.env.example

HOST=127.0.0.1
PORT=8080

# GitHub Classic PAT: check the "repo" scope when creating it.
GITHUB_TOKEN=ghp_xxx

# OpenAI-compatible API credentials. This example uses DeepSeek.
AI_API_KEY=sk-xxx
AI_BASE_URL=https://api.deepseek.com
AI_MODEL=deepseek-v4-flash
# Thinking mode: disabled (default) or enabled.
AI_THINKING=disabled
# Maximum number of tokens generated for one AI batch response.
AI_MAX_TOKENS=16000

# Telegram BotFather token and the destination chat ID.
TELEGRAM_BOT_TOKEN=123456:xxx
TELEGRAM_CHAT_ID=123456789

# Number of repositories sent to each AI request. All Stars are loaded, then
# review requests are sent in these batches.
REVIEW_BATCH_SIZE=50

# SQLite database for repositories and review results.
DATABASE_FILE=review-stars.db

# Optional 5-field cron expression. Example: every day at 09:00.
# @every 24h is also supported.
REVIEW_CRON=

# Number of different repositories to send to Telegram per scheduled run.
REVIEW_COUNT=1

# Deterministic pre-AI filter. All enabled rule groups must match simultaneously;
# statuses inside RULE_STATUSES are alternatives.
# Set RULE_STALE_DAYS or RULE_MAX_STARS to 0 to disable that rule.
RULE_STATUSES=archived
RULE_STALE_DAYS=180
RULE_MAX_STARS=1000