chatbot-webflux

Gradle Kotlin monorepo: Spring WebFlux backend, ChatGPT-like web UI, and local Ollama LLM.

chatbot-webflux/
├── apps/backend/      # Spring WebFlux API
├── apps/web/          # Vite + React chat UI
├── infra/ollama/      # Model pull scripts
└── docs/design/

Prerequisites

• JDK 25 (Gradle auto-downloads via Foojay if not installed; Gradle 9.1+ required to run builds on JDK 25)
• Node.js 22+ (or Gradle downloads via Node plugin)
• Ollama on Intel Mac (x86_64)

Quick start

# 1. LLM
ollama serve
./infra/ollama/pull-models.sh

# 2. Backend (default profile → Ollama; auto-reloads on code changes)
./gradlew :apps:backend:dev

# 3. Web UI (separate terminal)
./gradlew :apps:web:npm_run_dev
# → http://localhost:5173

API

Endpoint	Description
GET /heartbeat	Service health
GET /v1/status	Server + Ollama status
POST /v1/chat	Stream chat (SSE) — body: { "message", "conversationId?" }
GET /actuator/health	Includes Ollama indicator

Gradle tasks

Task	Description
./gradlew buildAll	Build UI + backend (UI copied into backend static/)
./gradlew :apps:backend:dev	API on :8080 (auto-reload on code change)
./gradlew :apps:backend:bootRun	API on :8080 (no reload)
./gradlew :apps:backend:test	Tests (mock-llm profile)
./gradlew :apps:web:npm_run_dev	Vite on :5173

Profiles

Profile	Use
default	Ollama at 127.0.0.1:11434
mock-llm	Tests / no Ollama — ./gradlew :apps:backend:bootRun --args='--spring.profiles.active=mock-llm'

Hardware

Tuned for 2.2 GHz quad-core Intel i7 — default model llama3.2:1b. See docs/design/04-llm-integration.md.

Design

docs/design/README.md