From 5dfe158bb0652c9a5ced4bdffb52445cdc3fc1e6 Mon Sep 17 00:00:00 2001 From: hyeokjun32 Date: Thu, 11 Jun 2026 22:14:47 +0900 Subject: [PATCH] docs: add local studio demo walkthrough --- README.ko.md | 3 + README.md | 4 + .../local_studio_demo_walkthrough.ko.md | 97 +++++++++++++++++++ .../local_studio_demo_walkthrough.md | 97 +++++++++++++++++++ 4 files changed, 201 insertions(+) create mode 100644 docs/portfolio/local_studio_demo_walkthrough.ko.md create mode 100644 docs/portfolio/local_studio_demo_walkthrough.md diff --git a/README.ko.md b/README.ko.md index 07820b4..5818503 100644 --- a/README.ko.md +++ b/README.ko.md @@ -63,6 +63,9 @@ InferEdge는 다음을 연결하는 validation pipeline입니다. Local Studio는 CLI/API/job workflow를 브라우저에서 조작하고 관찰하는 local-first interface입니다. cloud SaaS dashboard가 아니며, 사용자의 PC에서 실행되는 demo/review UI입니다. +역할별 브라우저 데모 순서는 [Local Studio demo walkthrough](docs/portfolio/local_studio_demo_walkthrough.md) +([한국어: Local Studio 데모 가이드](docs/portfolio/local_studio_demo_walkthrough.ko.md))를 기준으로 확인합니다. + ### 브라우저 데모 실행 1. `poetry run inferedgelab serve --host 127.0.0.1 --port 8000` 실행 diff --git a/README.md b/README.md index 372627c..365632a 100644 --- a/README.md +++ b/README.md @@ -136,6 +136,7 @@ Portfolio entry points: |---|---|---| | [Portfolio submission](docs/portfolio/inferedge_portfolio_submission.md) | [한국어: 포트폴리오 제출 문서](docs/portfolio/inferedge_portfolio_submission.ko.md) | submission-ready project narrative | | [Resume/interview summary](docs/portfolio/inferedge_resume_interview_summary.md) | [한국어: 이력서/면접 요약](docs/portfolio/inferedge_resume_interview_summary.ko.md) | short role-specific explanation | +| [Local Studio demo walkthrough](docs/portfolio/local_studio_demo_walkthrough.md) | [한국어: Local Studio 데모 가이드](docs/portfolio/local_studio_demo_walkthrough.ko.md) | browser demo path and role-specific talking route | | [1-page architecture summary](docs/portfolio/inferedge_1page_architecture.md) | [한국어: 1페이지 아키텍처 요약](docs/portfolio/inferedge_1page_architecture.ko.md) | ecosystem diagram and role split | | [Pipeline status](docs/portfolio/inferedge_pipeline_status.md) | [한국어: 파이프라인 상태](docs/portfolio/inferedge_pipeline_status.ko.md) | current implementation status | @@ -170,6 +171,9 @@ It runs on the user's machine through the FastAPI server and is intended as a lo InferEdge Local Studio can replay the bundled portfolio evidence without requiring a live Jetson device during an interview walkthrough. The `Load Demo Evidence` flow imports the ONNX Runtime CPU and TensorRT Jetson Runtime JSON fixtures from [examples/studio_demo](examples/studio_demo), refreshes Compare View, and keeps the demo pair selectable in Recent jobs while the local server process is running. +For a role-specific browser demo route, use [Local Studio demo walkthrough](docs/portfolio/local_studio_demo_walkthrough.md) +([한국어: Local Studio 데모 가이드](docs/portfolio/local_studio_demo_walkthrough.ko.md)). + ### Run the Browser Demo 1. Run `poetry run inferedgelab serve --host 127.0.0.1 --port 8000` diff --git a/docs/portfolio/local_studio_demo_walkthrough.ko.md b/docs/portfolio/local_studio_demo_walkthrough.ko.md new file mode 100644 index 0000000..ed4dcd7 --- /dev/null +++ b/docs/portfolio/local_studio_demo_walkthrough.ko.md @@ -0,0 +1,97 @@ +# InferEdge Local Studio Demo Walkthrough 한국어 Quick Guide + +언어: [English](local_studio_demo_walkthrough.md) | 한국어 + +이 문서는 Local Studio를 보여줄 때 어떤 순서로 설명하고, 역할별로 어떤 메시지를 강조할지 빠르게 확인하기 위한 요약본이다. 