Canine Behavioral Intelligence Platform
The open research infrastructure for structured canine behavioral data
BarkMind is not a pet social media platform. It is research infrastructure for canine behavioral science.
Professional dog trainers, shelter behaviorists, groomers, and veterinary staff encounter behavioral incidents every day. Those encounters generate expert knowledge that almost always disappears — into notebooks, case files, and memory. BarkMind captures that knowledge in a structured, searchable, auditable format and builds it into a dataset that behavioral AI can actually learn from.
The core loop:
Professional submits behavioral incident
→ Community annotates with controlled vocabulary
→ Verified experts review and file formal verdicts
→ Multi-expert consensus resolves complex cases
→ Evidence locks — case becomes an immutable dataset entry
The platform is operational, governed, and open to professional contributors.
| Service | URL | Status |
|---|---|---|
| Frontend | https://barkmind.jesseboudreau.com |  |
| Backend API | https://barkmind-api.jesseboudreau.com | |
| API Documentation | https://barkmind-api.jesseboudreau.com/docs | OpenAPI / Swagger |
| Governance Status | https://barkmind-api.jesseboudreau.com/governance/status | Aegis-compatible |
The live platform includes realistic demo cases with verified expert profiles. To access admin or expert features, open an issue using the Expert Verification template or contact the maintainer.
Visual architecture diagram:
docs/assets/architecture.svg
┌─────────────────────────────────────────────────────────────┐
│ Cloudflare Edge │
│ TLS 1.3 · DDoS Protection · WAF · HTTP/2 │
└─────────────────┬───────────────────────────────────────────┘
│ Named Tunnel: reselleros
│
┌─────────────┴──────────────────────────┐
│ VM: vmi3002990 │
│ │
│ ┌──────────────┐ ┌──────────────┐ │
│ │ Frontend │ │ Backend │ │
│ │ Next.js 16 │ │ FastAPI │ │
│ │ :3008 │◄─► :8108 │ │
│ │ App Router │ │ uvicorn │ │
│ └──────────────┘ └──────┬───────┘ │
│ │ │
│ ┌──────▼───────┐ │
│ │ PostgreSQL │ │
│ │ 22 tables │ │
│ │ 73-term │ │
│ │ taxonomy │ │
│ └──────────────┘ │
│ │
│ ┌──────────────────────────────────┐ │
│ │ Aegis AI — Governance Control │ │
│ │ Audit events · Topology registry │ │
│ └──────────────────────────────────┘ │
└────────────────────────────────────────┘
Supervision: systemd (barkmind-backend.service, barkmind-frontend.service)
Ingest: Cloudflare Named Tunnel (no raw ports exposed)
Auth: JWT Bearer — python-jose + passlib bcrypt
Media: Pillow + ffmpeg — local disk (S3-ready abstraction)
| Feature | Description |
|---|---|
| Case Submission | Upload video/images with incident description, setting, trigger context |
| Behavioral Taxonomy | 73 controlled terms across 14 categories (body posture, tail position, eye contact, stress indicators, arousal, social engagement, and more) |
| Structured Annotations | Typed annotations (observation/interpretation/concern/recommendation) with confidence levels and timestamp ranges |
| Timeline Markers | Named behavioral events pinned to exact video timestamps (trigger, escalation, de-escalation, handler intervention) |
| Expert Resolutions | Formal verdicts: Safe / Concern / Escalation Risk / Requires Intervention |
| Multi-Expert Consensus | Structured opinion aggregation — vote counting, not AI-generated |
| Evidence Locking | Resolved cases freeze with immutable snapshots → permanent dataset entries |
| Feature | Description |
|---|---|
| Immutable Audit Trail | Every action logged in append-only audit_events table |
| Expert Verification | Credential verification (CPDT, CAAB, CBCC, Fear Free) by admins |
| Reputation System | Event-driven accumulation — +5 resolution, +2 consensus alignment, etc. |
| Review Assignments | Claim/transfer/escalate workflow with traceability |
| Export Traceability | Every data export logged in export_jobs — who requested what, when |
| Dataset Snapshots | Point-in-time metadata capture for version tracking and citation |
| Feature | Description |
|---|---|
| Annotation Lineage | Full provenance chain: author credentials + confidence + revision history |
| Inter-Rater Analysis | Endpoint for IRR data extraction (Cohen's kappa ready) |
| NDJSON Export | Streaming dataset export for ML pipelines |
| Provenance API | GET /dataset/lineage/{case_id} returns complete annotation chain |
| Taxonomy API | GET /taxonomy returns structured vocabulary with severity and signal type |
