確定性的對抗臨床模擬結合即時多評審計。演練高風險的破裂修復、危機篩查與邊界對抗,使用自適應合成人物。
AI 翻譯
Deterministic adversarial clinical simulation meets real-time multi-judge auditing. Rehearse high-stakes rupture repairs, crisis screenings, and boundary confrontations with adaptive synthetic persona
健康TypeScript · PythonGitHub @daggerstuff上線中營運者未確認
試用過了嗎? 點一下就好,不需要登入。
尚未偵測到更新。GitHub 發布版本後會顯示在這裡。
個人 GitHub 帳號下公開儲存庫的擁有者可以登入並驗證所有權,自行管理介紹,無需管理員核准。
登入並產生驗證檔案,在 30 分鐘內將其提交至此儲存庫的預設分支。其他分支或分叉儲存庫中的檔案無效。然後在下方檢查檔案,即可無需管理員核准直接管理專案。請保留檔案以維持管理權限。
如需登入,將前往登入頁面。登入後請回到此處,再次點擊所選的驗證按鈕。
在專案資料夾中執行 /nomorevibe verify,即可驗證所有權並自行管理介紹。
一分鐘完成驗證確定性的對抗臨床模擬結合即時多評審計。演練高風險的破裂修復、危機篩查與邊界對抗,使用自適應合成人物——逐回合監督,並具即時重試迴路。
AI 翻譯
Deterministic adversarial clinical simulation meets real-time multi-judge auditing. Rehearse high-stakes rupture repairs, crisis screenings, and boundary confrontations with adaptive synthetic personas—supervised turn-by-turn with instant retry loops.
取自儲存庫 README
正在載入完整 README。 以下為已儲存的摘錄。
Pioneering the digital frontier of mental health.
Pixelated Empathy is fundamentally reshaping the mental health landscape.
Historically, the clinical world has pushed back against integrating deeply with
the digital realm—often out of valid concerns for safety, empathy, and privacy.
We are here to bridge that gap in a highly positive and beneficial way. By
merging clinical rigor with advanced, carefully aligned AI, we create
environments where high-stakes, human-centered teams can practice, analyze, and
evolve their emotional intelligence before stepping into real-world scenarios.檢視完整 README ↗Pioneering the digital frontier of mental health.
Pixelated Empathy is fundamentally reshaping the mental health landscape. Historically, the clinical world has pushed back against integrating deeply with the digital realm—often out of valid concerns for safety, empathy, and privacy.
We are here to bridge that gap in a highly positive and beneficial way. By merging clinical rigor with advanced, carefully aligned AI, we create environments where high-stakes, human-centered teams can practice, analyze, and evolve their emotional intelligence before stepping into real-world scenarios.
Our goal moving forward is to scale deep, genuine empathy through digital platforms without losing the profoundly human essence of care. We believe that by thoughtfully bringing therapeutic principles into the digital realm, we can unlock new paradigms for training, healing, and connection.
| Requirement | Version | Check |
|---|---|---|
| Node.js | >= 24 | node --version |
| pnpm | 12.16.1 | pnpm --version |
| Python | 3.14 | python3 --version |
| uv | latest | uv --version |
| PostgreSQL | 17+ | psql --version |
| Redis | 7+ | redis-cli --version |
| Docker | latest | docker --version |
# 1. Clone with submodules (or init them after cloning)
git clone --recursive https://github.com/daggerstuff/pixelated.git
cd pixelated
# If already cloned without --recursive:
git submodule update --init --recursive
# 2. Install dependencies
pnpm install
# 3. Start the dev server
pnpm dev
The app runs at http://localhost:5173.
Submodule Note: This repo uses three git submodules (
ai/,foresight/,docs/). After any clone or pull, rungit submodule update --init --recursiveto sync them. Without this step, those directories will be empty.
To run the full stack (web + AI services + bias detection + analytics + worker):
pnpm dev:all-services
See CONTRIBUTING.md for the complete developer guide.
