Verification Intelligence™for Pre‑Silicon Teams
Overnight failures to a morning plan. dVI™ clusters your regressions and DeepDebug™ traces each failing test to its root cause — with on‑prem privacy and cost transparency.
Seeking 1–2 pilot partners
Run dVI™ on your latest regression. 4–6 week pilot with measurable outcomes and on‑prem deployment.
Quickstart
Up and running in minutes
Private-by-design. No code changes to your testbench. Works with your existing regressions.
Point to your nightly root; we incrementally scan and index artifacts without overloading NFS.
# install
pip install dvi
# index last night's regressions
dvi monitor --root /nfs/nightly_nov_4,_2025
Create structured JSON/JSONL datasets from logs and signals—ideal for audits or model evals.
# extract structured failure data
dvi curate --index .dvi/index Ask the DeepDebug agent to reason across logs, wave markers, and prior occurrences.
# analyze a specific test
dvi debug --test tb_core_random_42 Security & On‑Prem
- Private-by-design: process logs locally; ship nothing off your network.
- On‑prem deploys and air‑gapped environments supported.
- Audit‑ready JSON/JSONL artifacts; configurable retention.
Supported Simulators
Drop‑in with no testbench changes.
Verification‑ready features
What you get on day one
Drop into your existing regressions and get instant signal: AI triage, DeepDebug™ root‑cause, grounded chat, and live health — all on‑prem.
- DeepDebug™ Root‑Cause
- LLM‑powered agentic mode that hunts the true reason each test failed, gathers all relevant context, and lets you iterate via a modern chat interface.
- Instant AI Triage
- Overnight analysis produces clustered failure groups with executive summaries and prioritized action items.
- Real‑Time Monitoring
- Live dashboard tracks pass rate, failure deltas, and cost — optimized for regression workflows.
- Multi‑Simulator Logs
- Out‑of‑the‑box patterns for VCS, Xcelium, Questa, Verilator and more using AI‑assisted parsing.
- Conversational Debug
- Ask natural questions about current failures. Chats are grounded in your latest triage, DeepDebug™ insights, and logs.
- On‑Prem Privacy
- Everything runs inside your network. Zero data exposure and full IP protection.
Built for verification teams
- Leads: Track new vs existing failures. Prioritize by impact.
- Engineers: DeepDebug™ root‑causes tests and reduces log spelunking.
- Managers: Clear rollups by run, test, and owner to plan the week.
Ready to see it on your regressions?
We’ll connect to a sample run, cluster failures, and walk your team through the morning plan.
Talk to usWhy dVI™ over scripts + spreadsheets
dVI™ unifies incremental monitoring, DeepDebug™ root‑cause, AI parsing, and contextual chat into a single, reproducible pipeline your whole team can trust.
Advantages
- DeepDebug™ automatically surfaces the true root cause of each failure
- Incremental scan avoids reprocessing and NFS overloads
- Versioned triage snapshots with “latest” convenience link
- Grounded chat that already knows your regression context
- Bounded concurrency to keep EDA servers stable
- Cost and token telemetry per regression run
- Plain JSON artifacts for long‑term portability