Private beta — joining now

Your team's knowledge — searchable, with receipts.

Ragesta is a self-hosted retrieval-augmented generation system for engineering teams. Index your codebase, docs, and tickets; ask questions in plain English; get answers grounded in real source material — every claim cited, traceable, deterministic.

What makes it different

Four ways Ragesta isn't another vector-search demo.

🏠

Self-hosted, end-to-end

Indexes, embeddings, and answers all stay on your hardware. Your codebase never leaves your perimeter, your queries never train someone else's model. One binary, your control.

📜

Every answer cited

No hallucinated facts dressed in confident prose. Each sentence in an answer points back to the file, line, ticket, or doc page it came from. Click a citation, see the source.

⚙️

Deterministic indexing

Same corpus, same query, same answer — every time. Indexing is content-addressed; updates are diff-based; retrieval is reproducible. No "ask twice and get two stories."

{ }

Code-aware retrieval

Knows the difference between a function and a comment, an import and a string literal. Treats commit messages, PR descriptions, and inline docs as distinct signal — not just more text.

ragesta · live query
Why does the upload gateway sometimes block .tar.gz files larger than 100 MB?
The upload gateway enforces a 100 MB hard cap on archive files, scanning them with ClamAV before any extraction. Files over the limit are rejected with HTTP 413 and an incident is written to the security log.
📎 packages/server/src/plugins/upload-gateway/index.ts:142–168
📎 docs/security-engine-runbook.md § "Upload limits"
📎 PR #2917 — "Cap archive scans at 100 MB"

Get on the beta waitlist.

We're rolling Ragesta out to a small group of engineering teams first. Drop a note and we'll send you access when your slot comes up.