <Uri>                        Set target address and port
  --log                                    Enable logging
  -s|--https                               Set enable HTTPS (Self-signed by default, not recommended)
  -pem|--pemfile[:<FilePath>]              Set your pem certificate file path <./cert.pem>
  -key|--keyfile[:<FilePath>]              Set your pem certificate key file path <./cert.key>
  -h|--host[:<Hostname>]                   Set the allowed host names. (Allow all by default)
  --no-think-trim                          Disable ThinkTrim
  --no-token                               Disable API key verification
  --use-model-manage                       Enable model management via API
  --use-model-info-public-path             Allows get model information via API without APIKEY
  --use-cors-any                           Allow cross-origin requests
  --use-rate-limit                         Enable request rate limiting (with ipratelimiting.json)
  --num-ctx[:<NumCtx>]                     Set the number of contexts per request
  --use-input-security                     Enable input security
  --risk-model[:<RiskModel>]               Set the input security risk model
  --risk-model-prompt[:<RiskModelPrompt>]  Set the input security risk model prompt
  --risk-keywords <RiskKeywords>           Set the input security risk model keywords

TODO

• [x] HTTPS
• [x] Host
• [x] Log
• [x] CORS
• [x] Rate Limit
• [x] Public URL
• [x] Trim <think/>
• [x] Override num_ctx
• [x] Input content security based on llamaguard/shieldgemma
• [ ] Auto context summary compress
• [ ] Apikey from DB
• [ ] ACME Auto HTTPS