Repository navigation
Expand file tree
/
Copy path.env.example
More file actions
42 lines (36 loc) · 1.62 KB
/
Copy path.env.example
File metadata and controls
42 lines (36 loc) · 1.62 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
# Backend
STORAGE_PATH=./storage
CORS_ORIGINS=["http://localhost:3000","http://localhost:4001"]
MAX_UPLOAD_MB=20
# AI API Key (optional - users can provide their own)
GOOGLE_API_KEY=
# Gemini model for mask generation (optional)
# gemini-3-pro-image-preview - best quality, accurate dimensions (default)
# gemini-2.5-flash-image - faster/cheaper, needs post-hoc alignment
GEMINI_IMAGE_MODEL=gemini-3-pro-image-preview
# Local tracing (optional)
# Comma-separated tracer list, e.g. birefnet-lite,isnet
# Add birefnet-general only when you have installed the optional GPU requirements.
TRACERS=
# auto prefers CUDA when available; cuda requires GPU; cpu disables GPU use
TRACEFINITY_ONNX_PROVIDER=auto
# Automatic tool names after tracing (optional)
# none disables naming; ollama uses a local vision model through ToolNamer;
# openrouter uses OpenRouter's chat completions API (see OPENROUTER_API_KEY /
# OPENROUTER_LABEL_MODEL below).
TOOL_LABEL_PROVIDER=none
TOOL_LABEL_MODEL=qwen3-vl:4b
TOOL_LABEL_OLLAMA_URL=http://localhost:11434
TOOL_LABEL_TIMEOUT_SECONDS=30
TOOL_LABEL_MAX_CROP_PX=512
# OpenRouter (optional) -- only needed when TOOL_LABEL_PROVIDER=openrouter
OPENROUTER_API_KEY=
# Comma-separated priority list; tried in order, falling through to the next
# model on HTTP 429 (free-tier ":free" models share a 20 req/min pool per
# model across all OpenRouter users, so a popular one can rate-limit fast).
OPENROUTER_LABEL_MODEL=google/gemini-2.0-flash-001
# Chat completions endpoint. Any OpenAI-compatible API that accepts image
# inputs works here; leave empty for OpenRouter.
OPENROUTER_URL=
# Frontend
NEXT_PUBLIC_API_URL=http://localhost:8000