reelflow-studio
ReelFlow Studio
A client's video publishing is fully automated, with no screen a person could use, so approving a video meant reaching into the machinery. This is that screen: one button per decision. Every button retries up to six times and counts a job done only when the automation confirms it, because those steps sometimes stall. The demo here runs that loop on a made-up queue, not the client's live system.
Built by our founder, Salman Adnan, for one of this agency's clients. Its source is private since it's a real client's operations tool. The linked demo replays the same approve, caption, publish, and schedule loop against a fabricated video queue, not the real client's workflows.
Overview
A single-file operator console for a set of n8n automation workflows: video approval and publishing, LinkedIn content generation, and a scripting agent. The browser talks directly to n8n webhooks, with retry logic so a flaky workflow run never silently drops an approval.
Key features
- Four tabs, each with its own start button: AI Videos, Hi-Tec Automations, Mr. Salman LinkedIn, and a Carousel sub-tab.
- index.html is everything: markup, CSS, and three IIFE script blocks, with state in module variables plus sessionStorage.
- A Netlify edge function proxies anyone-with-the-link Google Drive files so video and iframe embeds are same-origin, whitelisting MIME types and returning 409 when a file is not shared.
- Every call auto-retries: 6 attempts, a 120 s timeout per attempt (240 s for slow jobs), a 15 s gap; a call is completed only when the workflow responds with the literal string Completed.
- Two Python Pinecone loader scripts seed retrieval indexes, embedding with gemini-embedding-001 at 3072 dimensions.
Results
The four tabs call 33 distinct n8n endpoints from one base URL. End-to-end runs are not verified here, since they need a live n8n instance; the loaders were verified to start and fail cleanly.