How Repo Dive works

Repo Dive is a daily code-learning platform by DanixSoft. Every day it recommends one open-source repository matched to what you want to learn, explains why it picked that one, tells you which files to read first and in what order, and asks three questions about it. What you write down is saved to a searchable archive and shapes what you are recommended next.

Yes. There is no paid tier and no card required. That includes the daily recommendations, the reading plans, unlimited interest profiles, the archive, exclusions, the learning record and the email summaries. Putting a paywall in front of reading open source would defeat the point of it.

A deterministic scoring model, not an AI. Every candidate from the GitHub Search API is scored on eight signals — relevance to your profile, project health, documentation quality, maturity, readability, novelty against what you have already studied, diversity against the areas you usually read, and difficulty fit — and one is picked from the top of the ranking. The weights are published, and every dive shows the reasons that actually decided it.

Two OAuth scopes. The `repo` scope lets Repo Dive create one private repository on your account and commit snapshots of what you studied into it — GitHub offers no narrower scope that can write to a private repository, and that scope also grants access to your other private repositories. Repo Dive never reads them: it only ever calls the GitHub search endpoints, public repository endpoints, and the git endpoints for the single archive repository it maintains for you. The `user:email` scope is read-only and is used to send your daily dive and weekly summary. Your password is never seen or stored, and you can revoke the grant from your GitHub settings or disconnect from the Preferences page at any time.

No. Discovery only ever searches public, open-source repositories, and the only repository Repo Dive writes to is the single private archive it creates for you. Be aware, though, that the `repo` scope GitHub issues is broad enough to permit more than that — it is the narrowest scope that can create a private repository. If that is not a trade you want to make, you should not connect the account.

In a private repository called `daily-code-explorer` on your own GitHub account, one dated folder per repository studied, each with a SOURCE.md recording the exact commit you were pointed at, the licence and the reading plan. Nothing is ever deleted from it. Snapshots are partial by design — text files only, capped in size and count — because the archive is for reading, not for mirroring.

A day counts when you complete a dive: open it and write something, whether that is an answer to one of the questions or a note of your own. Scheduled jobs running, emails arriving and dashboard visits count for nothing. Days are counted in your own timezone, and a streak anchored on yesterday stays alive through today — so you are not shown a zero at nine in the morning.

The run ends and a new one starts whenever you next complete a dive. Your total dives, notes, archive and learning record are all untouched — the streak is the only thing that resets, and it is deliberately the least important number on the page. If it was a single day, you can repair it: two repairs a month, covering one missed day immediately before a run that is still going. They reset monthly, cannot be bought, and are never advertised with a countdown.

Yes, and the star range and how recently a project must have been pushed to. You can keep several interest profiles and switch between them. Note that a very narrow profile limits how much variety the recommendations can offer, since the engine can only choose from what your filters return.

Yes. Add either a full `owner/repository` or a bare `owner` to your exclusion list. An owner entry is a hard filter — nothing that account publishes will ever be selected — and repositories you have already been shown are excluded automatically.

Per weekday. Set zero on the days you know you will not read and more on the days you will, up to twenty. Where a day asks for several, they are spread across your waking hours rather than committed in one burst at an arbitrary hour of your night.

Pause on your dashboard stops all scheduled dives immediately and keeps everything else intact. Daily dive emails, weekly summaries and reminder emails are three separate switches in Preferences. Pausing also stops reminder emails, because asking to be left alone should mean it.

If GitHub rejects the token — because you revoked the grant, or it expired — Repo Dive marks the connection as needing attention, stops the scheduled job for your account rather than retrying against a dead credential, and shows a reconnect prompt on your dashboard. Nothing in your history, notes or archive is affected.

Every dive shows the repository's licence, resolved to its SPDX identifier where GitHub can match one, and says plainly whether it is permissive, copyleft, unrecognised or absent. Where there is no licence, the dive says so: publicly visible does not automatically mean reusable, and you should check the repository's own licensing information before reusing code. Repo Dive describes the licence; it does not give legal advice.

GitHub tokens are encrypted with AES-256-GCM before storage, in a table that no client-facing database role can read. Every table enforces row-level security scoped to the owning account, and the columns a browser session may write are restricted to actual preferences — streaks and history are server-owned and cannot be edited from the client. Tokens and note text are never logged and never sent to analytics.

Yes, from the Preferences page, without asking anyone. It removes your profile, every dive, your notes, your learning record and the stored GitHub token. You confirm by typing your GitHub username, because a year of notes should not be one misclick from gone. Your archive repository lives on your own GitHub account and is deliberately not touched — it is yours, and deleting a repository is not an authority Repo Dive should hold.

No, and the design deliberately works against it. Streaks count days you completed a dive, not days a job ran, so an untouched archive earns nothing. The commits are real and are yours, but the thing being measured is whether you read the code.