Music Sample Extractor Pro

I want to extract the latest samples from the website tracklib and grab samples from the most recent 5 songs, based on a rough description of the samples I want. I enter a brief description, the agent visits tracklib and then grabs my samples. To do this we will need to crawl https://www.tracklib.com/music/new-arrivals and extract, say, the latest list of 20 songs, then an LLM will filter them based on my description and find the 5 best matches on youtube, download them and chop them up into samples. we will want to extract stems (instrumentals, vocals), and then extract (say) 16 beats worth of audio - beginning from start of the song. We will need to analyse beatpoints and tempo, extract the stems, and then chop the sampes. this will then give us 5x songs x2 (instrumental and vocal) chops = 10 files to chop. then we could potentially extract the spectrograph image of each, analyze it to find the best place to chop, and then cut out samples from those points. finally it will return the audio samples chopped up in zip files.

First, pre-entered info...

final-output2

The final output is a single ZIP file containing the 10 audio samples (two samples—instrumental and vocal—from each of the five chosen songs). Each sample is 16 beats in length, neatly chopped at the beat boundaries.

subagents-extracted2

OVERVIEW OF AGENTS: Broad description of the agents required, and what each will do to generate the required component SUBAGENT 1: ScraperAgent • Purpose: Retrieve the 20 most recent Tracklib arrivals (title, artist, link, etc.). • Skill Used: #226 - Extract Structured Data From 1x URL • Process: 1) Call #226 with "https://www.tracklib.com/music/new-arrivals" and request the 20 newest songs. 2) Parse the extracted data into a structured list/dict. 3) Output the 20-song list to be passed along. SUBAGENT FINAL OUTPUT: [20-new-arrivals] SUBAGENT 2: FilterAgent • Purpose: Select the top five songs based on a user's descriptive query. • Skill Used: #223 - Powerful LLM Prompt-to-Text Response • Process: 1) Provide the 20-song list plus the user's description to #223. 2) Instruct the LLM to rank songs by relevance, returning the top five. 3) Output those five songs (title, artist, link) for the next steps. SUBAGENT FINAL OUTPUT: [5-filtered-songs] SUBAGENT 3: DownloaderAgent • Purpose: Obtain MP3 audio for the five selected songs. • Skills Used: 1) #224 - Oracle Ask A Question (to find YouTube URLs if necessary) 2) #193 - Download Entire YouTube Video As MP3 • Process: 1) For each selected song, optionally use #224 to find the best YouTube link. 2) Call #193 on that YouTube link to download the MP3 audio. 3) Return the five MP3 file URLs. SUBAGENT FINAL OUTPUT: [5-song-mp3-audio] SUBAGENT 4: StemExtractorAgent • Purpose: Separate each MP3 into vocal and instrumental stems. • Skill Used: #197 - Extract Instrumental & Vocal From MP3 • Process: 1) For each of the five MP3s, call #197. 2) Collect the resulting 10 MP3 URLs (vocal + instrumental each). 3) Output those 10 stems for beat analysis. SUBAGENT FINAL OUTPUT: [10-separated-stems] SUBAGENT 5: TempoAnalyzerAgent • Purpose: Identify tempo (BPM) and beat markers for each stem. • Skill Used: #180 - Extract Beatpoints & Tempo of MP3 • Process: 1) For each of the 10 stems (MP3), call #180. 2) Gather the BPM and beat marker data (e.g., array of timestamps). 3) Output that tempo data. SUBAGENT FINAL OUTPUT: [beats-and-tempo-data] SUBAGENT 6: ChopAndPackAgent • Purpose: Chop each stem into a 16-beat segment and package everything into a final ZIP. • Skills Used: 1) #219 - Cut Wav/mp3 Audio into Multiple Pieces/Samples (for exact trimming) 2) REQUIRED (Zip Creation) - No existing skill to bundle into a ZIP • Process: 1) Compute duration for 16 beats using tempo. 2) For each stem, call #219 to extract that 16-beat segment. 3) After obtaining the 10 final clips, use a new or external zip function (REQUIRED) to create "final_samples.zip." 4) Return the ZIP file as the final output. SUBAGENT FINAL OUTPUT: [final-zip-of-samples]

subagentnumber

[subagentnumber]

Then, agent response...

subagent1refined-new-json-core

G) TASKFLOW SUMMARY 1. Website Screenshot Task: - ID: #174 - Title: Capture Tracklib New Arrivals Page Screenshot - Description: Takes a full webpage screenshot of Tracklib's new arrivals page, ensuring all content is captured for analysis, including dynamically loaded content - Input(s): [tracklib_url] - Output token name: [page_screenshot] 2. Visual Analysis Task: - ID: #176 - Title: Analyze Tracklib Page Structure via GPT Vision - Description: Uses GPT Vision to analyze the screenshot and identify key page elements, content structure, and locations of relevant song information for efficient data extraction - Input(s): [page_screenshot] - Output token name: [page_structure_data] 3. Data Extraction Task: - ID: #226 - Title: Extract Structured Song Data from Tracklib - Description: Scrapes the webpage using identified structure to extract detailed song information including titles, artists, release dates, and metadata, organizing into structured format - Input(s): [tracklib_url, page_structure_data] - Output token name: [raw_song_data] 4. Data Enrichment Task: - ID: #224 - Title: Enrich Song Data via Oracle Queries - Description: Queries additional information for each song by visiting individual song URLs, filling in missing metadata like BPM, key, and price tier information - Input(s): [raw_song_data, song_urls] - Output token name: [enriched_song_data] Final Output Integration: The [enriched_song_data] is formatted into the specified JSON structure containing all 20 songs with complete metadata, forming the final output token [tracklib_songs_json] which is passed to the next subagent in the workflow. Chain Flow: [tracklib_url] → #174 → [page_screenshot] → #176 → [page_structure_data] → #226 → [raw_song_data] → #224 → [enriched_song_data] → [tracklib_songs_json]

