AiKit News

AI search, live sources, deep research, weather, files, and ChatJimmy speed stats

Live App • Repository • Features • Quick Start • Environment • API Routes • Deploy

---

🌟 Overview

AiKit News is a modern AI search and research app built with Next.js, ChatJimmy, and Exa. It can answer directly, search the live web, run multi-step Deep Research, render weather cards, read text/code attachments, keep local chat history, and show per-response generation stats.

Example stats:

241 Tokens   0.018s   13,514 TPS

Powered by:

• ChatJimmy official site
• ChatJimmy reverse API
• Exa AI search

---

✨ Features

• ChatJimmy answers using llama3.1-8B
• Free shared ChatJimmy config prefilled for quick demos
• Generation stats footer with completion tokens, decode time, and TPS
• Exa web search with titles, URLs, highlights, and page text
• Inline citation pills using clean [1], [2], [3] markers with source context
• Deep Research mode that searches, synthesizes, follows up, and writes a cited final answer
• Weather tool for live conditions and short forecast cards
• Text/code file attachments with token counting and context-safe limits
• Tool settings for Search, Weather, custom system prompt, and top_k
• Regenerate and response versions so older answers remain navigable
• Conversation search with highlighted matches and match navigation
• Copy actions for full messages and individual code blocks
• Auto-generated chat titles for cleaner history
• IndexedDB local chat history for saved conversations
• Light, dark, and system themes with refined source hover states
• Markdown rendering with GFM tables, code highlighting, math, and KaTeX
• Responsive app shell with sidebar history and mobile-friendly controls
• Portless dev script for Vercel Portless workflows
• Vercel and Cloudflare-ready scripts

---

🧱 Tech Stack

Layer	Tech
Framework	Next.js 16 App Router
UI	React 19, Tailwind CSS 4, Framer Motion
Language	TypeScript
AI backend	ChatJimmy OpenAI-compatible API
Search	Exa API
Weather	Open-Meteo based weather helper
Markdown	react-markdown, remark-gfm, remark-math, rehype-katex, rehype-highlight
Local storage	IndexedDB
Deployment	Vercel, OpenNext Cloudflare

---

🚀 Quick Start

Requirements

• Node.js 20+
• npm
• Exa API key for web search and Deep Research

Install

git clone https://github.com/tanu360/aikit-news.git
cd aikit-news
npm install
cp .env.example .env.local

Add your Exa key to .env.local, then run:

npm run dev

Open:

http://localhost:3000

Portless Dev

If you use Vercel Portless:

npm run dev:portless

---

🔐 Environment

AiKit needs an Exa key for search. ChatJimmy is prefilled with the free shared key below.

EXA_API_KEY=your_exa_api_key_here

CHATJIMMY_API_URL=https://jimmy.aikit.club/v1/chat/completions
CHATJIMMY_API_KEY=tarun-dummy
CHATJIMMY_MODEL=llama3.1-8B

NEXT_PUBLIC_MODEL_NAME=llama3.1-8B
NEXT_PUBLIC_SITE_URL=https://news.aikit.club

Environment Variables

Variable	Required	Scope	Purpose
EXA_API_KEY	Yes	Server	Used by /api/search and /api/deep-research
CHATJIMMY_API_URL	Yes	Server	OpenAI-compatible ChatJimmy completion endpoint
CHATJIMMY_API_KEY	Yes	Server	ChatJimmy API key or shared demo key
CHATJIMMY_MODEL	Yes	Server	Model sent to ChatJimmy
NEXT_PUBLIC_MODEL_NAME	No	Browser	Model label shown in the input UI
NEXT_PUBLIC_SITE_URL	No	Browser/server metadata	Canonical, Open Graph, and Twitter metadata origin

ChatJimmy Free Config

Use this exact change if your .env.local still has placeholders:

-CHATJIMMY_API_URL=your_chatjimmy_api_url_here
-CHATJIMMY_API_KEY=your_chatjimmy_api_key_here
-CHATJIMMY_MODEL=llama3.1-8B
+CHATJIMMY_API_URL=https://jimmy.aikit.club/v1/chat/completions
+CHATJIMMY_API_KEY=tarun-dummy
+CHATJIMMY_MODEL=llama3.1-8B

tarun-dummy is intentionally shared as a free ChatJimmy key for people trying AiKit News. You can swap it for your own ChatJimmy-compatible endpoint, or run the proxy from tanu360/chatjimmy-reverse-api.

Get an Exa API Key

1. Open exa.ai or the Exa dashboard.
2. Sign in or create an account.
3. Open the API keys section.
4. Create a new API key.
5. Put it in .env.local:

EXA_API_KEY=your_real_exa_api_key_here

Exa is only called from server-side API routes. Keep it in .env.local and do not expose it as a NEXT_PUBLIC_ variable.

