What Shall We Build Next?

Describe

Describe your task

Refine

Refine the plan

SubAgents

Review all agents

Deploy

Deploy your agent

A) SUBAGENT SUMMARY  
ScrapeNewArrivals gathers up to 10 track entries (title, artist, link) from the Tracklib "New Arrivals" page and consolidates them into a single structured output for later use.

B) FINAL TASK OUTPUT  
[tracklib-new-arrivals] – a single textual or JSON-formatted list of up to 10 recent tracks, including title, artist, and link.

C) SUBAGENT INPUT  
{variable1} – The user’s initial description of the tracks they want to sample (note: not strictly needed to perform the scraping, but included for completeness).

E) SUBAGENT TASK SUMMARY  
1. Use Skill #226 (Extract Structured Data From 1x URL)  
   • INPUT:  
       – URL: "https://www.tracklib.com/music/new-arrivals"  
       – Instruction: "Extract up to 10 of the most recent tracks, including their title, artist, and detail/link if available."  
   • OUTPUT: Raw extracted textual data (e.g. partial HTML or structured text).

2. Use Skill #223 (Powerful LLM Prompt-to-Text Response)  
   • INPUT:  
       – The raw extracted data from step 1  
       – (Optionally) {variable1} (the user description) for context if needed  
       – Prompt: "Summarize the extracted data into up to 10 entries in JSON form, each with {title, artist, link}."  
   • OUTPUT: A concise final text/JSON structure representing the 10 (or fewer) track entries.

3. Subagent Final Output:  
   • The structured list of tracks (title, artist, link), labeled “[tracklib-new-arrivals].”

F) SILOS  
Since this subagent only requires scraping and then summarizing, there is a single silo:  
• Silo 1: {Scrape & Summarize} – Steps 1 and 2 (Skill #226 → Skill #223)  
No further silos are necessary.

SubAgent #1 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
"FilterTracks" reads the user’s music-style description and the up-to-10 track metadata from Subagent 1, ranks and filters them, and returns the top 3 best matches.

B) FINAL TASK OUTPUT  
• [filtered-tracks]: A single textual/structured list (e.g. JSON) containing exactly three chosen tracks (title, artist, link).

C) SUBAGENT INPUT  
• {variable1} (the user’s initial description of the tracks they want to sample)  
• [tracklib-new-arrivals] (the up-to-10 track entries from Subagent 1)

E) SUBAGENT TASK SUMMARY  
1) Input: {variable1} and [tracklib-new-arrivals]  
2) Action Step (Skill #223: Powerful LLM Prompt-to-Text Response)  
   • Prompt the LLM to interpret {variable1} (the user’s style/genre description) alongside the list in [tracklib-new-arrivals].  
   • Rank the 10 tracks by their match to {variable1}, then select the best three.  
   • Return a JSON/text structure of these three entries, each with “title,” “artist,” and “link.”  
3) Output: [filtered-tracks]

F) SILOS  
• This subagent is straightforward (only one skill call). No additional silos required.

SubAgent #2 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
This subagent will take three filtered Tracklib entries (title, artist), locate each track on YouTube, download an MP3 of each, and finally produce a single textual/structured output listing the three local MP3 files.

B) FINAL TASK OUTPUT  
[downloaded-mp3-files] → A single textual or JSON-style reference that clearly indicates the file paths/URLs of the three MP3s (e.g. a list of three MP3 URLs).

C) SUBAGENT INPUT  
[filtered-tracks] (i.e., the list of three chosen tracks from Subagent 2, containing title, artist, link, etc.)

E) SUBAGENT TASK SUMMARY

1) Read in the three chosen tracks from [filtered-tracks].  
2) For each track (repeated three times):  
   • Use #224 (Oracle Ask A Question) to find the best official or close-match YouTube video link for the given track (title/artist).  
   • Optionally use #223 (Powerful LLM Prompt-to-Text Response) if needed to parse out the direct YouTube URL from the Oracle’s textual answer.  
   • Pass that YouTube URL to #193 (Download Entire YouTube Video As MP3) to download the audio as an MP3.  
3) Summarize the three resulting MP3 file paths into a single textual/structured output (e.g. JSON) using #223 (Powerful LLM Prompt-to-Text Response).

