AI · RAG Platform · 2024 — 2025
DevPulse — making 50,000 documents
feel like one searchable mind.
A retrieval-augmented platform for engineering teams: ingest scattered docs, threads, and PDFs into role-scoped workspaces — then ask questions and get citation-grounded answers. Built for recall, latency, and refusal — not demo-day magic.
Next.js · TypeScript · Hybrid Retrieval · LLM
4 workspaces preloaded
25 docs · 12 sessions
Try this once you're in
-
1
Pick a workspace
4 are preloaded: Backend Engineering · Frontend KB · DevOps · AI/ML Research.
-
2
Browse documents
Each workspace has ~6–9 docs in mixed states (indexed, processing, failed) — see how the UI handles each.
-
3
Open a chat session
12 multi-turn sessions are preloaded with realistic conversations across all workspaces.
-
4
Ask a new question
Citations are inlined with the answer — click a citation to jump to the source chunk.
Note for reviewers — this demo runs against a mock backend that mirrors the production RAG pipeline shape (Postgres + pgvector retrieval, hybrid scoring, JWT auth) but returns deterministic seeded responses so the demo stays snappy and predictable. Source on
GitHub ↗.