What Shall We Build Next?

Describe

Describe your task

Refine

Refine the plan

SubAgents

Review all agents

Deploy

Deploy your agent

A) SUBAGENT SUMMARY: This subagent converts an MP4 video file to an MP3 audio file.

B) FINAL TASK OUTPUT: The final output will be an MP3 audio file extracted from the input MP4 video file. The MP3 will have the same duration as the original video.

C) SUBAGENT INPUT: The subagent requires the URL of the MP4 video file hosted on the internal server as its initial input.

E) SUBAGENT TASK SUMMARY: The subagent will complete the following sequence of skills to convert the MP4 to MP3:

Subagent Input (MP4 URL) > Skill #241 - Convert mp4 to mp3 > Subagent Output (MP3 URL)

Detailed breakdown:
1. The subagent receives the internal URL of the MP4 video file as its input. 
2. It passes this MP4 URL to Skill #241 - Convert mp4 to mp3.
3. Skill #241 extracts the audio from the MP4 file and saves it as an MP3 file on the server.
4. The URL of the newly created MP3 file is returned as the final output of the subagent.

F) SILOS: This subagent does not require any silos as it only involves a single skill to complete the MP4 to MP3 conversion task.

SubAgent #1 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY: The Transcription Subagent generates a timestamped transcription from an input MP3 file.

B) FINAL TASK OUTPUT: A JSON file containing the timestamped transcription of the input MP3 file. The JSON should have the following structure:
[
  {
    "text": "Transcribed text for section 1",
    "start": "00:00:00",
    "end": "00:01:23"
  },
  {
    "text": "Transcribed text for section 2", 
    "start": "00:01:24",
    "end": "00:02:57"
  },
  ...
]

C) SUBAGENT INPUT: The URL of the MP3 file to be transcribed.

E) SUBAGENT TASK SUMMARY:
1. Receive the MP3 URL as input
2. Call skill #188 (Get Transcription Of MP3 (With Timings)) with the MP3 URL as input
3. Retrieve the timestamped transcription output from skill #188
4. Format the transcription into the specified JSON structure
5. Save the JSON file to the server
6. Return the URL of the saved JSON file as the final output

Detailed flow:
Input MP3 URL > #188 Get Transcription Of MP3 (With Timings) > Process transcription into JSON format > Save JSON file > Output JSON file URL

F) SILOS: This subagent does not require silos as it consists of a single linear sequence of actions.

SubAgent #2 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY: The Section Breakdown Subagent analyzes a timestamped video transcription and identifies six key sections, providing a title, summary, and timestamp for each section.

B) FINAL TASK OUTPUT: A JSON file containing an array of six objects, each representing a key section of the video. Each object should include a title (string), summary (string), and timestamp (number in seconds) for the corresponding section.

C) SUBAGENT INPUT:
- A JSON file containing the timestamped transcription of the video, where each line is an object with the following properties:
  - text (string): The transcribed text for the line
  - start (number): The start timestamp of the line in seconds
  - end (number): The end timestamp of the line in seconds

E) SUBAGENT TASK SUMMARY:
1. Extract the transcription text from the input JSON file (#236 - Extract content of a file)
2. Divide the transcription text into six parts of roughly equal length (#176 - Divide Text (Or Code) Into 4x Parts)
3. For each part:
   a. Analyze the text using an LLM to identify the main topic or step discussed in the section (#178 - Rewrite Text Based On Instructions)
   b. Generate a concise title for the section based on the identified topic/step (#178 - Rewrite Text Based On Instructions)
   c. Generate a brief summary of the section's content (#178 - Rewrite Text Based On Instructions)
   d. Determine the timestamp for the start of the section based on the first line's start time in the corresponding part of the transcription
4. Combine the generated titles, summaries, and timestamps for all six sections into a single JSON file (#180 - Rewrite JSON (or other code) Based On Instructions)

F) SILOS:
This subagent does not require silos as the tasks are sequential and interdependent.

SubAgent #3 - Diagram

Expand Flow

A) SUBAGENT SUMMARY: The Transcription Refinement Subagent refines the raw transcription using an LLM, dividing it into six sections based on the provided section breakdown.

B) FINAL TASK OUTPUT: A refined transcription divided into six sections, saved as a single TXT file.

