SimTrainer

SimTrainer is an interactive practice lab for Resident Assistants (RAs) and Teaching Assistants (TAs). Trainees chat with an AI‑driven student persona, receive real‑time scores for empathy/curiosity/structure/satisfaction, and save conversations for reflection. Everything runs locally with Next.js, Prisma, and Ollama so recruiters and collaborators can inspect the full stack.

Highlights

• Real RA/TA scenarios – 20+ prebuilt tracks plus support for custom prompts.
• Local LLM feedback – the /api/evaluate route uses Ollama to score the entire conversation and determine whether the student feels satisfied.
• Conversation history – authenticated users can save, reopen, and reset runs with full transcripts and rubric metrics.
• Customizable UI – reusable ChatPanel, history dashboard, and builder pages to tweak personas without redeploying.

Tech Stack

Layer	Tech
Frontend	Next.js App Router, Tailwind CSS, TypeScript
Backend	Next.js API routes (/api/chat, /api/evaluate, /api/history)
Database	SQLite via Prisma ORM
Authentication	NextAuth.js (credentials provider)
AI Integration	Local Ollama models (default llama3) for conversation + evaluation

Prerequisites

1. Node.js 18+ (recommend nvm use 18).
2. npm (ships with Node) or pnpm/yarn/bun if you prefer.
3. Ollama:
   • Download from ollama.com/download.
   • Install and start the Ollama service.
   • Pull the conversational model: ollama pull llama3 (or your chosen model).
4. SQLite (bundled with Prisma; no separate install needed).

Getting Started

# 1. Clone the repo
git clone https://github.com/your-org/simtrainer.git
cd simtrainer

# 2. Install dependencies
npm install

# 3. Copy environment template and set secrets
cp .env.example .env.local
# Fill in NEXTAUTH_SECRET, DATABASE_URL, SMTP creds (optional), etc.

# 4. Generate Prisma client & migrate
npx prisma migrate dev

# 5. Ensure Ollama is running
ollama serve  # (Mac/Linux) or launch the Ollama desktop app on Windows

# 6. Start the dev server
npm run dev

Visit http://localhost:3000 to explore scenarios. The default RA/TA chat loads a scripted student opener; use the right sidebar to switch tracks or custom prompts.

Configuring Ollama

• The app points to http://127.0.0.1:11434 by default. Override via .env.local:

  OLLAMA_URL=http://localhost:11434
  OLLAMA_MODEL=llama3.1:8b  # any installed model

• /api/chat streams student responses, while /api/evaluate sends the entire transcript to Ollama to receive JSON feedback with empathy/curiosity/structure/satisfaction/resolved/summary.

Project Structure

simtrainer/
├─ app/
│  ├─ api/               # REST endpoints (chat, evaluate, auth, history)
│  ├─ auth/, history/…   # Pages rendered via App Router
│  └─ scenarios/         # Prebuilt RA/TA prompts
├─ components/           # ChatPanel, HistoryList, Scenario builders
├─ data/tracks.ts        # Scenario metadata
├─ prisma/
│  └─ schema.prisma      # Conversation + user models
├─ lib/                  # Prisma client, auth helpers, session storage
└─ README.md

Running in Production

npm run build
npm run start

Set NODE_ENV=production, ensure DATABASE_URL points to a persistent SQLite/ Postgres instance, and keep Ollama running on the same network location as defined in your env vars. Deployments to platforms like Vercel require a reachable Ollama endpoint (self-hosted VM or container).

Custom Scenarios

1. Navigate to /builder to author a new prompt.
2. Provide a title, short description, and persona notes.
3. Save the generated slug in your database or run it ad hoc.

Alternatively, add entries directly to data/tracks.ts and redeploy.

Testing the Evaluation Flow

• Start a sample conversation in /ra or /ta.
• Click End scenario & save to trigger /api/evaluate.
• Watch the numeric scores and “Scenario complete” banner appear when the student is satisfied.

Troubleshooting

Issue	Fix
fetch ECONNREFUSED 127.0.0.1:11434	Ensure Ollama is running and reachable; verify OLLAMA_URL.
Feedback JSON parsing error	Check Ollama logs; confirm the model supports JSON-style responses.
Prisma migration fails	Delete prisma/dev.db (if safe) and rerun npx prisma migrate dev.
“Unauthorized” on history pages	Create an account via /auth/register, then log in.

Contributing

1. Fork the repository.
2. Create a feature branch: git checkout -b feat/new-scenario.
3. Commit changes with clear messages.
4. Open a PR describing the scenario/feature and attach screenshots or Looms of the new interaction.

License

MIT © SimTrainer Team. Feel free to adapt for your own RA/TA coaching needs—just share improvements!