22 tables across 6 layers:
Core: users, cases, case_media, tags, case_tags, comments
Annotation: annotations, annotation_taxonomy_refs, annotation_revisions
Taxonomy: taxonomy_terms
Timeline: timeline_markers
Resolution: expert_resolutions
Trust: expert_profiles, review_assignments, consensus_records,
expert_opinions, evidence_locks, audit_events, reputation_events
Operations: export_jobs, dataset_snapshots, organizations
Migration chain (Alembic):
249e14807858 → initial_schema
759b2d6bccd0 → add_media_thumbnails
9c08dac2316e → phase4_annotation_intelligence
c1d10499127d → phase5_trust_infrastructure
b698e6352af2 → phase6_operational_intelligence
git clone https://github.com/yourusername/BarkMind.git
cd BarkMind
docker compose up
# → Backend: http://localhost:8108/health
# → Frontend: http://localhost:3008python3 --version # 3.12+
node --version # 22+
psql --version # PostgreSQL 18+
ffmpeg -version # for video thumbnails# 1. Clone and configure
git clone https://github.com/yourusername/BarkMind.git
cd BarkMind
cp .env.example .env
# Edit .env: set DATABASE_URL, JWT_SECRET
# 2. Database setup
sudo -u postgres psql <<SQL
CREATE USER barkmind_user WITH PASSWORD 'yourpassword';
CREATE DATABASE barkmind OWNER barkmind_user;
GRANT ALL PRIVILEGES ON DATABASE barkmind TO barkmind_user;
\connect barkmind
GRANT ALL ON SCHEMA public TO barkmind_user;
SQL
# 3. Run migrations + seed
cd backend
pip install -r requirements.txt
alembic upgrade head
# Tags and taxonomy seed automatically on first backend startup
# 4. Start backend
cd backend
uvicorn app.main:app --host 127.0.0.1 --port 8108 --reload# 5. Install and build
cd frontend
npm install
npm run build # uses --webpack (required for VM stability)
# 6. Start frontend
npm run start # runs on :3008./start.sh # starts both services via systemd
./status.sh # shows service state, ports, health
./stop.sh # graceful stop
./restart.sh # restart both servicescurl http://127.0.0.1:8108/health # {"status":"ok","service":"barkmind"}
curl http://127.0.0.1:3008/ # 307 → /homeBarkMind is open to professional contributors. You don't need to be a developer.
- Dog trainers (CPDT, CBCC, KPA-CTP) — submit training session breakdowns
- Shelter staff — document intake behavioral assessments
- Groomers — submit fear response and stress escalation cases
- Veterinary staff — document restraint stress and fear protocol cases
- Daycare leads — submit group play escalation and overarousal incidents
- Students — annotate existing cases using the behavioral taxonomy
| Type | Who | How |
|---|---|---|
| Submit a case | Any registered user | Upload → /upload |
| Annotate a case | Any registered user | Browse /cases → add annotation |
| Expert review | Verified professionals | /expert — claim and resolve cases |
| Code contribution | Developers | See CONTRIBUTING.md |
# Backend tests
cd backend && pytest
# Frontend lint
cd frontend && npm run lint
# Type check
cd frontend && npx tsc --noEmitSee docs/CONTRIBUTOR_QUICKSTART.md for full onboarding.
BarkMind uses a curated 73-term controlled vocabulary across 14 categories:
body_posture tail_position ear_position
eye_contact mouth_tension stress_indicators
fear_indicators play_signals arousal_escalation
social_engagement avoidance resource_guarding
handler_intervention environmental_triggers
Each term has:
severity_hint(0–4): informational → mild → moderate → elevated → severesignal_type: threat / appeasement / stress / fear / arousal / play / social / avoidance / resource / handler / trigger / neutral
Browse the live taxonomy: barkmind.jesseboudreau.com/tags
Full taxonomy doc: docs/BEHAVIORAL_TAXONOMY.md
BarkMind is designed from the ground up to produce a trustworthy, citable research dataset.
Every annotation records:
- Author username and role at time of creation
- Expert verification status
- Confidence level (high/medium/low)
- Timestamp range (for video)
- Taxonomy terms referenced
- Full revision history
Resolved cases are locked with immutable snapshots. After locking:
- No new annotations can be added
- Existing annotations cannot be modified
- Media cannot be deleted
- A snapshot of complete case state is stored permanently
Every significant action creates an append-only AuditEvent record. Nothing can be
modified or deleted without a trace. Audit log is accessible to admin via API.
Contributors retain attribution on every annotation. If BarkMind's dataset contributes to published research or trained models, contributors are credited.
