Skip to content

EnvironmentVariables.md

Latest commit

 

History

History
177 lines (126 loc) · 8.29 KB

EnvironmentVariables.md

File metadata and controls

177 lines (126 loc) · 8.29 KB
Raw

Environment Variables Reference

This document summarizes the environment variables consumed by the SOCIALIZER platform. Copy .env.example to .env and override the values per environment (config/environments/development.env, staging.env, production.env).

Core Settings

  • ENV: Runtime environment (development, staging, production).
  • DEBUG: Enables verbose logging when true.
  • SECRET_KEY: 32+ character secret used for signing tokens.
  • API_VERSION: Prefix for API URLs (v1).

Server

  • BACKEND_HOST / BACKEND_PORT: FastAPI host and port.
  • FRONTEND_URL: Public URL for the web client.
  • CORS_ORIGINS: Comma-separated origins allowed to access the API.
  • PYTHONPATH: Ensures Celery/Uvicorn processes resolve backend modules without extra flags (src/backend in local environments).

Database

  • DATABASE_URL: PostgreSQL connection string for application traffic.
  • DATABASE_POOL_SIZE: SQLAlchemy connection pool size.
  • DATABASE_MAX_OVERFLOW: Additional connections allowed beyond pool size.
  • DATABASE_ECHO: Enable SQL logging when true.
  • TEST_DATABASE_URL: Separate connection string for tests.

Redis & Celery

  • REDIS_URL: Primary Redis instance for caching.
  • REDIS_PASSWORD: Optional password for Redis.
  • REDIS_MAX_CONNECTIONS: Max Redis connection pool.
  • CELERY_BROKER_URL: Redis broker for Celery workers.
  • CELERY_RESULT_BACKEND: Result backend for Celery.
  • CELERY_TASK_TIME_LIMIT: Per-task execution limit (seconds).
  • CELERY_WORKER_PREFETCH_MULTIPLIER: Worker prefetch behaviour.
  • FLOWER_PORT: Flower monitoring web UI port.

Instagram / Apify

  • APIFY_TOKEN: Apify API token with rights to run apify/instagram-profile-scraper. Required for /api/v1/ingest/instagram/profiles and the /instagram/ingest UI.

Reddit API

  • REDDIT_CLIENT_ID, REDDIT_CLIENT_SECRET, REDDIT_USER_AGENT: OAuth credentials.
  • REDDIT_USERNAME, REDDIT_PASSWORD: Script application credentials.
  • REDDIT_RATE_LIMIT_CALLS, REDDIT_RATE_LIMIT_PERIOD: Rate limit configuration.

AWS / Object Storage

  • AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_REGION: AWS credentials.
  • AWS_S3_BUCKET: Bucket used for persisted artifacts.
  • S3_RAW_PREFIX, S3_MODELS_PREFIX, S3_EXPORTS_PREFIX: Folder prefixes.

Machine Learning

  • PRIMARY_MODEL: Hugging Face model id used for primary sentiment analysis (cardiffnlp/twitter-roberta-base-topic-sentiment-latest by default).
  • FALLBACK_SERVICE: Label used when Azure Opinion Mining provides the sentiment result.
  • CONFIDENCE_THRESHOLD: Minimum primary-model confidence before triggering Azure fallback.
  • SENTIMENT_MIN_CONF, SENTIMENT_MIN_MARGIN: Overrideable Hugging Face confidence and margin thresholds before fallback.
  • FALLBACK_ENABLED: Toggle to disable Azure fallback entirely when set to false.
  • AZURE_TEXT_ANALYTICS_ENDPOINT, AZURE_TEXT_ANALYTICS_KEY: Credentials for Azure Text Analytics Opinion Mining (stored in secret manager in production).
  • MODEL_PATH, MODEL_VERSION: Legacy model binary metadata retained for archival workflows.
  • MODEL_DEVICE: Legacy inference device flag (cuda or cpu).
  • BATCH_SIZE, MAX_SEQUENCE_LENGTH: Inference/batching parameters.
  • ML_INFERENCE_URL, ML_TIMEOUT_SECONDS: Legacy external inference endpoint and timeout.
  • SPACY_MODEL_NAME: spaCy pipeline used for entity linking.
  • THREAD_ARCHIVE_IDLE_MINUTES: Minutes of inactivity before automatically archiving threads.
  • Thresholds: SENTIMENT_THRESHOLD, SARCASM_THRESHOLD, TOXICITY_THRESHOLD.

LLM Configuration (New)

  • LLM_MODEL: LLM model name or endpoint for sentiment/attitude/emotion/sarcasm analysis.
  • LLM_ENDPOINT: API endpoint URL for LLM inference service (if using external API).
  • CONFIDENCE_THRESHOLD: Minimum confidence score for LLM predictions (default 0.75).
  • SARCASM_THRESHOLD: Sarcasm probability threshold for labeling as sarcastic (default 0.5).
  • WEIGHTING_MODE: Mode for upvote weighting (linear, logarithmic, sqrt; default linear).
  • WEIGHT_CAP: Maximum upvote weight cap to prevent outlier dominance (default 200).

