- Website: www.unwind.dev
Travel planning is vital for a great vacation, but it can be a tedious and frustrating process. Sifting through maps and countless online resources is not only time-consuming but often yields less-than-ideal results. Juggling time and budget constraints adds another layer of complexity. Our goal is to streamline this process, empowering travelers to create unforgettable vacations that fit their budgets and schedules. Our planning APIs simplify the experience by allowing users to input their desired destination, date, and time preferences. The service then generates customized travel plans tailored to their needs. While the initial version focuses on single-day trips and generic rankings, future iterations will incorporate personalized recommendations.
- Save your favorite plans in your profile
- View trip details
- Make a plan yourself by creating a template
Two endpoints let a caller add a place to the shared cache by name, as an alternative to the category-based nearby-places search: search by free text, then confirm one result into the cache.
Runs a free-text Google Places search around a coordinate and returns every result as a confirmable candidate; nothing is written to the shared cache by this call. Every result (insertable or not) is stashed server-side under its Google place ID for 30 minutes so a subsequent confirm call can resolve it by ID alone rather than trusting place data an HTTP caller might send back.
Request:
{
"query": "Joe's Pizza",
"location": {"latitude": 40.7309, "longitude": -74.0021},
"radius": 5000,
"limit": 10
}query is required (2-120 characters). location (latitude/longitude) is required, with
no default — same as the nearby-places endpoints' own zero-location rejection — because an
unbiased text query like "konjoe" can resolve to the wrong continent without a coordinate to
anchor it. radius is in meters and is clamped to the service's max search radius (16,000 m /
~10 miles) when zero or larger than that maximum. limit defaults to 10 and is capped at 20.
Response (200, fields elided for brevity):
{
"results": [
{
"place": {
"ID": "ChIJd8BlQ2BZwokRAFUEcm_qrcA",
"Name": "Joe's Pizza",
"Status": "OPERATIONAL",
"LocationType": "restaurant",
"Types": ["restaurant", "food", "point_of_interest", "establishment"],
"FormattedAddress": "7 Carmine St, New York, NY 10014",
"Location": {"latitude": 40.7309, "longitude": -74.0021, "city": "", "adminAreaLevelOne": "", "country": ""},
"PriceLevel": 1,
"Rating": 4.5,
"UserRatingsTotal": 3200
},
"category": "Eatery",
"insertable": true
}
]
}category and insertable are always derived server-side from the place's own Google types.
A candidate whose primary type does not map to a known category is still returned (so the
caller can see it), but with category: "" and insertable: false.
Inserts a previously returned candidate into the shared cache — placeIDs:<category> plus a
place_details:place_ID:* record — making one Place Details call to fill in hours, address,
URL, summary, and (as a gap-fill only, never overwriting an existing photo) photo before
writing.
Request:
{"placeId": "ChIJd8BlQ2BZwokRAFUEcm_qrcA"}Response (200):
{
"place": { "ID": "ChIJd8BlQ2BZwokRAFUEcm_qrcA", "...": "same shape as place-search's place object, now enriched with hours/URL/summary" },
"category": "Eatery",
"alreadyCached": false
}Error responses:
404{"error": "...", "code": "candidate_expired"}— the place ID was never searched, or its 30-minute stash entry expired.422{"error": "...", "code": "unsupported_place_type", "placeType": "<google type>"}— the candidate's primary Google type does not map to any category; nothing is written.
Confirming a place writes it into the shared Redis cache immediately — you can verify this
directly, without waiting on any other endpoint: the confirm response itself echoes the
written place, and ZSCORE placeIDs:<category> <placeID> against Redis returns a score right
away.
Whether the place then shows up in /v1/nearby-places-by-category depends on that cell's
search freshness, not on the write above. In a warm cell (one whose MapsLastSearchTime
marker is still fresh), the confirmed place appears on the very next read. In a cold or
stale cell, the next read triggers a background Google search whose results replace — not
merge with — the cached bucket in that response, so the newly confirmed place is briefly
missing from that one response and only appears from the following read onward. When
verifying a confirm in a cell you are not sure is warm, check ZSCORE first and expect to
need up to two reads of /v1/nearby-places-by-category before the place shows up.
Never manually stamp a cell's MapsLastSearchTime to force this — doing so marks the cell
"searched" for 14 days and would suppress a real cold search the cell still needs.
Both endpoints require the same authentication as the other /v1 endpoints: a Personal
Access Token via Authorization: Bearer <token>, or a JWT session cookie as a browser
fallback.
The category a place lands under is always computed server-side from Google's own primary type on the place, never accepted from the caller, and a primary type that maps to no known category is refused outright rather than defaulted into some bucket — so nothing client-supplied ever reaches the shared place cache unclassified.
- git clone the repository
- update Homebrew with
brew update - Install Redis using Homebrew with
brew install redis- If Redis is already installed, execute
brew upgrade redis
- If Redis is already installed, execute
- Obtain Google Maps API key and set the
MAPS_CLIENT_API_KEY=YOUR_GCP_API_KEY,REDISCLOUD_URL=redis://localhost:6379environment variables. - Set environment variable
ENVIRONMENT=DEVELOPMENT,SENDGRID_API_KEY=NO_KEYas we do not create mailers in development environment. - Start (in background) Redis service with
brew services start redis. - Execute
go run main/main.goto start the server.
- Use command
docker-compose up -dto start the containers. - Make sure to set up environment variables
REDIS_URL=redis://redis:6379andMAPS_CLIENT_API_KEY=YOUR_GCP_API_KEY. Note that usinglocalhostfor redis URL does not work. - To stop the containers, use the command
docker-compose stop.
- The service can be deployed on any service platform. Particularly we have configured the code base and been deploying the service to Heroku.
- For deployment to Heroku, simply execute
git push heroku master.
- Multi-city, multi-day planning
- Redis
- Google Maps API
- GeoNames Web Services
- Backend: Golang
- Frontend: Bootstrap and Javascript
