As developers, we ship features every day, but summarizing technical Pull Requests or Release notes into engaging LinkedIn posts is often an afterthought or a manual chore.
To solve this, I built Git-To-Feed—an open-source Go microservice that listens to GitHub webhooks, processes technical context via Gemini AI, and requires a human approval tap in Discord before publishing to LinkedIn.
Here is how the event-driven pipeline flows:
flowchart TD
GH["GitHub Webhook"] -->|"HMAC Verified"| WH["Webhook Handler"]
WH --> PIPE["Pipeline Gemini AI"]
PIPE --> BOT["Discord Bot"]
BOT -->|"Approve / Reject"| APP["Approval Engine"]
APP -->|"Published"| LI["LinkedIn API"]
Package-by-Feature Layout: Domain logic (internal/approval, internal/pipeline, internal/publisher) is cleanly separated from infrastructure adapters (internal/platform).
HMAC Signature Verification: Webhooks are validated before processing payload memory.
Anti-Cringe & Few-Shot Prompting: Structured AI templates convert technical jargon into clear release announcements without corporate fluff.
Human-in-the-Loop: Discord WebSocket bot renders interactive action buttons (Approve / Reject) so no post goes live without dev review.
The repository is live on GitHub under the MIT license!
We have open issues labeled good first issue and help wanted for CI/CD setup, multi-channel routing, and test coverage. PRs and star support are highly appreciated!