Quick Start¶

Get up and running with Sentinel DV in minutes.

Prerequisites¶

• Python 3.10 or higher
• Verification artifacts (UVM logs, cocotb results, coverage reports)
• 500MB+ disk space for indexing

Installation¶

pip install "sentinel-dv>=2.3.0"

git clone https://github.com/kiranreddi/sentinel-dv.git
cd sentinel-dv
pip install -e ".[dev]"

Configuration¶

Required: Sentinel DV does not start without a config file (no silent demo/ default).

From the repository root, copy the template and edit paths for your environment:

cp config.example.yaml config.yaml

Alternatively set export SENTINEL_DV_CONFIG=/absolute/path/to/config.yaml, or place config.yaml / config.yml in the directory where you start the server.

Example config.yaml:

artifact_roots:
  - /path/to/verification/regressions

index:
  type: duckdb
  path: ./sentinel_dv.db

adapters:
  uvm: true
  cocotb: true
  assertions: true
  coverage: true

security:
  max_page_size: 200
  max_response_bytes: 2097152

redaction:
  enabled: true
  redact_emails: true
  redact_paths: true

Index Your Artifacts¶

python -m sentinel_dv.indexing.indexer --config config.yaml --index-all

This will:

1. Scan configured artifact roots
2. Parse verification artifacts using enabled adapters
3. Build normalized index in DuckDB
4. Generate failure signatures and topology

Expected output:

[INFO] Scanning artifact roots...
[INFO] Found 150 test logs
[INFO] Parsing UVM logs...
[INFO] Parsing cocotb results...
[INFO] Building index...
[INFO] Indexed 150 tests, 45 failures, 1200 assertions, 80 coverage reports
[INFO] Index complete: sentinel_dv.db (125MB)

Start the Server¶

python -m sentinel_dv.server --config config.yaml

Server will start on stdio (MCP protocol):

Sentinel DV v2.3.0 started
Schema version: 1.0.0
Tools registered: 15
Index ready: 150 tests indexed

Use with Claude Desktop¶

Add to your Claude Desktop configuration (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "sentinel-dv": {
      "command": "python",
      "args": [
        "-m",
        "sentinel_dv.server",
        "--config",
        "/absolute/path/to/config.yaml"
      ]
    }
  }
}

Restart Claude Desktop.

First Queries¶

Try these queries with Claude:

"Why did test axi_burst_test fail in the latest run?"

"Compare functional coverage between runs R123 and R124"

"Show me all AXI assertion failures from this week"

"What's the pass rate for the nightly regression?"

"Show clk toggles for test_counter_sim between 2 and 3 microseconds"

Verify It Works¶

Check that tools are accessible:

# test_connection.py
from sentinel_dv.config import load_config
from sentinel_dv.indexing.store import get_store

config = load_config("config.yaml")
store = get_store()

print(f"Tests indexed: {store.count_tests()}")
print(f"Runs indexed: {store.count_runs()}")
print(f"Failures indexed: {store.count_failures()}")

Next Steps¶

• MCP tool gallery — visual cards for all 28 tools (real JSON from demo/)
• Multi-project demo — UVM, cocotb, Verilator, and commercial simulator fixtures
• VCS, Questa, and Cadence — license-free artifact validation (python scripts/verify_all_mcp_tools.py --sim vcs)
• Verilator + VCD example — end-to-end waveform demo
• Waveform summaries — *.wave.json and *.vcd indexing
• Configure adapters for your specific simulator
• Explore tools available for queries
• Understand schemas for structured data
• Deploy to production with systemd

Troubleshooting¶

Full Troubleshooting Guide