Environment variables

Process-level settings come from DISCOVER_* env vars (pydantic Settings, app/settings.py). Provider and channel secrets are referenced by name in the YAML (${VAR}) — values live only here or in .env.

DISCOVER_* settings

Env var	Default	What it does
DISCOVER_CONFIG_PATH	config/profiles.yaml	Path to the profiles YAML
DISCOVER_ENV_FILE	.env	Which dotenv file local runs load
DISCOVER_DATABASE_URL	sqlite:///./data/discover.db	SQLAlchemy URL (/data/discover.db in Docker)
DISCOVER_SCHEDULER_ENABLED	true	Master switch for all APScheduler jobs
DISCOVER_LOG_LEVEL	info	Python logging level
DISCOVER_FETCH_TIMEOUT_SECONDS	10	Feed-fetch HTTP timeout
DISCOVER_REFRESH_JITTER_SECONDS	60	Spread first feed fetches and per-source jitter
DISCOVER_CONTACT_EMAIL	unset	Appended to the fetch User-Agent; SEC feeds expect it
DISCOVER_SOURCE_AUTO_DISABLE_FAILURES	10	Consecutive fetch failures before a source is disabled (0 = never)
DISCOVER_API_KEYS	unset	Comma-separated read keys for /api/v1 + legacy reads (auth)
DISCOVER_ADMIN_KEY	unset	Guards /api/admin/*, topic writes, Settings (/login)
DISCOVER_PIPELINE_INTERVAL_SECONDS	60	How often extract + cluster runs (only new/pending articles)
DISCOVER_EXTRACT_MAX_ARTICLES	50	Articles fetched for text per pipeline run
DISCOVER_EXTRACT_TIMEOUT_SECONDS	10	Article-page fetch timeout
DISCOVER_EXTRACT_MAX_BYTES	2097152	Page size cap (2 MB)
DISCOVER_EXTRACT_MIN_WORDS	80	Fewer extracted words = text_status failed
DISCOVER_EXTRACT_DOMAIN_DELAY_SECONDS	2	Politeness delay between requests to one domain
DISCOVER_IMAGES_DIR	data/images	Card-image cache (/data/images in Docker)
DISCOVER_NO_IMAGE_STYLE	unset	Overrides images.no_image_style (logo/text_only/gradient)
DISCOVER_IMAGES_ALLOW_HOSTS	unset	Comma-separated hosts exempt from the image SSRF guard
DISCOVER_OUTLETS_PATH	config/outlets.yaml	Outlet registry (names/countries/kinds/groups) used for coverage labels
DISCOVER_CLUSTER_THRESHOLD	0.45	Min cosine similarity to join a story
DISCOVER_CLUSTER_WINDOW_HOURS	72	Article must be within this of the story's first/latest member
DISCOVER_CLUSTER_SAME_SOURCE_THRESHOLD	0.7	Stricter threshold when the story already has that source
DISCOVER_CLUSTER_MIN_SHARED_TOKENS	2	Significant tokens shared with the story title
DISCOVER_SUMMARIZE_ENABLED	true	Turn brief/report generation on/off
DISCOVER_SUMMARIZE_PROFILE	first profile	Whose llm.summarizer endpoint the worker uses
DISCOVER_SUMMARIZE_INTERVAL_SECONDS	60	Summary worker tick
DISCOVER_ENRICH_INTERVAL_SECONDS	60	Scheduled enrichment sweep: admits eligible stories, starts the worker for anything waiting
DISCOVER_ENRICH_WATCHDOG_SECONDS	120	Restarts the enrichment worker thread if it is not alive while stories wait
DISCOVER_SUMMARIZE_MAX_PER_RUN	25	Stories processed per worker tick
DISCOVER_SUMMARIZE_RESUMMARIZE_MIN_MINUTES	30	Min spacing before a story may regenerate
DISCOVER_SUMMARIZE_MAX_ARTICLES_PER_STORY	5	Members sent to a brief prompt
DISCOVER_SUMMARIZE_MAX_WORDS_PER_ARTICLE	1200	Words per member sent to a brief prompt
DISCOVER_SUMMARIZE_FAILURE_LIMIT	3	Consecutive endpoint failures before the worker pauses
DISCOVER_SUMMARIZE_PAUSE_MINUTES	10	Pause length after the failure limit
DISCOVER_REPORT_MODE	unset	Overrides every profile's summaries.report_mode

Other variables (.env.example)

Referenced from profiles.yaml as ${NAME} — only the name is stored.

Env var	Used for
OPENAI_API_KEY	Cloud provider key (market-monitor example)
LOCAL_LLM_BASE_URL, LOCAL_LLM_API_KEY, LOCAL_LLM_MODEL	OpenAI-compatible endpoint for llm.* roles
GUARDIAN_API_KEY, NEWSDATA_API_KEY, X_BEARER_TOKEN	API ingestion sources (api_key: on a source)
TINYFISH_API_KEY, BRAVE_API_KEY, EXA_API_KEY	Web-search providers (api_key: on a provider)
DISCOVER_TIMEZONE, DISCOVER_PUBLIC_URL	Alert timezone / base URL for links in notifications
NTFY_SERVER, NTFY_TOPIC, NTFY_TOKEN	ntfy channel
ALERT_WEBHOOK_URL, ALERT_WEBHOOK_SECRET	Webhook channel + HMAC key
SMTP_HOST, SMTP_USER, SMTP_PASSWORD, ALERT_EMAIL_FROM, ALERT_EMAIL_TO	Email channel
LLM_API_KEY_<NAME>	Per-endpoint key for python -m app.summarize.compare

Missing ${ENV} values for alert channels and sources are logged once at startup and listed (names only) on the Alerts page and Settings > Pipeline.