This platform operates as a cohesive unit across four primary repositories. Together, they form the secure, intelligent backbone of our clinical AI engine:
| Repository | Role |
|---|---|
| pixelated (Main Core) | Primary orchestrator: application surfaces, clinical dashboards, API routes, shared product logic. |
| ai (Cognitive Engine) | Model research and training: inference, emotional signal analysis, clinical validity pipelines. |
| foresight (Continuity & Memory) | Persistent contextual memory for AI agents across longitudinal therapeutic simulations. |
| docs (Knowledge Base) | HIPAA compliance, security posture, clinical feedback loops, architecture decisions. |
ai/, foresight/, and docs/ are git submodules — run
git submodule update --init --recursive after cloning to populate them.
pixelated/
├── apps/
│ ├── web/ # Main Astro 6 + React 19 application
│ └── business-strategy-cms/ # Business strategy CMS
├── ai/ # Submodule: AI inference & training (Python)
├── foresight/ # Submodule: Foresight MCP memory server (Python)
├── docs/ # Submodule: Documentation knowledge base
├── agents/ # AI agent packages (advisor, content, eve, intake, pipeline)
├── packages/ # Shared workspace packages
├── config/ # Vitest, Playwright, and other configs
├── content/ # Astro content collections
├── data/ # Seed data and fixtures
├── infra/ # Infrastructure (Docker, K8s)
├── scripts/ # DevOps and utility scripts
└── tests/ # Integration and E2E tests
pnpm dev # Start dev server (port 5173)
pnpm dev:all-services # Start all services concurrently
pnpm build # Production build
pnpm lint # Lint (oxlint, type-aware)
pnpm format # Format code (oxfmt + prettier)
pnpm test:unit # Unit tests (Vitest)
pnpm test:integration # Integration tests
pnpm e2e # End-to-end tests (Playwright)
pnpm e2e:ui # Interactive E2E
uv run pytest # Run Python tests
uv run ruff check . # Lint Python
uv run ruff format . # Format Python
ai/ is the Python training and research engine (Python 3.14, managed with
uv). It is not part of the root uv workspace — it ships its own
pyproject.toml and uv.lock:
cd ai
uv sync # Create the submodule venv
uv run pytest # Full test suite
make test # Training test suite (matches CI)
uv run ruff check . # Lint
Training datasets are versioned with DVC (remote: Cloudflare R2,
pixel-archive/ai-data); fetch them with cd ai && dvc pull.
CI gate: the Quality workflow's ai submodule strict ratchet job
(scripts/ci/ai-strict-ratchet.mjs) pins the submodule's legacy
strict-typing debt per file and per error code in
scripts/ci/ai-strict-baseline.json. It fails when that debt grows or a new
.py file fails mypy --strict — new code must be strict-clean, and debt
reductions are re-pinned with --update / --prune.
git submodule update --init --recursive # Init after fresh clone
pnpm submodules:sync # Re-sync after pull
| Layer | Technology |
|---|---|
| Frontend / SSR | Astro 6 + React 19 (TypeScript, Tailwind CSS) |
| Backend / AI | FastAPI / Express / Flask (Python 3.14 via uv) |
| Database | PostgreSQL 17 + pgvector |
| Caching | Redis |
| State | Zustand |
| Package Manager | pnpm 12.16.1 |
| Runtime | Node.js >= 24 |
See CONTRIBUTING.md for branch naming, commit conventions, testing, and security rules.
@ts-ignore, @ts-nocheck, # noqa, or
# type: ignorepnpm lint and pnpm format:check before pushingAGENTS.md for AI coding assistant conventionsThe docs/ submodule contains 980 files across 28 subdirectories. See
DOCS.md for a full index of all subdirectories, or browse the key
areas below:
apps/web/src/content-store/docs/):
product Gemini usage routes through Vertex AI only, a Google Cloud HIPAA BAA is signed before PHI touches GCP, and
no PHI passes through the Gemini API/AI Studio or logs/metadataRun
git submodule update --init --recursivefirst ifdocs/is empty.
Report vulnerabilities to [email protected].
See docs/SECURITY.md and docs/THREAT_MODEL.md for details.
© 2026 Pixelated Empathy. All rights reserved.