Full governance documentation: docs/GOVERNANCE_WORKFLOW.md
Phase 1-6: ✅ Platform foundation (backend, frontend, media, annotation, trust, governance)
Phase 7-9: ✅ Deployment, stability, systemd supervision
Phase 10: ✅ Port governance and lifecycle hardening
Phase 11: ✅ Demo dataset — 11 cases, 5 experts, 73-term taxonomy
Phase 12: ✅ Public presentation — landing page, contributor hub, about page
Phase 13: 🔄 Community activation (this phase)
Open source docs, ethics statement, annotation standards
Phase 14: 📋 Multimodal annotation
Frame extraction · Claude API integration · per-frame behavioral labels
Phase 15: 📋 Research dataset release
Open dataset · NDJSON export · contributor attribution · DOI
Phase 16: 📋 Behavioral AI foundations
Escalation prediction · risk scoring · trained on locked cases
BarkMind is committed to responsible development of behavioral AI.
- No autonomous conclusions. BarkMind produces human-labeled data. AI does not generate verdicts.
- No veterinary diagnosis. All platform output is behavioral observation, not medical diagnosis.
- No misrepresentation. AI summaries (when implemented) include explicit disclaimers.
- No hallucination risk. Structured scoring uses bounded enumerations, not free generation.
- Contributor attribution. Experts are credited on every annotation they create.
Full ethics statement: docs/ETHICS_AND_SAFETY.md
| Layer | Technology |
|---|---|
| Frontend | Next.js 16 (App Router, webpack) · TypeScript · Tailwind CSS 4 · SWR |
| Backend | FastAPI · Python 3.12 · SQLAlchemy 2 (async) · Pydantic v2 |
| Database | PostgreSQL 18 · asyncpg · Alembic |
| Auth | JWT Bearer · python-jose · passlib bcrypt |
| Media | Pillow · ffmpeg · local disk (S3-ready storage abstraction) |
| Deployment | systemd · Cloudflare Named Tunnel · uvicorn |
| Governance | Aegis AI control plane · immutable audit_events |
BarkMind/
├── backend/
│ ├── app/
│ │ ├── models/ # 22 SQLAlchemy models
│ │ ├── routers/ # 15+ FastAPI route modules
│ │ ├── schemas/ # Pydantic schemas
│ │ ├── services/ # Business logic layer
│ │ └── scripts/ # Seeding and utilities
│ ├── alembic/ # Migration chain
│ └── requirements.txt
├── frontend/
│ └── src/
│ ├── app/ # Next.js App Router pages
│ ├── components/ # Shared component library
│ ├── contexts/ # Auth context
│ └── lib/ # API client, types, utils
├── config/ # Aegis manifests, registry entry
├── docs/ # 70+ documentation files
├── prompts/ # AI prompt library
├── media/ # Local media storage
├── start.sh / stop.sh / restart.sh / status.sh
└── README.md
| Document | Purpose |
|---|---|
docs/ARCHITECTURE_OVERVIEW.md |
Full technical architecture |
docs/BEHAVIORAL_TAXONOMY.md |
73-term vocabulary reference |
docs/GOVERNANCE_WORKFLOW.md |
Case lifecycle and governance |
docs/ETHICS_AND_SAFETY.md |
Ethics statement |
docs/CONTRIBUTOR_QUICKSTART.md |
Onboarding for new contributors |
docs/FIRST_CASE_REVIEW.md |
Expert reviewer guide |
docs/ANNOTATION_STANDARDS.md |
Annotation quality rubric |
docs/DATASET_GOVERNANCE.md |
Dataset integrity documentation |
docs/LIVE_DEMO_SCRIPT.md |
15-minute demo walkthrough |
docs/LOCAL_DEV_GUIDE.md |
Local development guide |
docs/DEPLOYMENT_GUIDE.md |
Production deployment guide |
MIT License. See LICENSE for details.
Behavioral vocabulary terms are released under CC BY 4.0.
Dataset exports include full contributor attribution per the terms in docs/CONTRIBUTOR_PHILOSOPHY.md.
Co-Founder, Product Lead, and Principal Architect
Responsible for platform vision, AI architecture, engineering strategy, software development, and product direction.
Co-Founder, Canine Behavior Consultant, and Customer Experience Lead
Responsible for canine behavior expertise, pet care consulting, workflow design, customer experience strategy, product validation, documentation, and operational planning.
Together, Jesse and Darcee founded BarkMind to improve canine behavior understanding, pet care operations, and outcomes for pets, pet parents, trainers, groomers, boarding facilities, and veterinary teams through responsible use of AI and technology.
BarkMind is a founder-led project created by Jesse Boudreau and Darcee Sellers.
Combining decades of experience in pet care operations, canine behavior, customer experience, leadership, compliance, and technology, BarkMind is designed to help pet professionals better understand, document, and improve canine behavior through practical AI-powered tools.
BarkMind is a governed service running under the Aegis AI control plane.
# Governance status (no auth required)
curl https://barkmind-api.jesseboudreau.com/governance/status
# Aegis metadata
curl https://barkmind-api.jesseboudreau.com/.well-known/aegis-metaFounded by Jesse Boudreau & Darcee Sellers · Built by the behavioral professional community