sonagram

Ask your AI assistant for a playlist from your own music collection — and get a file any music app can play.

sonagram listens to every track you own (tempo, energy, mood, key, how songs feel), organizes what it learns into a fast, searchable map of your library, and lets an AI agent translate your sentence into a typed request. Sonagram's deterministic library engine—not the agent—then selects, orders, audits, and explains the playlist:

"Make me a deep-focus work playlist." "A party mix for Saturday that builds." "Songs like Teardrop, but calmer."

Your music files are never modified, moved, or uploaded — everything happens on your machine.

For AI agents

You are an AI agent and a user asked you to work with their music library — start here:

  1. Install — pip install sonagram. Prebuilt wheels cover macOS (Apple Silicon + Intel), Linux x86_64, and Windows x86_64 on Python ≥ 3.9 — no toolchain needed. Only on other platforms does pip fall back to the sdist, which compiles the native core and needs a Rust toolchain (rustup).
  2. Install the skill — sonagram skill install. This writes the sonagram-playlist skill to ~/.claude/skills/ (pass --dir <skills_root> for a non-default skills location).
  3. Read the installed SKILL.md and follow it now. It covers library detection, scanning, the optional Last.fm walkthrough, and playlist delivery. A skill only auto-loads at the next session start, so act on it immediately in this one.

Use it through your agent (no code)

This is the main way to use sonagram. One-time setup:

  1. Install it — pip install sonagram. That's it on Mac, Windows, and regular Linux — prebuilt packages, nothing to compile. (On unusual platforms pip builds from source, which needs the free Rust toolchain — your agent can sort that out.)
  2. Point it at your music — sonagram sources add ~/Music (repeat for other folders).
  3. Give your agent the skill — sonagram skill install.

That's it. From now on, just ask:

  • "make me a deep-focus work playlist"
  • "a party mix for Saturday that builds"
  • "songs like Teardrop but calmer"
  • "which songs do I have multiple versions of? pair them"
  • "what's even in my library?"

The first request analyzes your whole collection (about an hour for ~10,000 songs — your agent will tell you and can let it run in the background). Every request after that takes seconds: sonagram notices what changed, re-reads only that, and keeps the map current — including songs you've added or deleted.

Each playlist is saved with its name, your original request, and the full track list, so you can always ask for it again later (sonagram playlists). The .m3u8 files point straight at your own music files and open in any player.

Optional: better picks with Last.fm. A free Last.fm API key adds richer genre info, song popularity, and "fans also like" connections. Your agent can walk you through getting one — just ask.

CLI (scriptable)

pip install sonagram also gives you the standalone sonagram command (the same shared code path the agent uses, so the two can't drift). Once a source is registered, commands need no path arguments:

sonagram sources add ~/Music     # register a library folder (repeatable)
sonagram status                  # is everything up to date? (exit 0/1/2)
sonagram scan                    # analyze new/changed files → local cache
sonagram enrich                  # optional: fold in Last.fm metadata (needs a key)
sonagram build                   # merge all sources → the central graph
sonagram profile --format json   # curation-relevant coverage/distributions
sonagram curate --preset focus --tracks 25 \
    --name "Deep Focus" --description "a calm work playlist" --format json
sonagram playlists               # list stored playlists (newest first)
sonagram mcp install             # native kglite manifest + revealed music skills

sonagram config shows where everything lives (defaults under ~/.sonagram/) and whether a Last.fm key is set up. Explicit-path forms (sonagram scan ~/Music, sonagram status ~/Music --format json, …) still work for scripting a single library without touching the config, and sonagram playlist ... --copy-to <dir> produces a portable folder — the tracks copied next to the playlist file, ready for a USB stick or another device.

Everything is incremental: a rescan of an unchanged library analyzes nothing and finishes in well under a second.

For building your own agents / integrations

Install Sonagram's safe kglite manifest/revealed skills, then use the absolute launch command it prints—or drive the same library contract from Python:

sonagram mcp install
# RUN: '/absolute/path/sonagram-mcp-server' --graph '/.../music.kgl'

The thin Sonagram frontend embeds KGLite 0.16.5's server and registers typed profile/policy/curate/audit/explain/store tools against its live graph. KGLite still owns MCP, Cypher, graph lifecycle, and generic tools; Sonagram owns only the music-domain handlers.

import sonagram
sonagram.scan("~/Music")
g = sonagram.build("~/Music", out_path="music.kgl")   # a live kglite graph
brief = {"preset": "focus", "target_tracks": 25,
         "target_duration_sec": None, "seed_ids": [],
         "seed_role": "pinned", "unsupported_intents": []}
result = sonagram.curate_playlist("music.kgl", brief)
assert result["exportable"] and result["audit"]["passed"]

Agents get a full manual (AGENT-GUIDE.md: the schema, a query cookbook, and a typed curation contract) plus live-gated kglite skills. Both route final selection/order/audit through the library rather than agent-authored heuristics.

How it works (the short version)

  • sonara does the listening: tempo, key, energy, mood, loudness, structure, and an audio "fingerprint of feel" for every track.
  • kglite stores the map: a graph database with search, similarity, and an agent-friendly query interface.
  • sonagram is the part in between: it decides what the map contains — every song with all its signals, connected to artists, genres, decades, moods, detected styles, and its 10 most similar tracks — and keeps the map exactly reproducible, byte for byte, no matter how often you rescan.

Full documentation (CLI reference, Python API, the graph's schema, and the engineering details) lives on Read the Docs — see docs/.

