CodeTrans/.env.example at main · cld2labs/CodeTrans · GitHub

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
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
# ============================================================
# CodeTrans — Environment Configuration
# ============================================================

# Backend port
BACKEND_PORT=5001

# ============================================================
# Inference Provider
# ============================================================
# "remote"  — Cloud or enterprise OpenAI-compatible API (e.g. CodeLlama via gateway)
# "ollama"  — Local Ollama running natively on the host machine (recommended for Mac)
INFERENCE_PROVIDER=remote

# ============================================================
# Option A: Remote OpenAI-compatible API (INFERENCE_PROVIDER=remote)
# ============================================================
# INFERENCE_API_ENDPOINT: Base URL of your inference service (no /v1 suffix)
#   - GenAI Gateway:   https://genai-gateway.example.com
#   - APISIX Gateway:  https://apisix-gateway.example.com/CodeLlama-34b-Instruct-hf
INFERENCE_API_ENDPOINT=https://your-api-endpoint.com/deployment
INFERENCE_API_TOKEN=your-pre-generated-token-here
INFERENCE_MODEL_NAME=codellama/CodeLlama-34b-Instruct-hf

# ============================================================
# Option B: Ollama — native host inference (INFERENCE_PROVIDER=ollama)
# ============================================================
#
# IMPORTANT — Why Ollama runs on the host, NOT in Docker:
#   On macOS (Apple Silicon / M-series), running Ollama as a Docker container
#   bypasses Metal GPU acceleration. The model falls back to CPU-only inference
#   which is dramatically slower. Ollama must be installed natively so the Metal
#   Performance Shaders (MPS) backend is used for hardware-accelerated inference.
#
# Setup:
#   1. Install Ollama: https://ollama.com/download
#   2. Pull your model (see options below)
#   3. Ollama starts automatically; confirm it is running:
#        curl http://localhost:11434/api/tags
#   4. Set the variables below in your .env
#
# The backend container reaches host-side Ollama via the special DNS name
# `host.docker.internal` which Docker Desktop resolves to the Mac host.
# (On Linux with Docker Engine this requires the extra_hosts entry in docker-compose.yaml,
#  which is already configured.)
#
# --- Production / high-quality translation ---
# INFERENCE_PROVIDER=ollama
# INFERENCE_API_ENDPOINT=http://host.docker.internal:11434
# INFERENCE_MODEL_NAME=codellama:34b
# ollama pull codellama:34b          # ~20 GB, best quality
#
# --- Testing / SLM performance benchmarking ---
# INFERENCE_PROVIDER=ollama
# INFERENCE_API_ENDPOINT=http://host.docker.internal:11434
# INFERENCE_MODEL_NAME=codellama:7b
# ollama pull codellama:7b           # ~4 GB, fast — use this for gauging SLM perf
#
# --- Other recommended code models ---
# ollama pull deepseek-coder:6.7b    # ~4 GB, strong at code tasks
# ollama pull qwen2.5-coder:7b       # ~4 GB, excellent multilingual code
# ollama pull codellama:13b          # ~8 GB, good balance of speed vs quality
#
# Note: INFERENCE_API_TOKEN is not required when using Ollama.

# ============================================================
# LLM Settings
# ============================================================
LLM_TEMPERATURE=0.2
LLM_MAX_TOKENS=4096

# ============================================================
# Code Translation Settings
# ============================================================
MAX_CODE_LENGTH=8000
MAX_FILE_SIZE=10485760

# ============================================================
# CORS Configuration
# ============================================================
CORS_ALLOW_ORIGINS=["http://localhost:5173", "http://localhost:3000"]

# ============================================================
# Local URL Endpoint
# ============================================================
# Only needed if your remote API endpoint is a private domain mapped in /etc/hosts.
# Otherwise leave as "not-needed".
LOCAL_URL_ENDPOINT=not-needed

# ============================================================
# SSL Verification
# ============================================================
# Set to false only for development with self-signed certificates.
VERIFY_SSL=true