🌍 AI Travel Concierge (Agent Framework + Redis + Mem0)

A travel planning assistant with dual-layer memory: Redis-backed chat history and Mem0-powered long‑term memory. It provides time‑aware research via Tavily, uses OpenAI models for planning, and can export finalized itineraries to an ICS calendar file, all wrapped in a polished Gradio UI with per‑user contexts.

🧠 Key features

• Dual-layer memory: Short‑term chat history in Redis; long‑term preferences via Mem0 (OpenAI LLM + embeddings)
• Per‑user isolation: Separate memory contexts and chat history for each user
• Time‑aware search: Tavily integration for logistics (flights/hotels/transport) and destination research
• Calendar export (ICS): Generate calendar files for itineraries and open the folder via UI
• Gradio UI: Chat, user tabs, live agent event logs, clear‑chat control
• Configurable: Pydantic settings via environment variables, .env support

🧩 Architecture overview

• gradio_app.py: Launches the Gradio app, builds UI, wires event streaming, calendar open, and user switching
• agent.py: Implements TravelAgent using Agent Framework
  • Tools: search_logistics, search_general, generate_calendar_ics
  • Mem0 long‑term memory per user; Redis chat message store for short‑term context
  • Tavily search/extract for fresh web info; ICS generation via ics
• config.py: Pydantic settings and dependency checks
• context/seed.json: Seeded users and initial long‑term memory entries
• assets/styles.css: Custom theme and styling

A diagram of a software company:

✅ Prerequisites

• Python >=3.11
• Redis instance (local Docker, Redis Cloud, or Azure Managed Redis)
• API keys: OpenAI, Tavily, Mem0

🔐 Required environment variables

Provide via your environment or a .env file in the project root. Minimum required:

• OPENAI_API_KEY (must start with sk-; validated)
• TAVILY_API_KEY
• MEM0_CLOUD (default false). If true, you must set MEM0_API_KEY and Mem0 Cloud will be used. If false, local Mem0 runs with Redis vector store.
• MEM0_API_KEY (required only when MEM0_CLOUD=true)

Recommended/optional overrides (defaults shown):

• TRAVEL_AGENT_MODEL = gpt-4o-mini
• MEM0_MODEL = gpt-4o-mini
• MEM0_EMBEDDING_MODEL = text-embedding-3-small
• MEM0_EMBDDING_MODEL_DIMS = 1536
• REDIS_URL = redis://localhost:6379
• MAX_CHAT_HISTORY_SIZE = 6
• MAX_SEARCH_RESULTS = 5
• SERVER_NAME = 0.0.0.0
• SERVER_PORT = 7860
• SHARE = false

Example .env template:

OPENAI_API_KEY=sk-...
TAVILY_API_KEY=...
MEM0_CLOUD=false
# If using Mem0 Cloud, set the API key and flip the flag above to true
# MEM0_API_KEY=...
REDIS_URL=redis://localhost:6379
TRAVEL_AGENT_MODEL=gpt-4o-mini
MEM0_MODEL=gpt-5-nano
MEM0_EMBEDDING_MODEL=text-embedding-3-small
MEM0_EMBDDING_MODEL_DIMS=1536
MAX_CHAT_HISTORY_SIZE=6
MAX_SEARCH_RESULTS=5
SERVER_NAME=0.0.0.0
SERVER_PORT=7860
SHARE=false

🧠 Mem0 modes

• Local (default): MEM0_CLOUD=false
  • Uses Redis as the vector store defined by REDIS_URL
  • Embeddings and LLM calls use your OpenAI key and the configured models
• Cloud: MEM0_CLOUD=true
  • Uses Mem0 Cloud. You must set MEM0_API_KEY
  • No Redis vector store is used for long‑term memory (Redis is still used for chat history)

🗄️ Redis setup options

• Azure Managed Redis: This is an easy way to get a fully managed service that runs natively on Azure. You will require an Azure subscription to get started. Achieve unmatched performance with costs as low as USD 12 per month. Alternative methods for deploying Redis are outlined below. See quickstart guide through Azure portal: https://learn.microsoft.com/en-us/azure/redis/quickstart-create-managed-redis

  Note: Enable access key-based authentication on your cache if you want to run the agent locally. Note the endpoint URL and the password; this will be used later.

• Local (Docker):

docker run --name redis -p 6379:6379 -d redis:8.0.3

• Redis Cloud: create a free database and set REDIS_URL

To clear all app data in Redis (chat history, summaries):

make redis-clear

▶️ Install & run (uv)

This project uses uv for environment and dependency management.

# Install uv (if not installed)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install brew
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

# Install dotenv
brew install dotenv

# From the project root
echo "Creating and syncing environment..."
uv sync

# Create .env from example (then fill in values)
cp -n .env.example .env 2>/dev/null || true

# Start the app (opens browser)
dotenv run -- uv run gradio_app.py

The app launches at http://localhost:7860.

👤 Seed users and memory

• Users are defined in context/seed.json under user_memories
• On first run, each user's long‑term memory is seeded via Mem0
• The default selected user is the first key in seed.json
• Switch users via the tabs at the top of the UI

Example context/seed.json:

{
  "user_memories": {
    "Alice": [ { "insight": "Prefers boutique hotels and walkable neighborhoods" } ],
    "Bob": [ { "insight": "Loves food tours and early morning flights" } ]
  }
}

💬 Using the app

• Ask for trip ideas, date‑bound logistics, or destination research
• The agent will call tools as needed:
  • search_logistics(query, start_date?, end_date?) for flights/hotels/transport
  • search_general(query) for activities, neighborhoods, dining, events
  • generate_calendar_ics(...) once your itinerary is finalized to produce an .ics file
• The right panel shows live agent events and tool logs
• Use “Clear Chat” to wipe the current user’s short‑term history from Redis

📅 Calendar export

• When an itinerary is finalized, the agent can export an .ics file
• Click “Open Calendar” to open the per‑user calendars folder in your OS file explorer
• Files are stored under assets/calendars/<USER_ID>/

Recommended user flow

Try the following query flow to see the agent in action:

Blank User (Mark) - No preferences (cold start case)

1. I like bears and the places where bears can be found.
2. I think Christmas is quite a magical time of year and I enjoy visiting winter wonderland villages
3. What's a place to visit that I might like?
4. Can you give me a solid 3 day itinerary based on online recommendations? For what to do in (Pick one of the locations it mentions)
5. Okay, can you recommend a flight and hotel? I live in (Pick a location) and I'm going to leave on 22nd December and will return 26th December.
6. Give me the calendar.

Pre-filled User (Shreya) - User preferences known at startup

1. I'm planning to hit up Redis Released in London this year. Can you find more info about it?

🐛 Troubleshooting

• Missing API keys: app exits with a configuration error and hints for .env
• OpenAI key must start with sk- (validated in config.py)
• Redis connection errors: verify REDIS_URL and that Redis is reachable
• Mem0 errors when seeding: check MEM0_API_KEY and OpenAI settings
• Browser doesn’t open: navigate to http://localhost:7860 manually

---

Built with Redis, Agent Framework, OpenAI, and Tavily. Enjoy!