Affected Component
Analytics Platform Integration (Langfuse)
Describe the bug
Steps to Reproduce
无法使用，公网目标授权门禁非常严格
System Configuration
无法使用，公网目标授权门禁非常严格
Logs and Artifacts
No response
Screenshots or Recordings
No response
Verification
- I have checked that this issue hasn't been already reported
- I have provided all relevant configuration files (with sensitive data removed)
- I have included relevant logs and error messages
- I am running the latest version of PentAGI
Activity
2
AlaeddineMessadi15d ago这个 400 不是 PentAGI 返回的。"This content was flagged for possible cybersecurity risk" 是 LLM 服务商 API 返回的内容，PentAGI 的代码不会生成这句话，只是在模型调用失败时加上 "agent model call failed" 前缀后显示出来（performer.go#L553-L555）。
可以按提示向该服务商申请 Trusted Access for Cyber；或者换一个服务商/模型来跑这个 flow：在 Settings → Providers 里新建一个配置，或者用自定义的 OpenAI 兼容接口（LLM_SERVER_URL）、本地 Ollama（OLLAMA_SERVER_URL）。如果用的是 OpenAI 早期的 GPT-5 / 5.1 / 5.2 模型，README 里有相关说明。
This 400 does not come from PentAGI. "This content was flagged for possible cybersecurity risk" is the response from the LLM provider's API. PentAGI never produces that text, it only prefixes the failed model call with "agent model call failed" (performer.go#L553-L555).
Either apply for the provider's Trusted Access for Cyber program as the message says, or run the flow on a different provider or model: add one under Settings → Providers, or use a custom OpenAI-compatible endpoint (LLM_SERVER_URL) or a local Ollama (OLLAMA_SERVER_URL). If you are on OpenAI's older GPT-5/5.1/5.2 models, the README has a note on exactly this.
遇到类似场景的话，建议补充几项信息会更容易定位：1) 具体报错文案/截图（授权门禁拦截时返回什么状态码或提示）；2) PentAGI 版本和部署方式（docker / 源码）；3) 公网目标的接入方式（agent gateway / 直连 / 自定义证书）。这类"公网目标授权"通常卡在 URL 白名单或 TLS 校验，贴出配置片段和失败日志，维护者可以直接判断是配置问题还是代码限制。同时可以确认一下问题是否只影响 Langfuse 集成——如果仅回调场景报错，可能是回调地址校验过严。