How I Built Source-Backed Expert-Video Search on Cloudflare

Description and use

Base2026 uses Cloudflare Workers, D1, and R2 to build a deterministic video search system that delivers verified excerpts and source links while keeping raw media private.

Problem

Discovering short‑form video clips is easy, but remembering who said what, comparing statements, and providing verifiable sources is hard.

Outcome

Base2026 delivers deterministic, reproducible evidence‑based answers from practitioner videos, exposing only reviewed excerpts and source links while keeping raw media private.

Workflow

• 1. Public discovery request hits Cloudflare Worker
• 2. Worker routes to private D1/R2 intake for deduplication
• 3. Queued capture and AI transcription via Workers AI
• 4. Evidence validation and review state management
• 5. Policy‑bound excerpt projection
• 6. Public D1 FTS5 index updated
• 7. Worker serves static assets or public D1 data
• 8. Visitor accesses stable evidence page or original source

Evidence

The public release excludes: raw media, raw captions, and raw ASR; full private transcripts and private review packets; credentials, logs, client material, and lead data; unreviewed or policy‑ineligible model output.

Source

Verified August 29 public readback recorded: 2,175 public search documents; 1,574 distinct source videos; 50 dynamic evidence routes; 0 full transcripts published.

Source

The public core is free to use and read‑only. I do not describe the entire operation as cost‑free: infrastructure, processing, review, and maintenance still consume resources.

Source