Multi-LLM Provider Configuration (Benchmarking)

OpenAI:

  • OPENAI_API_KEY: OpenAI API key
  • OPENAI_MODEL: Model name (default gpt-4o-mini)
  • OPENAI_ENDPOINT: API endpoint (default https://api.openai.com/v1/chat/completions)

Anthropic:

  • ANTHROPIC_API_KEY: Anthropic API key
  • ANTHROPIC_MODEL: Model name (default claude-3-5-sonnet-20241022)
  • ANTHROPIC_ENDPOINT: API endpoint (default https://api.anthropic.com/v1/messages)

Gemini:

  • GEMINI_API_KEY: Google Gemini API key
  • GEMINI_MODEL: Model name (default gemini-1.5-pro)
  • GEMINI_ENDPOINT: API endpoint (default https://generativelanguage.googleapis.com/v1beta/models)

Control Settings:

  • LLM_PROVIDERS: Comma-separated list of providers to use (default openai, options: openai,anthropic,gemini)
  • LLM_EVAL_MODE: Evaluation mode (default parallel, options: parallel, sequential)
  • LLM_BENCHMARK_SAMPLES: Fraction of comments to benchmark (default 0.25 for 25%, range 0.0-1.0)

LLM Provider Automation (Phase 2)

Active Provider:

  • PROVIDER_PREFERRED: Currently selected provider (set by automation, default openai)

Automation Schedules:

  • BENCHMARK_CRON_SCHEDULE: Cron schedule for quarterly benchmarking (default 0 3 1 */3 * - first day of quarter at 3 AM)
  • DRIFT_CHECK_SCHEDULE: Cron schedule for drift monitoring (default 0 0 * * 0 - Sunday midnight)
  • PROVIDER_SELECT_SCHEDULE: Cron schedule for automated provider selection (default 0 3 * * * - daily 3 AM)

Drift Detection:

  • DRIFT_SAMPLE_RATE: Fraction of comments to re-analyze for drift detection (default 0.05 for 5%)
  • DRIFT_AGREEMENT_THRESHOLD: Agreement threshold for drift alerts (default 0.8)

Cost Controls:

  • COST_ALERT_THRESHOLD: Monthly cost limit in USD before alerting (default 500)
  • ALERT_EMAIL: Email address for cost and drift alerts
  • SLACK_WEBHOOK_URL: Slack webhook for alerts (optional)

Hugging Face & OpenAI Tokens

  • HUGGINGFACE_ACCESS_TOKEN, HF_TOKEN, HUGGINGFACE_HUB_TOKEN: Access tokens forwarded to Hugging Face downloads (pipeline also honours HF_TOKEN for local CLI parity).
  • HF_HOME: Explicit cache directory for Hugging Face model artifacts (defaults to .hf_cache).
  • OPENAI_API_KEY: Token for optional OpenAI integrations.

Auth0

  • AUTH0_DOMAIN, AUTH0_CLIENT_ID, AUTH0_CLIENT_SECRET, AUTH0_AUDIENCE, AUTH0_ALGORITHM: Authentication configuration.

Integrations

  • SENDGRID_API_KEY, FROM_EMAIL, FROM_NAME: Transactional email.
  • SLACK_WEBHOOK_URL, SLACK_CHANNEL, SLACK_BOT_TOKEN: Slack alerts.

Monitoring

  • SENTRY_DSN, SENTRY_TRACES_SAMPLE_RATE: Error reporting.
  • DATADOG_API_KEY, DATADOG_APP_KEY, DATADOG_SERVICE, DATADOG_ENV: Metrics ingestion.
  • LOG_LEVEL, LOG_FORMAT: Structured logging.

Feature Flags

  • ENABLE_REAL_TIME_UPDATES, ENABLE_WEBSOCKET, ENABLE_ALERTS, ENABLE_BACKFILL, ENABLE_BRIGADE_DETECTION, ENABLE_BOT_DETECTION.

Limits & Retention

  • API_RATE_LIMIT, API_RATE_LIMIT_PERIOD: Request throttling.
  • MAX_CONCURRENT_THREADS: Max simultaneous Reddit threads to analyze.
  • RAW_COMMENT_RETENTION, AGGREGATE_RETENTION, CACHE_TTL_THREAD, CACHE_TTL_AGGREGATE: Retention policies.

Security

  • ALLOWED_UPLOAD_EXTENSIONS, MAX_UPLOAD_SIZE: Upload restraints.
  • CORS_MAX_AGE: CORS preflight cache lifetime.
  • SESSION_TIMEOUT: Web session lifetime in seconds.
  • AUTH0_DOMAIN, AUTH0_AUDIENCE, AUTH0_CLIENT_ID, AUTH0_ALGORITHMS: Backend Auth0 configuration (JWT issuer, API audience, optional custom algorithm list).

Testing & Tooling

  • PYTEST_WORKERS, COVERAGE_THRESHOLD: Testing defaults.
  • ENABLE_DEBUG_TOOLBAR, ENABLE_SWAGGER_UI, ENABLE_REDOC: Development tooling toggles.

Frontend

  • VITE_API_URL: Base REST API URL.
  • VITE_WS_URL: WebSocket URL.
  • VITE_AUTH0_DOMAIN, VITE_AUTH0_CLIENT_ID, VITE_AUTH0_AUDIENCE: Auth0 config shared with client.

Deployment

  • DEPLOYMENT_ENV, VERSION, BUILD_NUMBER, GIT_COMMIT: Build metadata.

Optional

  • WANDB_API_KEY, WANDB_PROJECT, WANDB_ENTITY: Weights & Biases.
  • K8S_NAMESPACE, K8S_SERVICE_NAME, K8S_REPLICA_COUNT: Kubernetes hints.
  • CUSTOM_CAST_DICTIONARY_URL, CUSTOM_SLANG_LEXICON_URL, TIMEZONE: Customization.