License: MIT

Project links

Data verified by PyPI on Aug 20, 2026
Data provided by the project maintainers, verified at the time the release was uploaded to PyPI.

Key dates

PyPI data
Data sourced directly from PyPI's database.
  • Released:
Latest release

1 maintainer

PyPI data
Data sourced directly from PyPI's database.
Avatar for kkollsg from gravatar.com kkollsg

GitHub Statistics

Data verified by PyPI on Aug 20, 2026
The GitHub source repository was provided by the project maintainers and verified by PyPI at the time of upload. Stars, forks, and open issues/PRs are derived from that repository and have not been independently verified.

License

MIT License

Requires

Python >=3.9

Tags

music knowledge-graph graph audio playlist cypher mir rust

Classifiers

Report project as malware

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

sonagram-0.2.5.tar.gz (929.0 kB view details)

Uploaded Source

Built Distributions

If you're not sure about the file name format, learn more about wheel file names.

sonagram-0.2.5-cp39-abi3-win_amd64.whl (10.7 MB view details)

Uploaded CPython 3.9+Windows x86-64

sonagram-0.2.5-cp39-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (11.0 MB view details)

Uploaded CPython 3.9+manylinux: glibc 2.17+ x86-64

sonagram-0.2.5-cp39-abi3-macosx_11_0_arm64.whl (9.8 MB view details)

Uploaded CPython 3.9+macOS 11.0+ ARM64

sonagram-0.2.5-cp39-abi3-macosx_10_12_x86_64.whl (10.7 MB view details)

Uploaded CPython 3.9+macOS 10.12+ x86-64

File details

Details for the file sonagram-0.2.5.tar.gz.

File metadata

  • Download URL: sonagram-0.2.5.tar.gz
  • Upload date:
  • Size: 929.0 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for sonagram-0.2.5.tar.gz
Algorithm Hash digest
SHA256 6a8084b0564cccb0cc25936112451b25ee0dbe378e6771082528345d02371e1f
MD5 26cfd48ab145fdffa6acd542d028f003
BLAKE2b-256 03ae07d566c3acf14406ecdda2f466b72f4081618c0c760504cc93d66bea31d0

See more details on using hashes here.

Provenance

The following attestation bundles were made for sonagram-0.2.5.tar.gz:

Publisher: release.yml on kkollsga/sonagram

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file sonagram-0.2.5-cp39-abi3-win_amd64.whl.

File metadata

  • Download URL: sonagram-0.2.5-cp39-abi3-win_amd64.whl
  • Upload date:
  • Size: 10.7 MB
  • Tags: CPython 3.9+, Windows x86-64
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for sonagram-0.2.5-cp39-abi3-win_amd64.whl
Algorithm Hash digest
SHA256 d028507db71f486bbcfdffbc334f989ed7cb700b0bc198623526e1e55822de56
MD5 c6a6fbfaab85a26174f695b7c7898b84
BLAKE2b-256 33c645f037cb9a162de195273fb5a19ef4947ca7f00192ebeaee01a9b9f3c841

See more details on using hashes here.

Provenance

The following attestation bundles were made for sonagram-0.2.5-cp39-abi3-win_amd64.whl:

Publisher: release.yml on kkollsga/sonagram

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file sonagram-0.2.5-cp39-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.

File metadata

File hashes

Hashes for sonagram-0.2.5-cp39-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
Algorithm Hash digest
SHA256 25a7557247df00b78c335436180874b944221c1b3c1cbedd35df024f059f6d3f
MD5 ad4dc8b5231a3d4f3190e7c094cbeea5
BLAKE2b-256 494298a7c23ff65b50ff0762612a7918ede5b1030f0a819715bad926b79c5bee

See more details on using hashes here.

Provenance

The following attestation bundles were made for sonagram-0.2.5-cp39-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl:

Publisher: release.yml on kkollsga/sonagram

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file sonagram-0.2.5-cp39-abi3-macosx_11_0_arm64.whl.

File metadata

File hashes

Hashes for sonagram-0.2.5-cp39-abi3-macosx_11_0_arm64.whl
Algorithm Hash digest
SHA256 8abb5b672d697330e71beeb197a8f68d1235bfea5cbe9a79131e40e7bd14d122
MD5 02fb5607451f810051fb8c32dfda7fd0
BLAKE2b-256 7fdc0a6c7c8a61aa1dcd3a546322e2690bac3ad0476f0c061e59783f349127c8

See more details on using hashes here.

Provenance

The following attestation bundles were made for sonagram-0.2.5-cp39-abi3-macosx_11_0_arm64.whl:

Publisher: release.yml on kkollsga/sonagram

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file sonagram-0.2.5-cp39-abi3-macosx_10_12_x86_64.whl.

File metadata

File hashes

Hashes for sonagram-0.2.5-cp39-abi3-macosx_10_12_x86_64.whl
Algorithm Hash digest
SHA256 11b8342ad81b1e595141167bcd19a1a8493948ffe416c82f246483d37e7ef15c
MD5 28895673accb3fbd0bf5455c9bbe6799
BLAKE2b-256 d422e954f79d906a144d73b7f893661bdd58d63d965cd9f16ce55cffdae19b19

See more details on using hashes here.

Provenance

The following attestation bundles were made for sonagram-0.2.5-cp39-abi3-macosx_10_12_x86_64.whl:

Publisher: release.yml on kkollsga/sonagram

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page