2e14562bd2
Refactor toxicity analysis implementation to be independent of specific LLM providers. Update configuration and code to use generic terminology and allow flexibility in choosing language models. Changes: - Update environment variable naming for API credentials - Generalize documentation to support multiple LLM providers - Improve configuration flexibility for model selection - Add project documentation files to gitignore 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
26 lines
559 B
Text
26 lines
559 B
Text
# PostgreSQL
|
|
POSTGRES_USER=bluesky
|
|
POSTGRES_PASSWORD=changeme
|
|
POSTGRES_PORT=5433
|
|
|
|
# Collector settings
|
|
LOG_LEVEL=INFO
|
|
MAX_PAGES_PER_ACCOUNT=50
|
|
MENTION_LOOKBACK_HOURS=12
|
|
|
|
# Bluesky authentication (required for mention search)
|
|
# Create an app password at: Settings → App Passwords
|
|
BSKY_HANDLE=
|
|
BSKY_APP_PASSWORD=
|
|
|
|
# Toxicity Analyzer (LLM)
|
|
LLM_API_KEY=
|
|
ANALYZER_MODEL=gpt-4.1-nano
|
|
ANALYZER_CONCURRENCY=3
|
|
ANALYZER_BATCH_SIZE=10
|
|
|
|
# Web UI
|
|
WEB_PORT=5001
|
|
|
|
# Scheduling is controlled by ofelia cron in docker-compose.yml
|
|
# Default: every 4 hours ("0 0 */4 * * *")
|