Inference and agents, in the KingdomPay per token · Saudi RiyalDCP-Agent for Saudi business · agents.dcp.saAgents can rent a GPU · npx -y github:dhnpmp-tech/dcp-mcpEarn Riyal from your GPUPDPL · Saudi data residencyInference and agents, in the KingdomPay per token · Saudi RiyalDCP-Agent for Saudi business · agents.dcp.saAgents can rent a GPU · npx -y github:dhnpmp-tech/dcp-mcpEarn Riyal from your GPUPDPL · Saudi data residency
DCP
Sign inStart free →
DCP
Explore
01Overviewنظرة عامةSovereign Arabic AI runtime02InferenceالاستدلالOpenAI-compatible · live catalog03Fine-Tuningالضبط الدقيقLoRA contracts · proof-gated serving04BatchالدُفعاتJSONL validation · execution gated05DeploymentsالنشرEndpoint proof · traffic gated06BenchmarksالقياساتMeasured rows · quality claims gated07PricingالأسعارPer-million-token · SAR08GPU Podsحاويات GPURent a whole GPU on demand09AgentsالوكلاءZero-human onboarding · MCP10DocsالتوثيقOpenAI-compatible API11EarnاكسبEarn Riyal from your GPU12SupportالدعمTalk to the team
In-Kingdom · PDPL© 2026 · Riyadh

Dedicated endpoints for custom models and LoRA adapters, only after load proof.

DCP's dedicated-deployment rail connects Pods, Fine-Tuning, and Inference: create an adapter, request a deployment intent, prove the serving endpoint loaded the right artifact, smoke it with hashed response evidence, then route billed traffic. Today the intent, load-proof, endpoint-smoke, usage, and billing contracts are visible; public traffic remains gated.

Inspect deployment intents ->Use live inference
Adapter readiness contractsGET /api/adapters/*/readiness

Loading adapter deployment readiness...

deployment_intent

Deployment intent rows

Renter-owned adapter deployment records exist before traffic is routed.

Statuscontract visible
load_proof

vLLM load proof

Endpoint proof must match deployment id, adapter id, base model, mode, and artifact checksum.

Statuscontract visible
endpoint_smoke

Endpoint smoke readiness

A funded deterministic smoke must prove response hash, latency, token totals, and adapter trace before route or billing claims.

Statusgated
route_traffic

Route traffic gate

Traffic stays off until the backend records matching proof for the deployment, endpoint, and artifact.

Statusgated
usage_attribution

Usage attribution readiness

Usage rows stay disabled until they can prove deployment, adapter, endpoint, checksum, provider, request, scoped-key, token, cost, and pending settlement fields.

Statuscontract visible
billing_readiness

Billing readiness

Adapter billing stays off until endpoint smoke, funded principal, usage attribution, and settlement policy are approved.

Statuscontract visible
multi_lora

Multi-LoRA later

Live merge and multi-LoRA are product targets, not public serving claims until controlled smoke proof exists.

Statuscontract visible
GPU rack visual representing a future dedicated DCP endpoint for custom model servingfig. 04 - adapter intent to endpoint proof

The product contract is ready for operators. The endpoint must still prove itself.

The deployed endpoint becomes real only when the backend receives matching load proof from the serving layer for the deployment id, adapter id, base model, mode, endpoint id, and checksum. Endpoint smoke then has to prove a funded deterministic request, response hash, latency, token totals, and adapter trace. The POST route is live only as a disabled validation contract that returns 409 and records nothing. Until then, deployment rows are planning and audit records, not public route promises.

curl -s "https://api.dcp.sa/api/adapters/deployments?limit=25" \
  -H "Authorization: Bearer $DCP_RENTER_KEY"

curl -s https://api.dcp.sa/api/adapters/adpt_support_arabic/deployments \
  -H "Authorization: Bearer $DCP_RENTER_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "deployment_id": "adpl_support_arabic_001",
    "mode": "single_adapter_live_merge",
    "endpoint_id": "endpoint_qwen_arabic_01",
    "route_traffic": false
  }'

curl -s https://api.dcp.sa/api/adapters/endpoints/smoke/readiness

curl -s https://api.dcp.sa/api/adapters/$ADAPTER_ID/deployments/$DEPLOYMENT_ID/endpoint-smoke \
  -H "Authorization: Bearer $DCP_RENTER_KEY"

curl -s https://api.dcp.sa/api/adapters/$ADAPTER_ID/deployments/$DEPLOYMENT_ID/endpoint-smoke \
  -X POST \
  -H "Authorization: Bearer $DCP_RENTER_KEY" \
  -H "Content-Type: application/json" \
  -d '{"funded_smoke_principal":true,"smoke_result":{"request_id":"req_smoke_001"}}'

curl -s https://api.dcp.sa/api/adapters/usage/attribution/readiness

curl -s https://api.dcp.sa/api/adapters/settlement/readiness

curl -s https://api.dcp.sa/api/adapters/billing/approval/readiness

curl -s https://api.dcp.sa/api/adapters/billing/readiness
  • Single-adapter live merge is first; multi-LoRA waits for controlled vLLM smoke.
  • Route traffic remains false until deployment, adapter, base model, mode, endpoint, and checksum proof match.
  • Usage writes and billed inference start only after endpoint smoke proves response hash, latency, token totals, adapter trace, funded principal, usage attribution, and settlement policy.
Fireworks-style boundary

Dedicated deployments are the bridge between LoRA and revenue.

Fireworks separates serverless inference from fine-tuned LoRA deployment. DCP should do the same: public inference for general models, dedicated endpoints for customer adapters, and traffic only after proof.

Prepare an adapterRent a GPU pod
01adapter_ready

Adapter artifact metadata must be registered and ready.

02serving_load_proof

The endpoint reports matching deployment id, adapter id, base model, mode, endpoint id, and checksum.

03endpoint_smoke

A funded deterministic request proves response hash, latency, token totals, and adapter trace.

04route_traffic

Traffic and billing stay off until the proof row marks the deployment running.

05usage_attribution

Usage rows must carry deployment, adapter, endpoint, checksum, provider, request, token, cost, and pending-settlement proof.

06billing_readiness

Billing stays disabled until usage rows carry adapter and endpoint attribution.