Privacy Policy

DHIMIX AI is a music recommendation platform that uses AI scoring to suggest tracks matched to your mood, energy level, and genre preferences. It does not require an account to use.

Music is streamed from two sources:

• Jamendo — a catalog of Creative Commons–licensed tracks from independent artists around the world.
• Cloud Showcase — a small, curated demonstration collection hosted on Cloudflare R2 object storage.

All personalisation — your profile, Music DNA, favorites, and listening history — is computed and stored in your browser. DHIMIX AI does not operate a backend that stores personal user data.

DHIMIX AI uses localStorage — a standard browser storage mechanism — to persist your experience across sessions. The following data is stored locally on your device only and is never transmitted to our servers:

Preferences and profile. Your selected genre, mood, energy level, and scoring mode are saved so you do not need to re-enter them on each visit. Stored under the key dhimix-app.

Recommendation history. The results of your recommendation sessions — track titles, artists, confidence scores, and explanations — are stored locally to power the Insights panel. This data is not transmitted to our servers.

Favorites. When you tap the heart button on a track, the following information is written to localStorage under the key dhimix-feedback:

• Track ID
• Title and artist name
• Catalog source (Jamendo or Cloud Showcase)
• Artwork URL
• Timestamp of when you saved it

Favorites exist only in your browser. They are not synced to any server. Clearing browser data removes them permanently.

Listening history. Each time a track successfully begins playing, a record is written to localStorage under the key dhimix-history containing:

• Track ID
• Title and artist name
• Catalog source
• Artwork URL
• Playback timestamp

History is capped at 100 entries. No audio content, playback duration, or position is recorded. History is not transmitted to our servers.

Music DNA. Your Music DNA is a lightweight preference vector computed from your feedback (likes, favorites, skips). It is used only to re-rank recommendations locally and is stored in your browser.

Volume and UI settings. Player volume, theme selection, and display preferences are stored locally so your setup persists between visits.

When you request recommendations, your preference profile (genre, mood, energy, scoring mode) is sent to the DHIMIX AI recommendation API — a server-side endpoint on the same host as this application. The API uses this data to fetch and score matching tracks from Jamendo and the Cloud Showcase, then returns the results to your browser.

The server does not log your preferences to a database, associate them with your identity, or share them with any third party. The request is stateless — each recommendation session is independent.

DHIMIX AI fetches track metadata and stream URLs from the Jamendo public API using a read-only client ID. No personal data about you is sent to Jamendo.

When you stream a Jamendo track, your browser connects directly to Jamendo's storage servers to receive the audio. Jamendo's own Privacy Policy applies to that connection. DHIMIX AI does not proxy or cache Jamendo audio.

Jamendo tracks are licensed under Creative Commons. Attribution requirements and commercial use terms vary by track and are displayed in the License Center.

The Cloud Showcase is a small, curated collection of audio files hosted on Cloudflare R2. It is a demonstration catalog — not a full commercial music library. When you stream a Cloud Showcase track, your browser requests a short-lived presigned URL from the DHIMIX AI server and then streams the audio directly from Cloudflare.

Cloudflare may collect connection metadata (such as IP address and request headers) as part of their standard CDN operation, subject to Cloudflare's Privacy Policy. DHIMIX AI does not log which tracks you stream.

License metadata for Cloud Showcase tracks is derived from filenames and a local manifest. Each track's license status is visible in the License Center. Tracks with an "Unknown" license status are used for demonstration purposes and have not yet been formally reviewed.

Spotify (optional). If you connect a Spotify account, DHIMIX AI receives your display name, profile image, and top listening data via Spotify's OAuth flow. This is stored in an encrypted server-side session for the duration of your visit and is not written to a persistent database. You can disconnect Spotify at any time from the Profile page, which immediately clears the session. Spotify's own Privacy Policy governs data on their side.

PostHog (optional). If a PostHog analytics key is configured in the deployment, anonymised usage events — such as recommendation requests and session duration — may be sent to PostHog. No personally identifiable information is included. See PostHog's Privacy Policy for details. If PostHog is not configured, no analytics data is collected.

DHIMIX AI sets one session cookie — dhimix_session — when you connect Spotify. It is HttpOnly, Secure, and SameSite=Lax. It is used only to maintain your Spotify authentication state and is removed when you sign out. No advertising or cross-site tracking cookies are used.

Because all personalisation data lives in your browser, you are in full control:

• Clear favorites, history, and DNA individually from Settings → Data.
• Clear everything at once by clearing your browser's localStorage via its built-in privacy or developer settings.
• Disconnect Spotify from the Profile page to remove the session cookie.

DHIMIX AI does not retain server-side copies of your local data. Once cleared, it cannot be recovered.

DHIMIX AI is not directed at children under 13 and does not knowingly collect data from children.

If this policy changes materially, the "Last updated" date at the top of this page will be revised. We will not introduce changes that expand data collection beyond what is described here without clearly communicating that change. Continued use of DHIMIX AI after a policy update indicates acceptance of the revised terms.

For privacy questions or data requests, reach us at:

System type. DHIMIX AI operates as a rule-based deterministic scoring system. Recommendations are produced by a fixed, human-authored formula that weights eight musical features — genre, mood, energy, acousticness, popularity, decade, tempo, and instrumentalness — against your stated preferences. There is no neural network, no probabilistic model, and no automated learning from user behaviour running on the server.

No black-box inference. Every recommendation result includes a Why[] explanation block that lists the exact scoring reasons and their contribution weights. The output you see is the direct and complete product of the formula. There are no hidden signals, no undisclosed training data, and no opaque embeddings influencing the result.

Human oversight. Because scoring logic is expressed as readable, version- controlled code, any developer or auditor can inspect and verify the complete decision path for any recommendation. No specialised AI tools are required to audit the system.

No high-risk application. DHIMIX AI does not make decisions affecting employment, credit, education, healthcare, law enforcement, or any other domain listed in Annex III of the EU AI Act. It is a music discovery tool operating in a low-risk, non-consequential context.

User agency. You can inspect your active scoring weights at any time via the Preference Panel, adjust them freely, and immediately see how changes affect results. The system has no ability to override or ignore your stated preferences.

Music DNA (client-side only). The optional Music DNA feature re-ranks results using a lightweight preference vector computed from your feedback (likes, favorites, skips). This computation runs entirely in your browser. No learning model is trained on your data on any server.

Contact for AI-related queries. Questions about the recommendation system, scoring methodology, or this compliance statement may be directed to dhimixai@gmail.com.