Datadog RCA Assistant (Embabel + DICE)

AI-assisted root cause analysis for production incidents using Embabel agents and DICE (Domain-Integrated Context Engineering) for memory and reasoning.

Quick links

What	Where
Demo videos (GitHub Pages)	Live site · workflow — set Settings → Pages → Source: GitHub Actions or deploy returns 404; committed recordings/*.mp4 only
Video bundle (TTS, text visuals, compose)	docs/documentation-generator/datadog-drilldown-video-docs/
Narrative script (architecture + D-Wave)	docs/VIDEO_NARRATIVE_ARCHITECTURE_AND_DWAVE.md
D-Wave Leap (token, venv, verify)	docs/DWAVE_LEAP_SETUP.md
QUBO spec & milestones	dwave.md · milestones/milestone-1.md, milestone-2.md
JVM ↔ Python bridge (ADR)	docs/adr/0002-dice-leap-subprocess-bridge.md
CI / branch protection	docs/CI_BRANCH_PROTECTION.md
QUBO e2e smoke	./scripts/qubo-e2e-smoke.sh · scripts/README.md

If you fork the repo, replace menkelabs/datadog-drilldown in the URLs above with your fork.

D-Wave & QUBO (optional)

Problems can be compiled to QUBO and solved locally or, optionally, on D-Wave Leap. Optional for core RCA; enriches ranking when enabled.

Topic	Doc / path
Leap API token & config	docs/DWAVE_LEAP_SETUP.md
Python PoC (instance → QUBO → solve)	dice-leap-poc/ · dwave.md
JVM bridge & RCA flag	embabel-dice-rca/README.md (QUBO section)
Telemetry / metrics	docs/QUBO_METRICS_V1.md · docs/DWAVE_REAL_WORLD_METRICS.md

Default path needs no Leap account — local classical solver in dice-leap-poc.

Architecture

• embabel-dice-rca — Analysis engine: Datadog telemetry, Embabel workflow, candidates, optional QUBO enrichment, DICE ingest.
• dice-server — Proposition extraction and reasoning over incident memory.
• test-report-server — REST + UI for test runs, solver JSONL, analysis (see test-report-server/README.md).

Modules (summary)

Module	Role
embabel-dice-rca	Datadog REST, logs/metrics/traces, Embabel, DICE, optional mock Datadog profiles
dice-server	Ingestion, LLM propositions, semantic query
test-report-server	H2 run history, UI, optional test triggers, solver run views
dice-leap-poc	Python: instance JSON → QUBO → local_classical or leap_hybrid

Setup

Prerequisites: Java 21+, Maven 3.8+; LLM key for agents; Datadog keys or Spring profile mock-datadog / mock-datadog-scenarios (see embabel-dice-rca/README.md).

Build & test (reactor root):

mvn test

Single module: cd embabel-dice-rca && mvn test (etc.). Aggregator: pom.xml.

Typical env:

export OPENAI_API_KEY="sk-..."
export DD_API_KEY="..."
export DD_APP_KEY="..."
export DD_SITE="datadoghq.com"

Run services:

cd dice-server && mvn spring-boot:run
cd embabel-dice-rca && mvn spring-boot:run

VS Code / IDE: Java 21 — .vscode/settings.json.

Testing

Kind	Command / pointer
Full Java reactor	mvn test (root)
DICE integration	cd embabel-dice-rca && mvn test -Dtest=SystemIntegrationTest (needs DICE_SERVER_URL, keys, …)
Test Report UI E2E	cd test-report-server && npm run e2e:ci (see e2e README)
QUBO Python + JVM	./scripts/qubo-e2e-smoke.sh

Tuning / H2 / analysis: test-report-server/README.md · embabel-dice-rca/docs/TEST_REPORT_ANALYSIS.md.

Documentation index

Doc	Topic
docs/documentation-generator/	Shared video-framework templates + datadog-drilldown-video-docs bundle
docs/DWAVE_LEAP_SETUP.md	Leap token & venv
dwave.md	DICE + QUBO + Leap architecture
docs/adr/0002-dice-leap-subprocess-bridge.md	Subprocess bridge
embabel-dice-rca/docs/architecture/README.md	Diagrams
docs/puml-setup.md	PlantUML

Optional MkDocs: if you add mkdocs.yml under docs/documentation-generator/, workflow docs-site will run mkdocs build --strict again; otherwise it skips with a notice.