subagent1mermaid

<!DOCTYPE html> <html> <head> <title>Mermaid Diagram Example</title> <script src="https://cdnjs.cloudflare.com/ajax/libs/mermaid/9.4.3/mermaid.min.js"></script> </head> <body> <div class="mermaid"> graph TD A[Start: tracklib_url] --> B[Task #174: Screenshot] B --> C[page_screenshot] C --> D[Task #176: GPT Vision Analysis] D --> E[page_structure_data] E --> F[Task #226: Data Extraction] A --> F F --> G[raw_song_data] G --> H[Task #224: Data Enrichment] H --> I[enriched_song_data] I --> J[Final: tracklib_songs_json] subgraph ScraperAgent B D F H end subgraph Inputs/Outputs A C E G I J end </div> <script> mermaid.initialize({ startOnLoad: true }); </script> </body> </html>

Mermaid Image - https://static.aiz.ac/1735395370-mermaid/mermaid1.png

prompts-required

[prompts-required]

subagent1-json-array

[ { "taskID": 174, "semanticTitleOfTask": "Capture Full Screenshot of Tracklib New Arrivals Page", "taskDescription": "Takes a complete webpage screenshot of Tracklib's new arrivals page, ensuring all dynamically loaded content is captured", "inputDescription": "The base URL for Tracklib's new arrivals page", "inputRequired": ["tracklib_url"], "outputDescription": "A PNG format screenshot of the entire webpage, including all song entries and metadata", "outputName": "page_screenshot", "promptInstruction": "" }, { "taskID": 176, "semanticTitleOfTask": "Analyze Page Structure Using GPT Vision", "taskDescription": "Analyzes the screenshot to identify page layout and element structure for efficient data extraction", "inputDescription": "The full webpage screenshot from the previous task", "inputRequired": ["page_screenshot"], "outputDescription": "A detailed technical analysis of the page structure, including element locations and HTML patterns", "outputName": "page_structure_data", "promptInstruction": "You are analyzing a screenshot of Tracklib's new arrivals page. Please identify and describe: 1. The layout pattern for individual song entries 2. The exact location and format of: - Song titles - Artist names - Release dates - Genre/style information - Sample category/price tier - Duration - BPM/Key information (if visible) 3. Any pagination elements or infinite scroll mechanisms 4. The HTML structure or class names visible for these elements Format your response as a structured technical analysis that can be used for precise data extraction." }, { "taskID": 226, "semanticTitleOfTask": "Extract Structured Song Data", "taskDescription": "Scrapes the webpage using the identified structure to extract detailed information for 20 most recent songs", "inputDescription": "Original Tracklib URL and the page structure analysis from GPT Vision", "inputRequired": ["tracklib_url", "page_structure_data"], "outputDescription": "Raw structured JSON data containing 20 songs with available metadata fields", "outputName": "raw_song_data", "promptInstruction": "Using the provided page structure analysis, extract data from Tracklib's new arrivals page. For each of the 20 most recent songs, capture: 1. Title (exactly as shown) 2. Artist name (primary artist only) 3. Release date (in ISO format) 4. Genre/style (as labeled) 5. Sample category/price tier 6. Direct Tracklib URL 7. Duration (in seconds) 8. BPM (if shown) 9. Key (if shown) Return the data in a clean JSON structure, with each song as a separate object containing all available fields. Handle missing data gracefully by using null values." }, { "taskID": 224, "semanticTitleOfTask": "Enrich Song Metadata", "taskDescription": "Queries additional information for songs by visiting individual URLs to complete missing metadata", "inputDescription": "The raw song data JSON and list of individual song URLs that need additional metadata", "inputRequired": ["raw_song_data", "song_urls"], "outputDescription": "Complete JSON array of 20 songs with fully enriched metadata, including any previously missing information", "outputName": "tracklib_songs_json", "promptInstruction": "For each song where metadata is incomplete: 1. Visit the individual song URL 2. Locate and extract any missing information from: - BPM - Musical key - Price tier - Additional genre tags - Sample restrictions - Release information Focus only on filling gaps in the existing data. Do not overwrite already captured information. Return only the new, supplementary data for each song that needed enrichment." } ]

subagent1-json-array-mermaid

<!DOCTYPE html> <html> <head> <title>Mermaid Diagram Example</title> <script src="https://cdnjs.cloudflare.com/ajax/libs/mermaid/9.4.3/mermaid.min.js"></script> </head> <body> <div class="mermaid"> flowchart TD A[Start] --> B[Base URL Input: tracklib_url] B --> C[Task #174: Take Screenshot] C --> D[Output: page_screenshot] D --> E[Task #176: GPT Vision Analysis] E --> F[Output: page_structure_data] F --> G[Task #226: Extract Song Data] G --> H[Output: raw_song_data] H --> I[Task #224: Enrich Metadata] I --> J[Output: tracklib_songs_json] J --> K[End: Pass to Next Subagent] subgraph Screenshot Phase B --> C --> D end subgraph Analysis Phase D --> E --> F end subgraph Data Extraction Phase F --> G --> H end subgraph Enrichment Phase H --> I --> J end </div> <script> mermaid.initialize({ startOnLoad: true }); </script> </body> </html>