Initialize Anthropic client. Args: config: AI configuration

Initialize OpenAI-compatible client. Args: config: AI configuration

Initialize Azure OpenAI client. Args: config: AI configuration

Initialize Gemini client. Args: config: AI configuration

Initialize RSS scraper. Args: sources: List of RSS feed configurations http_client: Shared async HTTP client

Initialize the scraper. Args: config: OpenBB source configuration. http_client: Shared httpx client (unused here; kep

Initialize the scraper. Args: config: GDELT source configuration. http_client: Shared async HTTP client.

Initialize scraper. Args: config: Scraper-specific configuration http_client: Shared async HTTP client

Initialize GitHub scraper. Args: sources: List of GitHub source configurations http_client: Shared async HTTP client

Initialize scraper. Args: config: OSS Insight source configuration http_client: Shared async HTTP client

Initialize the scraper. Args: config: Google News source configuration. http_client: Shared async HTTP client.

Generate daily summary. Args: items: Important items to include (already enriched with background/related) date: Date

Flatten matched domains into a deduplicated list of source configs. Args: matched_domains: Output from match_domains(). Returns:

Generate completion using Claude. Args: system: System prompt user: User prompt temperature: Sampling tem

Generate completion using Gemini. Args: system: System prompt user: User prompt temperature: Sampling tem

Fetch items from all enabled OpenBB watchlists. Args: since: Only return items published strictly after this time. Retur

Fetch content items published since the given time. Args: since: Only fetch items published after this time Returns:

Fetch GitHub content items. Args: since: Only fetch items published after this time Returns: List[ContentIte

Fetch trending repos for each configured language and apply filters.

Enrich filtered items into the enriched stage.

Fetch and deduplicate content into the raw stage.

Filter scored items into the filtered stage.

Generate a markdown summary from a stage.

Read run metadata.

Read items from a run stage.

Read a generated run summary.

List recent runs and stage states.

Run fetch -> score -> filter -> enrich -> summarize in one call.

Score a stage into the scored stage.

Send a webhook notification with the given variables. Uses the webhook URL (from environment variable), request_body template, and headers fr

Validate Horizon config and required environment variables.

List recent runs and stage availability.

Main entry point for the setup wizard.

Match user interest description against preset domains. Performs case-insensitive keyword matching against domain keywords and source tags. R

Print configuration template.

Effective default config resolved from local Horizon path.

In-memory metrics snapshot.

Run stage items resource.

Run metadata resource.

Run summary resource.

Recent run list.

Server metadata resource.

Reset all accumulated usage (useful for tests).

Resolve a tag or alias to its canonical form (lowercased).

Search HN + Reddit for each item concurrently. Returns {item.id: [related_stories]}. Deduplicates by URL against each item's own URL.

DingTalk returns HTTP 200 with errcode=400 in body — should be yellow warning.

Feishu returns HTTP 200 with code=19001 in body — should be yellow warning.

Non-JSON 2xx response body prints as green (no error code check possible).

Slack returns HTTP 200 with ok=false — should be yellow warning.