Design partner program

Enterprise onboarding for production RAG debugging — maturity assessment, pilot qualification, and scoped remediation priorities.

Book technical walkthroughApply for pilotRequest architecture review

Apply for pilot

Complete the checklist, stack selection, and maturity assessment for your readiness score.

Onboarding checklist
Stack
Incident maturity assessment (0–4 per category)

Architecture

Operational debugging platform layers.

  1. Operational RAG debugging API (symptom, root cause, remediation, trust)
  2. RAG failure debugger (paste logs/traces → evidence-grounded analysis)
  3. Investigation workflow (timeline, share links, compare)
  4. Private incident + design-partner stores (server-side)
  5. Citation evidence index (allowlisted sources, no raw customer logs in public index)
┌─────────────────────────────────────────┐
│  Enterprise UI (debugger, investigations) │
├─────────────────────────────────────────┤
│  Operational APIs + export (Slack/Jira)   │
├─────────────────────────────────────────┤
│  Private stores (incidents, investigations)│
├─────────────────────────────────────────┤
│  Citation evidence (allowlisted corpus)   │
└─────────────────────────────────────────┘

Deployment models

Hosted

Fastest pilot — shared operational intelligence layer with private incident stores.

+ Low setup time · Managed updates · Standard export workflows

− Shared control plane · Data residency by region

VPC

Dedicated environment in your cloud with private networking and customer-managed keys.

+ Network isolation · CMK option · Audit log export to your SIEM

− Higher setup effort · Customer-operated upgrades

On-prem

Air-gapped or regulated deployments with offline operational corpus options.

+ Maximum control · Custom retention · No egress by default

− Longest integration timeline · Customer-operated HA

Supported stacks

LangChainLlamaIndexLangGraphQdrantpgvectorPineconeWeaviateOpenSearchother

Ingestion boundaries

  • Public citation corpus is allowlisted expert + documentation — not your proprietary documents unless ingested to a private workspace.
  • Operational incident and investigation stores are server-side JSON — never sitemap-listed or public API.
  • Debugger paste inputs are not used for model training.
  • VPC/on-prem pilots can disable shared multi-tenant corpus paths entirely.

Security FAQ

Do you train on our incidents?
No. Submissions stay in your private store. Anonymized patterns require explicit consent.
What is the pilot SLA draft?
99.5% API availability target during pilot, 4h Sev-2 response, weekly operational review — formal SLA in enterprise agreement.
How is data deleted?
Deletion requests processed within 30 business days for incident, investigation, and application stores.

Pilot process timeline

  1. Week 0Architecture review

    Stack map, deployment model, readiness assessment.

  2. Week 1–2Debugger + investigations

    Wire save flow, exports, eval gate design.

  3. Week 3–4Incident workflow

    Slack/Jira/postmortem templates, remediation tracking.

  4. Week 5–6Pilot review

    Readiness rescoring, rollout or VPC hardening plan.

Enterprise SLA draft: 99.5% pilot availability target, 4h Sev-2 response — formal terms in agreement.