https://github.com/poojan38380/trekmate
Your AI-powered outdoor companion. Plan treks, check live weather, and chat for trail-ready tips—all in a sleek, lightning-fast UI.
https://github.com/poojan38380/trekmate
ai chatbot groq groq-api groq-llm hiking llm nextjs openmeteo openmeteo-api outdoors react tailwindcss travel-planner typescript weather weather-api
Last synced: about 1 month ago
JSON representation
Your AI-powered outdoor companion. Plan treks, check live weather, and chat for trail-ready tips—all in a sleek, lightning-fast UI.
- Host: GitHub
- URL: https://github.com/poojan38380/trekmate
- Owner: Poojan38380
- License: mit
- Created: 2025-10-05T10:17:36.000Z (8 months ago)
- Default Branch: main
- Last Pushed: 2025-10-05T14:28:32.000Z (8 months ago)
- Last Synced: 2025-10-05T16:12:27.668Z (8 months ago)
- Topics: ai, chatbot, groq, groq-api, groq-llm, hiking, llm, nextjs, openmeteo, openmeteo-api, outdoors, react, tailwindcss, travel-planner, typescript, weather, weather-api
- Language: TypeScript
- Homepage: https://trekmate-ten.vercel.app
- Size: 6.11 MB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Contributing: CONTRIBUTING.md
- License: LICENSE
- Code of conduct: CODE_OF_CONDUCT.md
- Security: SECURITY.md
Awesome Lists containing this project
README
TrekMate AI — Weather‑aware Trek & Travel Copilot (日本語音声入力対応)
Japanese voice input · Real‑time weather · Generative AI trek plans
---
TrekMate is a modern, AI-assisted trip planner for hikers and campers. Chat for trail advice, get real-time weather insights, and plan confidently with a fast, responsive interface.
### What it does (at a glance)
- Fetches weather from a free public API (Open‑Meteo) with no API key required
- Generates concise trek plans and suggestions using a Groq‑hosted LLM
- Supports Japanese voice input (and English) using the Web Speech API
- Detects Japanese text automatically and replies in natural Japanese
- Provides safety cues and time windows based on forecast analysis
---
## Demo
Left: Landing page • Right: AI chat interface
- Ask: 「明日、富士山で登山できる？」 or “Weather for Kyoto and a 2‑day plan?”
- Speak in Japanese by tapping the mic, then switch EN/JA as needed.
Note: Web Speech API support varies by browser. Chrome works best on desktop.
---
## Features
- Japanese voice input (ja‑JP) and English (en‑US) toggle
- Location geocoding with language awareness (Japanese or English search)
- 7‑day forecast + 72‑hour hourly analysis
- Weather summary with warnings, risk level, and best time windows
- Groq LLM prompt adapts to Japanese/English automatically
- Clean UI with chat history and follow‑up suggestions
---
## Tech Stack
- Next.js 14 (App Router) + React 18 + TypeScript
- Tailwind CSS for styling
- Open‑Meteo for geocoding and forecast (free, no key)
- Groq API for chat completions (I default to `llama-3.1-8b-instant`)
- Web Speech API for voice input (client‑side)
---
## Architecture overview
```text
app/
actions/
weather.ts # Server action: geocode + forecast + summary
ai.ts # Server action: Groq chat completion with compact prompt
components/
chat/ # Chat UI, voice controls, and client logic
lib/
weather.ts # Open‑Meteo integration and forecast analysis
types/ # TS types for weather and chat
```
Key flows I implemented:
- `getWeatherForLocation(name)` → geocodes (ja/en aware), fetches 7‑day forecast, builds hourly/daily arrays, then `generateWeatherSummary` returns warnings, risk, and best time window.
- `generateTrekPlan({ destination, weather, summary, userQuestion? })` → builds a compact prompt (to keep tokens low and speed high) and calls Groq.
- Client hook `useChat()` orchestrates phases, voice input via Web Speech API, and adaptive suggestions.
---
## Getting started
### 1) Install dependencies
```bash
npm install
```
### 2) Environment variables
Copy `.env` from `example.env` and set your Groq key.
```bash
cp example.env .env
```
Required:
- `GROQ_API_KEY`: your Groq API key
Optional:
- `GROQ_MODEL`: override model (default: `llama-3.1-8b-instant`)
### 3) Run locally
```bash
npm run dev
```
Open `http://localhost:3000`.
### 4) Voice input support
- Works in browsers that support the Web Speech API. Desktop Chrome recommended.
- Use the EN/JA toggle in the input bar to switch recognition language.
---
## How I solved the core requirements
- Japanese voice input: I wired the Web Speech API inside `components/chat/useChat.ts`, created a `recognition` instance lazily, and bound it to a toggle button with an EN/JA language switch in `ChatInput.tsx`.
- Weather retrieval: I used Open‑Meteo’s free endpoints (`lib/weather.ts`) for both geocoding and forecast. It’s fast and reliable, and it keeps the project free to run.
- Proposal generation by generative AI: I call Groq’s Chat Completions in a server action (`app/actions/ai.ts`). I send a compacted weather payload to keep latency low and responses focused.
---
## Performance choices I made
- Compact prompt: I slice hourly to 12 and daily to 3 days in the AI prompt to minimize tokens.
- Server actions with `next: { revalidate }`: I cache weather where safe and disable caching for AI for snappy, fresh responses.
- Early returns and clear control flow: Less nesting in `useChat()` for readability and speed.
---
## Configuration and customization
- To default to Japanese replies for Japanese input, I detect Japanese characters and set the system prompt accordingly.
- You can change the model via `GROQ_MODEL` without code changes.
- The features UI comes from `components` and can be adapted to other themes (fashion, sports, etc.).
---
## Contributing
I’m actively improving this project. Check out CONTRIBUTING.md for setup, branch naming, and commit conventions. PRs that improve voice reliability, UX, or add safety heuristics are very welcome.
---
## Security
Please read SECURITY.md to report vulnerabilities. Do not open public issues for sensitive reports.
---
## License
MIT — see LICENSE for details.
---
## Acknowledgements
- Open‑Meteo for free weather data
- Groq for blazing‑fast LLM inference
- The hiking community in Japan for trail inspiration