Overview
Token Usage
Requests & Latency
Top Modelsby requests
Endpointsby requests
Failures by Status
Failed Requests0
Endpoints Status
Recent Requests
Top Users (7 days)
Manage Endpoints
Manage Users
User Groups
Config group áp dụng tức thời. Group Admin không cấp quyền dashboard.
Request Logs
User Tracking
Select User
Select a user to view their activity
Claude Converter
Global Toggle
When enabled, any request where the model name contains claude will be automatically
converted from OpenAI Chat Completions format to Anthropic's native Messages API format,
with full SDK identity headers (Anthropic/JS 0.94.0 + x-stainless-* headers).
This handles: system message extraction, message role alternation merging, streaming SSE translation, image conversion, and thinking block filtering.
Non-Claude models are never affected and pass through unchanged.
When enabled, Claude text is streamed live as it is generated (asks upstream for
stream:true and translates Anthropic SSE to OpenAI chunks).
When disabled, the converter waits for the full answer and sends it in one piece —
the old behaviour. Turn this OFF if a provider misbehaves while streaming;
the converter keeps working either way. No restart needed.
Legacy. Simulates streaming by slicing an already-complete answer into 6-character pieces with a 10ms pause between each, so it is SLOWER than just waiting. Ignored when Real Streaming is on. Leave OFF.
Other Converter
Custom Model Converter
Converts only the exact model IDs listed below from OpenAI Chat Completions to the upstream native Anthropic Messages API. Matching is case-insensitive. Unlisted models pass through unchanged.
Streaming sends an immediate OpenAI SSE handshake, keeps waiting clients alive, and starts one delayed backup request when the upstream does not return headers within 15 seconds.
Settings
Custom Error Messages
Khi upstream trả lỗi, proxy sẽ hiện message tùy chỉnh cho user thay vì raw error. Để trống = dùng message mặc định.
Output Token Limit (max_tokens)
Giới hạn output tokens cho mọi request chat completion.
Proxy sẽ inject hoặc clamp giá trị max_tokens về đây trước khi gọi upstream.
0 = tắt (dùng default của upstream).
Venom Prompt
Bật: proxy chèn prompt Venom làm system message đầu tiên của mọi request chat. AI chỉ xưng Venom khi từ chối, lỗi hoặc đề xuất nội dung khác; prompt của khách luôn được ưu tiên. Lưu ngay khi bấm, không cần restart.
Loading...