---

🧠 How It Works

User prompt
  |
  v
Next.js chat UI
  |
  +--> Direct answer
  |      |
  |      v
  |   /api/chat -> ChatJimmy answer
  |
  +--> Search answer
  |      |
  |      v
  |   Router -> /api/search -> Exa -> /api/chat -> cited answer
  |
  +--> Weather
  |      |
  |      v
  |   ChatJimmy tool call -> weather helper -> weather card + answer
  |
  +--> Deep Research
         |
         v
     /api/deep-research -> search tree -> synthesis -> final cited answer

Generation Stats Formula

AiKit reads ChatJimmy usage metadata and calculates decode time with:

Decode Time = completion_tokens / generation_speed

Then the UI shows completion tokens, time in seconds, and TPS:

241 Tokens   0.018s   13,514 TPS

---

🛠️ API Routes

Method	Route	What it does
POST	/api/chat	Calls ChatJimmy, handles tools, streams UI chunks, emits generation stats
POST	/api/search	Searches Exa with instant results and source text
POST	/api/deep-research	Runs multi-step Exa research and final ChatJimmy synthesis
GET	/api/config	Returns the active model name for the input UI
POST	/api/title	Generates chat titles

Chat Request

curl http://localhost:3000/api/chat \
  -H "Content-Type: application/json" \
  -d '{
    "mode": "answer",
    "messages": [
      { "role": "user", "content": "Explain AI search in simple words" }
    ]
  }'

Search Request

curl http://localhost:3000/api/search \
  -H "Content-Type: application/json" \
  -d '{
    "query": "latest AI search products",
    "toolSettings": { "search": true, "weather": false }
  }'

Deep Research Request

curl http://localhost:3000/api/deep-research \
  -H "Content-Type: application/json" \
  -d '{
    "query": "How are AI agents changing software development?"
  }'

---

🔎 Deep Research

Deep Research turns one prompt into a compact research tree:

1. Search the original query.
2. Synthesize the most useful findings.
3. Generate follow-up questions.
4. Search the follow-ups.
5. Deduplicate sources.
6. Write a final answer with citations.

Stream events include:

Event	Meaning
step_start	A research query started
search_complete	Exa returned sources
synthesizing	ChatJimmy is analyzing the sources
step_done	One research step finished
research_complete	Source gathering finished
answer_start	Final answer generation started
answer_chunk	Final answer text chunk
generation_stats	Tokens, decode time, and TPS
all_sources	Final citation source list
done	Research completed

---

📜 Scripts

Command	Description
npm run dev	Start Next.js dev server
npm run dev:fresh	Clean caches, then start dev server
npm run dev:portless	Clean caches, then run portless run next dev
npm run build	Build the Next.js app
npm run build:fresh	Clean caches, then build
npm run start	Start production server
npm run lint	Run ESLint
npm run build:cloudflare	Build for Cloudflare with OpenNext
npm run preview:cloudflare	Preview Cloudflare build
npm run deploy:cloudflare	Deploy with OpenNext Cloudflare tooling
npm run clean	Remove build/cache artifacts

---

🚀 Deploy

Vercel

1. Push this repo to GitHub.
2. Import tanu360/aikit-news in Vercel.
3. Add the environment variables from Environment.
4. Deploy.

Cloudflare

This repo includes OpenNext Cloudflare support:

• open-next.config.ts
• wrangler.jsonc
• npm run build:cloudflare
• npm run deploy:cloudflare

Add the same environment variables to your Cloudflare deployment before going live.

Production Checklist

• Set NEXT_PUBLIC_SITE_URL to your deployed origin.
• Keep EXA_API_KEY server-only.
• Rotate the shared ChatJimmy key if you operate your own endpoint.
• Run npm run lint.
• Run npm run build.

---

🔒 Security Notes

• Keep .env.local private.
• .env.example is safe to commit because the ChatJimmy key is intentionally public for demos.
• Never put Exa keys in NEXT_PUBLIC_ variables.
• Search and weather are tool-controlled, so the model is prompted not to invent live facts when those tools are disabled.
• Treat attached files as user-provided context. Do not upload secrets or private files to a public deployment.

---

🧯 Troubleshooting

Problem	Fix
Search returns “Search is not configured.”	Add EXA_API_KEY to .env.local and restart the dev server.
Chat returns missing env vars	Make sure CHATJIMMY_API_URL, CHATJIMMY_API_KEY, and CHATJIMMY_MODEL are present.
Model label looks wrong	Set NEXT_PUBLIC_MODEL_NAME or CHATJIMMY_MODEL, then restart.
Port is already in use	Use npm run dev:portless or start Next on another port.
Cloudflare build fails	Run npm run build:cloudflare locally and confirm env vars are available.

---

🤝 Contributing

Contributions are welcome.

1. Fork the repository.
2. Create a feature branch.
3. Keep changes focused.
4. Run npm run lint and npm run build.
5. Open a pull request with screenshots for UI changes.

---

🙏 Credits

• Built by tanu360
• Powered by ChatJimmy
• Search by Exa
• ChatJimmy-compatible backend reference: chatjimmy-reverse-api

---

📄 License

MIT License. See LICENSE.