§ Documentation

Data sources

Where the data comes from.

Every Sentinellis report is assembled from a small set of public and licensed data providers. We don't scrape, we don't resell raw feeds, and we don't hide our sources. Here's the full list.

§ I

Our data principles

Every claim in a report carries an inline citation.
We prefer T1 sources (Reuters, Bloomberg, AP) when stories overlap.
When a source is unavailable, we ship the report with a flagged gap rather than guessing.
We never sell, share, or aggregate user data with these providers.

§ II

Active providers

Yahoo Finance

Accessed via yfinance Python package

Free

Primary financial data

What it provides

·Ticker resolution (company name → symbol)
·Real-time quote: market cap, P/E, dividend yield
·Income statement (revenue, EBIT, net income, EPS)
·Balance sheet (assets, liabilities, equity)
·Cash flow statement (operating, investing, financing)
·Company officers (CEO, CFO, key VPs + total comp)
·Sector, industry, employee count, website (used for logo)

Freshness

Updated within minutes for quotes; quarterly for statements (after company filings).

Gaps

Sparse for non-US tickers, very small caps, and companies not listed on major exchanges. Returns HTTP 500 occasionally on individual fields — handled by fail-soft.

Visit provider

NewsData.io

Accessed via REST API

Paid (free tier on MVP)

Primary news source

What it provides

·Articles from 60,000+ sources globally
·7-day recent window (free tier)
·Source tier metadata (T1 Reuters/Bloomberg, T2 CNBC/Forbes)
·Country + language filters

Freshness

Articles indexed within hours of publication. We cap to a 7-day window for relevance.

Gaps

Free-tier limit of 10 results per query. When NewsData is unreachable or returns 0 articles for a niche company, we fall back to Google News RSS.

Visit provider

Google News RSS

Accessed via Public RSS endpoint

Free fallback

News fallback when NewsData fails

What it provides

·Same major sources (Reuters, CNBC, MarketWatch, etc.)
·Headline + publication date + source
·URL to original article

Freshness

Real-time aggregator. Same recency window enforced post-fetch.

Gaps

Description fields contain HTML anchor tags rather than article snippets — we strip them and fall back to the headline. No relevance metadata, so all stages 2-6 of our pipeline still run.

Visit provider

Anthropic Claude

Accessed via Anthropic API

Paid (per-token)

AI synthesis & relevance scoring

What it provides

·Claude Sonnet 4.5 → final report synthesis with inline citations
·Claude Haiku → news article relevance scoring (1–10)
·Claude Haiku → per-article plain-English summaries
·Strict JSON schema enforcement on all outputs

Freshness

On-demand. Each report run triggers fresh inference — no AI output is cached separately from the report.

Gaps

Subject to upstream rate limits and content policy. On hard errors, the report ships fail-soft with an "analysis unavailable" banner instead of a missing section.

Visit provider

Google Favicons

Accessed via Public endpoint

Free

Company logos

What it provides

·128px favicon resolved from company website domain
·Used in report headers and OG share images

Freshness

Fetched at report time, cached implicitly with the report.

Gaps

Some companies have generic placeholder favicons. We previously used Clearbit Logo API but it was deprecated; Google Favicons is a near-perfect free replacement (a public company's favicon is almost always their logo).

Visit provider

§ III

What we don't use (yet)

SEC EDGAR direct ingestion

10-K and 10-Q filings flow through Yahoo Finance today. Direct EDGAR integration is on the roadmap — it would unlock richer segment data and longer historical series.

OpenCorporates / national registries

Legal-entity ground truth for private companies (ONRC for RO, Companies House for UK, KvK for NL). Designed but not yet wired into the pipeline. Priority for the B2B tier.

LinkedIn / Glassdoor scraping

We don't scrape. Executive data comes from Yahoo Finance (which sources from SEC DEF 14A filings for US-listed companies). LinkedIn ToS prohibits automated access without authorization, and we respect that.

How it works Methodology