Partner Onboarding

DrSimon Engine is a behavior-analysis API and SDK contract for education products that need tenant-scoped learning or proctoring signals. The current portal supports partner review, reference-client downloads, API exploration, and readiness verification.

Production SDK sales are not yet claim-ready. Use this page to run a controlled partner evaluation while the remaining platform and video-evidence gates are closed.

What partners can evaluate now

Area	Available now	Boundary
API surface	Session lifecycle, primitive ingest, proctoring session APIs, reports, and OpenAPI docs	Partner token required; no raw media upload
SDK downloads	Production integration package, versioned reference package, JavaScript sample, Python sample	Approved partner evaluation only; not a complete external sales claim
Platform evidence	Web and Android evidence complete in the current adapter matrix; Chrome mostly complete	Chrome low-end Chromebook proof remains pending
Review evidence	Study-video CSV review packet for rights, privacy, and label review	The CSV does not approve sources or allow training claims

Evaluation flow

Request partner access from 3R Innovation operations.
Review the API boundary in API Reference and the current blockers in Readiness.
Download the production integration package or reference SDK package from SDK Downloads.
Run the public package smoke:

python scripts/smoke_public_sdk_download.py --json
python scripts/build_sdk_partner_install_smoke.py --json

Run the release handoff commands for SDK distribution and platform attach:

python scripts/build_production_claim_evidence_intake.py --write --json
python scripts/verify_production_claim_evidence_intake.py --json
python scripts/build_sdk_distribution_candidate.py --json
python scripts/build_sdk_production_distribution_package.py --json
python scripts/check_sdk_production_package_distribution.py --json
python scripts/check_sdk_release_artifact_integrity.py --json
python scripts/check_sdk_distribution_readiness.py --json
python scripts/check_sdk_auth_tenant_session_contract.py --json
python scripts/check_sdk_offline_reliability_contract.py --json
python scripts/check_sdk_live_e2e_verification_contract.py --json
python scripts/build_windows_student_adapter_handoff.py --json
python scripts/check_sdk_platform_adapter_distribution_contract.py --json
python scripts/sdk_service_attach_handoff.py --json
python scripts/build_chrome_low_end_evidence_handoff.py --json

The evidence intake command creates .handoff/production-claim-evidence-intake/ with the current blocker list, owner handoff JSONs, study-video reviewer CSV, and SHA ledger for transfer to another machine. It does not mark the SDK production-claim ready. Run the verifier after creation and after transfer before using any attach/apply command. When the sibling tripler_monorepo worktree is available, the bundle also carries the Windows collector PowerShell under windows-student/scripts/ so a Windows operator can run the field collection from the transfer bundle.

On a managed low-end Chromebook, collect the Chrome extension CPU budget evidence and validate it from the engine repo:

cd ../focuspang-chrome-extension
npm run verify:chromebook-proctoring-field -- \
  --ceu-remote-debugging-url <managed-chromebook-cdp-url> \
  --teacher-detail-media-result <teacher-detail-media.json> \
  --seat-evidence <seat-flow.json> \
  --session-id <session-id> \
  --step8-screen-share-picker-accepted \
  --device-model <model> \
  --chromeos-version <ChromeOS version> \
  --cpu-class <atom|celeron|pentium_gold|snapdragon_4gen|exynos_8> \
  --ram-mb <4096> \
  --logical-cpu-count <4> \
  --output .handoff/chrome-low-end/low-end-chromebook-cpu-budget.json \
  --out-dir .handoff/chrome-low-end

python scripts/check_sdk_low_end_chromebook_cpu_budget_evidence.py \
  --report <low-end-chromebook-cpu-budget-json> \
  --json
python scripts/attach_sdk_platform_adapter_evidence.py \
  chrome_low_end_chromebook \
  --artifact <low-end-chromebook-cpu-budget-json> \
  --write \
  --json

On a Windows student device or managed VM, collect the Windows student adapter field evidence from the desktop candidate:

cd apps/focuspang-desktop
pnpm evidence:windows-student -- `
  -Output .handoff/drsimon-windows-student-evidence.json `
  -OfflineReplayArtifact C:\evidence\offline-replay.json `
  -LicenseGovernanceArtifact C:\evidence\license-governance.json `
  -OfflineReplayValidated `
  -MediaPipeLicenseCleanStackValidated `
  -TenantSessionLifecycleValidated `
  -NoRawMediaUploadValidated `
  -ClientSuppliedScoresRejected `
  -DesktopNoBatteryReason desktop_lab_device_no_battery `
  -ThermalZoneNotSupportedReason thermal_zone_unavailable_on_device

Then copy or sync the JSON artifact into the engine handoff location and run:

python scripts/check_windows_student_adapter_evidence.py \
  --report <windows-student-evidence-json> \
  --json
python scripts/attach_sdk_platform_adapter_evidence.py \
  windows_student \
  --artifact <windows-student-evidence-json> \
  --write \
  --json
python scripts/check_sdk_platform_adapter_distribution_contract.py --json
python scripts/check_sdk_distribution_readiness.py --json

For study-video work, generate the reviewer CSV packet:

python scripts/build_study_video_review_packet.py --json \
  --manifest docs/research/drsimon-study-video-training-corpus.json \
  --csv .handoff/study-video-review/study-video-review-packet.csv

The CSV includes blank approval-input columns for rights, privacy, behavior labels, split evidence, eval evidence, and hash refs. Blank columns do not approve any source; they only define the reviewer handoff schema.

After reviewers fill the CSV, validate it first, then write only if the validation report is ready:

python scripts/apply_study_video_review_packet.py --json \
  --manifest docs/research/drsimon-study-video-training-corpus.json \
  --review-csv .handoff/study-video-review/study-video-review-packet.csv
python scripts/apply_study_video_review_packet.py --json \
  --manifest docs/research/drsimon-study-video-training-corpus.json \
  --review-csv .handoff/study-video-review/study-video-review-packet.csv \
  --write

The apply step is fail-closed. It requires two distinct reviewers, rights and privacy evidence, behavior-label evidence, split evidence, hash pins, and the correct model_training or false_positive_eval role before a source row can count.

API key and tenant setup

Partner keys are issued by 3R Innovation operations. Keys should be injected by the host application or trusted native shell. Do not embed partner keys in browser-only code.

Each integration must provide:

Product scope, such as edu
Tenant or school scope
Device profile and SDK tier
Session start and end lifecycle
Batch identifiers and retry policy
No raw sensor or media payloads to the engine

Production promotion checklist

Gate	Required before sales claim
SDK distribution	Windows student adapter field evidence from the desktop collector
Platform attach	Chrome low-end Chromebook CPU budget evidence
Study-video training	Rights, privacy, trainability, two-reviewer label, split, and hash evidence for 50 approved training sources
False-positive eval	Rights, privacy, label, split, and eval evidence for 60 approved holdout sources

Explicitly disallowed claims

Drop-in production SDK
Raw media upload to DrSimon Engine
Client-supplied factor scores or BN scores
Clinical diagnosis SDK
Platform adapters are optional
Study-video candidates are approved for training or evaluation