대표/canonical 문서는 [InferEdge Local Studio Demo Walkthrough](local_studio_demo_walkthrough.md)이다. + +## 데모 경계 + +Local Studio는 사용자의 PC에서 committed evidence를 재생하고 API/job/report contract를 확인하는 local-first workflow UI다. + +production SaaS dashboard, cloud control plane, production worker service, production remote execution proof가 아니다. + +## 실행 순서 + +```bash +poetry run inferedgelab serve --host 127.0.0.1 --port 8000 +``` + +`http://localhost:8000/studio`를 열고 `Load Demo Evidence`를 클릭한다. + +이 경로는 live Jetson 없이도 다음 evidence를 보여준다. + +- ONNX Runtime CPU FP32 baseline fixture +- TensorRT Jetson FP16 25W candidate fixture +- Jetson 15W/25W power-mode context +- review/block problem case +- available optional AIGuard portfolio evidence + +## 보여줄 순서 + +1. TensorRT Jetson vs ONNX Runtime 비교부터 보여준다. +2. 이 demo pair가 committed fixture이며 live Jetson 없이 재생 가능하다고 말한다. +3. Lab-owned deployment decision context를 확인한다. +4. 문제 케이스로 annotation missing, invalid structure, contract mismatch, latency regression이 review/block evidence로 남는다는 점을 보여준다. +5. Runtime Intelligence는 Studio live dashboard가 아니라 Lab-owned report chain으로 설명한다. + +## 인용할 수치 + +| Evidence | 값 | +|---|---:| +| ONNX Runtime CPU FP32 mean | `45.4299 ms` | +| ONNX Runtime CPU FP32 p99 | `49.2128 ms` | +| ONNX Runtime CPU FP32 FPS | `22.0119` | +| TensorRT Jetson FP16 25W mean | `10.066401 ms` | +| TensorRT Jetson FP16 25W p99 | `15.548438 ms` | +| TensorRT Jetson FP16 25W FPS | `99.340373` | +| Studio demo speedup | 약 `4.51x` | + +이 pair는 backend/device/precision context가 다른 deployment review evidence이며, same-condition regression으로 말하지 않는다. + +## Runtime Intelligence 연결 + +Studio evidence에서 Runtime Intelligence로 넘어갈 때는 아래 Lab-owned report chain을 사용한다. + +```text +InferEdgeOrchestrator operation feed / operation_risk_rollup +-> InferEdgeEnv telemetry history / comparability-first regression context +-> optional InferEdgeAIGuard deterministic runtime evidence +-> InferEdgeLab Runtime Intelligence Risk Summary / deployment risk report +``` + +함께 볼 문서: + +- [EdgeEnv runtime regression Lab handoff](edgeenv_runtime_regression_lab_handoff.md) +- [Resume/interview summary](inferedge_resume_interview_summary.md) +- [Pipeline status](inferedge_pipeline_status.md) + +핵심 문장: Orchestrator, EdgeEnv, AIGuard는 evidence provider이고, final deployment decision owner는 Lab이다. + +## 역할별 경로 + +| 역할 | 먼저 보여줄 것 | 명확히 말할 것 | +|---|---|---| +| AI Inference Engineer | Runtime comparison, latency/p99/FPS, compare identity | 단순 benchmark가 아니라 provenance-aware inference validation이다. | +| Embedded / Edge Engineer | Jetson FP16 25W/15W evidence, device-local preservation context | demo는 live hardware 없이 재생 가능하지만, 새 live evidence에는 device가 필요하다. | +| Backend / AI Platform | API/job/report contract, worker boundary, Lab decision context | contract/evidence orchestration이지 DB/queue/auth/billing production SaaS가 아니다. | + +## 피할 표현 + +- production SaaS 완성 +- Local Studio는 cloud dashboard +- Runtime Intelligence는 production observability +- remote dispatch가 production remote execution을 증명 +- AIGuard나 Orchestrator가 final deployment decision owner +- Studio pair가 same-condition regression + +## CLI 확인 + +브라우저를 열지 않을 때는 아래 명령으로 같은 evidence를 확인한다. + +```bash +poetry run inferedgelab demo-evidence-summary +poetry run inferedgelab portfolio-demo-check +poetry run inferedgelab export-demo-evidence --output reports/studio_demo_evidence.md +``` + +`portfolio-demo-check`는 committed Studio fixture, README metric, portfolio docs, local Studio asset을 빠르게 검증하는 guard다. diff --git a/docs/portfolio/local_studio_demo_walkthrough.md b/docs/portfolio/local_studio_demo_walkthrough.md new file mode 100644 index 0000000..41789b5 --- /dev/null +++ b/docs/portfolio/local_studio_demo_walkthrough.md @@ -0,0 +1,97 @@ +# InferEdge Local Studio Demo Walkthrough + +Language: English | [한국어](local_studio_demo_walkthrough.ko.md) + +Use this walkthrough when reviewing InferEdge through the local browser UI. It keeps the demo focused on committed evidence, role-specific talking points, and Lab-owned decision boundaries. + +## Demo Boundary + +Local Studio is a local-first workflow UI. It replays committed evidence