F) SILOS

–––––––– SILO 1 (Track #1) ––––––––  
 • Input: The first track’s info: Title, Artist  
 • Action 1: #224 (Oracle Ask A Question)  
   – Prompt: “Find me the best or official YouTube link for ‘{Artist} - {Title}’. Provide the final direct YouTube link in your answer.”  
   – Output: Detailed text from Oracle (likely including or referencing a YouTube URL)  
 • Action 2 (optional if needed): #223 (Powerful LLM Prompt-to-Text Response)  
   – Prompt: “Extract the direct YouTube link from the text: [oracle answer]. Return only the link.”  
   – Output: Direct YouTube link  
 • Action 3: #193 (Download Entire YouTube Video As MP3)  
   – Input: Direct YouTube link  
   – Output: MP3 URL for track #1

–––––––– SILO 2 (Track #2) ––––––––  
 • Same steps as SILO 1 (Oracle → parse link → download #193)  
 • Output: MP3 URL for track #2

–––––––– SILO 3 (Track #3) ––––––––  
 • Same steps as SILO 1 (Oracle → parse link → download #193)  
 • Output: MP3 URL for track #3

–––––––– SILO 4 (Combine Results) ––––––––  
 • Action: #223 (Powerful LLM Prompt-to-Text Response)  
   – Prompt: “Combine these three MP3 URLs into a single textual output or JSON list. Label them as ‘mp3_file_1’, ‘mp3_file_2’, ‘mp3_file_3’.”  
   – Output: Final single textual/JSON structure with the three MP3 references → [downloaded-mp3-files]

SubAgent #3 - Diagram

Expand Flow

A) SUBAGENT SUMMARY:  
Analyze each of the three MP3 files to detect tempo and beatpoints, then cut and return the first 16-beat snippet (in WAV format) for each track.

B) FINAL TASK OUTPUT:  
[wav-snippets] → A single textual/structured listing of three WAV snippet file URLs (each containing 16 beats).

C) SUBAGENT INPUT:  
[downloaded-mp3-files] → A structured reference (e.g., list or JSON) with three MP3 URLs.

E) SUBAGENT TASK SUMMARY (Step-by-Step Flow):

1) FOR EACH OF THE 3 MP3 FILES:  
   a) USE SKILL #180 - Extract Beatpoints & Tempo of MP3  
      • INPUT: The MP3 URL for the track.  
      • OUTPUT: A list (or text) of beatpoints and tempo (the exact timestamps of each beat).

b) PARSE THE 16TH BEAT TIME  
      • The subagent will examine the result from skill #180 to locate the time (in seconds) of the 16th beat.

c) USE SKILL #219 - Cut Wav/mp3 Audio into Multiple Pieces/Samples  
      • INPUT: The same original MP3 plus the start/end times (0 seconds to the 16th-beat timestamp).  
      • OUTPUT: A single WAV snippet (URL) containing the extracted audio from 0 until the 16th beat.

2) CONSOLIDATE THE THREE WAVE SNIPPETS  
   a) USE SKILL #223 - Powerful LLM Prompt-to-Text Response  
      • INPUT: The three WAV snippet URLs.  
      • OUTPUT: A concise textual/structured summary referencing all three WAV snippet URLs (e.g., JSON or text list).

3) RETURN THE FINAL OUTPUT  
   • Subagent returns [wav-snippets] → The textual/structured listing of the three WAV snippet file URLs.

F) SILOS (IF APPLICABLE):  
• SILO A: Beat Detection & End Time Calculation (Steps 1a & 1b for each MP3).  
• SILO B: Audio Cutting to WAV (Step 1c for each MP3).  
• SILO C: Summarization / Consolidation (Step 2).