C) SUBAGENT INPUT:
1. Timestamped transcription JSON file from the Transcription Subagent
2. JSON file outlining the six key sections from the Section Breakdown Subagent

D) SUBAGENT TASK SUMMARY:
1. Extract the raw transcription text from the timestamped transcription JSON file.
2. Use Skill #178 (Rewrite Text Based On Instructions) to refine the raw transcription text, with instructions to improve clarity, grammar, and coherence.
3. Extract the section titles, timestamps, and summaries from the section breakdown JSON file.
4. Use Skill #178 (Rewrite Text Based On Instructions) to divide the refined transcription into six sections based on the provided section breakdown, with instructions to:
   a. Insert the section titles as headings
   b. Ensure the text content aligns with the section summaries
   c. Maintain the overall flow and coherence of the transcription
5. Save the refined and divided transcription as a single TXT file.

E) SILOS:
This subagent does not require silos as the tasks are sequential and interdependent.

Subagent Flow:
1. Timestamped transcription JSON file > Extract raw transcription text
2. Raw transcription text > Skill #178 (Rewrite Text Based On Instructions) > Refined transcription text
3. Section breakdown JSON file > Extract section titles, timestamps, and summaries
4. Refined transcription text + Section titles, timestamps, and summaries > Skill #178 (Rewrite Text Based On Instructions) > Refined and divided transcription text
5. Refined and divided transcription text > Save as TXT file > Subagent final output (refined-transcription.txt)

4 Template & Links

Expand Flow

A) SUBAGENT SUMMARY: Extract thumbnail images from the video at the timestamps specified in the Section Breakdown JSON.

B) FINAL TASK OUTPUT: Six thumbnail images (JPEG) corresponding to the start of each key section, saved on the server with both internal and external URLs provided.

C) SUBAGENT INPUT:
1. MP4 URL from the Video Processing Subagent
2. JSON file outlining the six key sections from the Section Breakdown Subagent, including the timestamp for each section

E) SUBAGENT TASK SUMMARY:
1. Input MP4 URL and JSON file with section breakdowns and timestamps > 
2. Skill #202 - Extract Thumbnail Images Of MP4 Video (repeated 6 times, once for each timestamp) >
3. Output six thumbnail images (JPEG) saved on the server, with both internal and external URLs provided

Detailed breakdown:
1. The Thumbnail Extraction Subagent receives the MP4 URL from the Video Processing Subagent and the JSON file containing the six key sections and their timestamps from the Section Breakdown Subagent.

2. For each of the six timestamps:
   a. Call Skill #202 - Extract Thumbnail Images Of MP4 Video, providing the MP4 URL and the specific timestamp.
   b. Save the resulting thumbnail image (JPEG) on the server.

3. After all six thumbnail images have been extracted and saved, output a list of the six thumbnail images, including both their internal and external URLs.

F) SILOS: This subagent does not require silos as it focuses on a single repeated task (extracting thumbnail images at specific timestamps) using the same skill (#202) for each timestamp. The iterations of the skill can be performed sequentially within the same subagent.

5 Template & Links

Expand Flow

Here is the response for the "Article Assembly Subagent":

A) SUBAGENT SUMMARY: The Article Assembly Subagent takes a refined transcription, section breakdown, and thumbnail images to assemble a well-formatted HTML article.

B) FINAL TASK OUTPUT: A single HTML file containing a well-structured article with the transcription divided into six sections, each with a relevant thumbnail image. The HTML should have proper formatting for readability and visual appeal.

C) SUBAGENT INPUT: 
- Refined transcription TXT file from the Transcription Refinement Subagent 
- JSON file outlining the six key sections from the Section Breakdown Subagent
- Six thumbnail images (JPEG) from the Thumbnail Extraction Subagent

E) SUBAGENT TASK SUMMARY:
1. Input the refined transcription TXT file, section breakdown JSON file, and six thumbnail image URLs
2. Use skill #236 to extract the content of the refined transcription TXT file 
3. Use skill #236 to extract the content of the section breakdown JSON file
4. Use skill #178 to rewrite the transcription text based on the section breakdown, inserting the section titles and summaries in the appropriate locations
5. Use skill #227 to format the rewritten text into pretty HTML, specifying where to insert the thumbnail images
6. Output the final HTML article file URL

F) SILOS: This subagent does not require silos as it is a straightforward sequential process.

So in summary, the detailed flow would be:

Refined transcription TXT file URL, section breakdown JSON file URL, 6 thumbnail image URLs 
> #236 Extract refined transcription text
> #236 Extract section breakdown JSON 
> #178 Rewrite transcription text based on section breakdown
> #227 Format rewritten text into HTML, inserting thumbnail images
> Output final HTML article file URL