Architecture

System Overview

The Thrive ecosystem is built around a simple idea:

Have one central source of truth for sermon and church content, and publish it out to every place people engage.

To make that work, we use a small set of focused components:

Diagram

        Attenders                          Staff
            |                               |
            v                               v
   +----------------+              +-----------------+
   |  Mobile App    |              |   Admin Tool    |
   +--------+-------+              +--------+--------+
            \                           /
             \                         /
              \       HTTPS          /
               v                     v
                 +---------------------+
                 |      Global API     |
                 | (sermons, config)   |
                 +----+-----------+----+
                      |           |
                data  |           | audio
                      v           v
                +-----------+  +--------+
                | Database  |  |  S3    |
                +-----------+  +--------+
                      |
                      v
            +------------------------+
            | AI Processing Pipeline |
            | (summary, 90+ tags,    |
            |  waveform, podcast)    |
            +------------------------+
                      |
                      v
                Podcast Platforms

High‑Level Flow

  1. Sermon is recorded
    Audio is edited and prepared by the production team.

  2. Staff upload the sermon
    Using the Admin Tool, staff upload audio and fill in basic details (title, date, series, speaker, etc.).
    The Global API securely stores this information and manages links to the audio file.

  3. AI enhances the content The AI Processing Pipeline:
    • Generates a TLDR-style summary (130-180 words) with straightforward, educational tone
    • Assigns topical tags from 90+ categories (e.g., hope, marriage, anxiety, finances)
    • Prepares waveform data for audio player visualization
    • Generates podcast descriptions written for non-church audiences
    • Creates series-level summaries when a series is complete
  4. Content goes live everywhere
    • The Mobile App and Public Website call the Global API to:
      • List sermon series
      • Show individual messages
      • Display summaries, tags, and artwork
    • Podcast platforms use the generated podcast feed to stay in sync.
  5. Attenders engage
    • In the Mobile App, people can:
      • Browse and search sermons
      • Play audio (online or downloaded)
      • View summaries and tags
      • Access “Connect”, “Give”, and other church resources
    • On the Website, people see the same core content, plus marketing and visitor‑focused pages.

Why This Design?

This architecture is intentionally:

Example Weekly Workflow

Here is a typical rhythm for a normal week:

  1. Sunday – The sermon is preached and recorded.
  2. Early week (Mon–Tue) – Audio is edited and exported by the production team.
  3. Mid‑week – A staff member uses the Admin Tool to:
    • Upload the audio file
    • Attach the message to the right series
    • Enter title, date, speaker, and passage
  4. Shortly after upload – The AI Processing Pipeline runs automatically to:
    • Transcribe the message using Azure Speech-to-Text
    • Generate a TLDR-style summary and assign tags from 90+ categories
    • Prepare waveform data and generate a podcast description for non-church audiences
  5. Same day – The Mobile App and Website automatically show the new message via the Global API; podcast apps receive the updated feed.

From the team’s perspective, the main manual task is one upload in the Admin Tool. Everything else fans out automatically.

Key Technologies (High‑Level)

Without going deep into implementation details, the system uses:

The rest of this documentation looks at each part of the system in more detail.

This site is open source. Improve this page.