Prototype · Phase 1 · Static
Two manuals.
One patch bay.
A working sandbox for cross-referencing technical documentation — built here on Telviva Enswitch 4.2 and QueueMetrics 26.01, two systems that already talk to each other over Asterisk. The same mechanics scale to legal document sets: chunking, search, cross-referencing, and eventually retrieval-backed narrative synthesis.
Library
Every document loaded into this sandbox, chunked on ingest.
Search
Full-text, client-side, across both documents at once. This is real — no mock data.
Patch Bay
Terms that surface meaningfully in both manuals, patched together like a switchboard. Click a jack to route the connection.
Select a term above to see the matched passage from each manual, side by side.
Chunk Explorer
How each document was split for retrieval. Every segment below is a real chunk with page provenance — hover to inspect.
Hover or tap a segment to preview that chunk.
AI Insights
This is where retrieval meets generation. Phase 1 is a preview of the interaction model with realistic sample output — wiring it to live inference is a Phase 2 step (see below).
Ask across documents
Demo outputAsk a question that spans both manuals. In Phase 2 this retrieves the top matching chunks and asks Claude to answer strictly from them, with citations back to page numbers.
Story Mode
Demo outputTurns a cross-referenced cluster of chunks into a plain-language narrative — the same mechanic that, on a legal document set, would turn a folder of filings into a coherent case timeline.
Making this live
The panel above is intentionally static so this page is safe to host publicly with zero ongoing cost. To make it real, the browser needs to call a Cloudflare Worker (not the Claude API directly — direct browser calls would expose your API key and get blocked by CORS). The Worker holds the key server-side, retrieves the right chunks, and calls Claude on the page's behalf. That's the natural Phase 2 step once you're happy with this direction.