and inspects API/job/report contracts on the user's machine. + +It is not a production SaaS dashboard, cloud control plane, production worker service, or production remote execution proof. + +## Run The Demo + +```bash +poetry run inferedgelab serve --host 127.0.0.1 --port 8000 +``` + +Open `http://localhost:8000/studio`, then click `Load Demo Evidence`. + +The stable browser path loads: + +- ONNX Runtime CPU FP32 baseline fixture from `examples/studio_demo` +- TensorRT Jetson FP16 25W candidate fixture from `examples/studio_demo` +- Jetson 15W/25W power-mode context +- validation problem cases for review/block paths +- optional AIGuard portfolio evidence when available + +## Review Order + +1. Start with the TensorRT Jetson vs ONNX Runtime comparison. +2. Confirm the demo pair uses committed fixtures and does not require a live Jetson device. +3. Open the Lab-owned deployment decision context. +4. Use the problem cases to show that missing annotations, invalid structures, contract mismatch, and latency regression become explicit review/block evidence. +5. For Runtime Intelligence, pivot to the report path rather than treating Studio as a live observability dashboard. + +## Evidence To Quote + +| Evidence | Value | +|---|---:| +| ONNX Runtime CPU FP32 mean | `45.4299 ms` | +| ONNX Runtime CPU FP32 p99 | `49.2128 ms` | +| ONNX Runtime CPU FP32 FPS | `22.0119` | +| TensorRT Jetson FP16 25W mean | `10.066401 ms` | +| TensorRT Jetson FP16 25W p99 | `15.548438 ms` | +| TensorRT Jetson FP16 25W FPS | `99.340373` | +| Studio demo speedup | about `4.51x` | + +Interpret the pair as deployment review evidence across backend/device/precision context, not same-condition regression. + +## Runtime Intelligence Hand-Off + +When the walkthrough moves from browser evidence to Runtime Intelligence, use the Lab-owned report chain: + +```text +InferEdgeOrchestrator operation feed / operation_risk_rollup +-> InferEdgeEnv telemetry history / comparability-first regression context +-> optional InferEdgeAIGuard deterministic runtime evidence +-> InferEdgeLab Runtime Intelligence Risk Summary / deployment risk report +``` + +Point reviewers to: + +- [EdgeEnv runtime regression Lab handoff](edgeenv_runtime_regression_lab_handoff.md) +- [Resume/interview summary](inferedge_resume_interview_summary.md) +- [Pipeline status](inferedge_pipeline_status.md) + +The key sentence is: Orchestrator, EdgeEnv, and AIGuard provide evidence; Lab remains the final deployment decision owner. + +## Role-Specific Route + +| Role | Show first | Say clearly | +|---|---|---| +| AI Inference Engineer | Runtime comparison, latency/p99/FPS, compare identity | This is provenance-aware inference validation, not only a benchmark. | +| Embedded / Edge Engineer | Jetson FP16 25W/15W evidence and device-local preservation context | The demo can be replayed without live hardware; new live evidence still requires the device. | +| Backend / AI Platform | API/job/report contract, worker boundary, Lab decision context | This is contract and evidence orchestration, not DB/queue/auth/billing production SaaS. | + +## Avoid Saying + +- "production SaaS is complete" +- "Local Studio is a cloud dashboard" +- "Runtime Intelligence is production observability" +- "remote dispatch proves production remote execution" +- "AIGuard or Orchestrator owns the final deployment decision" +- "the Studio pair is same-condition regression" + +## CLI Checks + +Use these checks when the browser is not needed: + +```bash +poetry run inferedgelab demo-evidence-summary +poetry run inferedgelab portfolio-demo-check +poetry run inferedgelab export-demo-evidence --output reports/studio_demo_evidence.md +``` + +`portfolio-demo-check` is the quick guard for committed Studio fixtures, README metrics, portfolio docs, and local Studio assets.