Happy Bull - Investment Research Platform

About this project
A personal, AI-first investment research workspace. The Company Research Desk, takes a company from its ticker to a dated, evidence-backed assessment you can question. Later on, it makes that research personal and keeps it current.
What it is for
Picking your own companies means reading filings, prices and news that are scattered, slow to read and easy to misjudge, while most tools answer with a score or an opinion that doesn't show its working. Happy Bull does an analyst's groundwork and shows every step. Code reads SEC filings and market prices and computes every figure, ratio and check, each with a receipt. An agent team then turns that evidence into three separate judgments, business quality, valuation and entry timing: a filings analyst and a news analyst search the company's own filings and releases, a writer drafts the judgments with every claim citing its evidence, and a critic argues the other side, while code verifies every citation and quoted number. Once switched on, the platform keeps that research current by itself as companies file, and each person's profile (their goal, horizon, holdings and how much detail they want) decides what they see and what a verdict means for them, never the judgments. It is for information and learning, not investment advice.
What it does
AI marks the features where a model does part of the work. Code computes every figure, ratio and score in all of them.
The company page: figures from SEC filings and market prices, each opening a receipt with its inputs and sources, in six views (For you, Price tag with a what-if lab, Cash, Business, Timing, Research), by fiscal year or quarterly TTM.

Research notes · AI: a filings analyst and a news analyst search the company's filings and releases, a writer drafts three separate judgments (business quality, valuation, entry timing) and a critic argues the other side; code checks every citation and quoted number.

Follow-ups and Refresh · AI: questions answered from what the note could see at its cutoff; Refresh makes a new dated note and shows what changed.
Research that keeps itself current · AI: once switched on, a worker reads SEC's filing feed every ten minutes, and a company people follow gets a new note when it files a periodic report.
Your profile and For you: a sample or a thirteen-card interview sets your goal, horizon and the weight of each judgment; every company and note then says whether it fits you, why and how much, read Short, Explained or Everything.
Portfolios, theses and watchlists · AI: holdings with the reasons you own them and what would make you sell; Happy Bull proposes what would test each reason and checks your thesis.
Personal runs · AI: your own deeper research on a note, on up to three focus areas, without changing its judgments.
Lessons: one term at a time, taught on the company in front of you, for a profile that wants to learn.
Alerts, the 07:00 briefing and email: your conditions, sharp falls, changed labels, thesis points and red-flag filings, in the app and by email, once switched on.
What moved: after a sharp fall, what changed since the note, judgment by judgment.
Costs before the call: each run shows its time and cost before it starts, under a monthly budget, and lists every model call after.
The command palette: Ctrl+K finds any company or command from any page.
Built with
SvelteKit
shadcn-svelte
LayerChart and Lightweight Charts
FastAPI
Pydantic
LangGraph
PostgreSQL 18 with pgvector
models and embeddings through OpenRouter
Data comes from SEC, Yahoo, Finnhub (logos) and the ECB (exchange rates).