TravelAI.Core 2.0.1

TravelAI.Core

A .NET 10 library for building AI-native travel platforms. Supports OpenAI, Anthropic, Ollama, and Azure OpenAI out of the box - swap providers without changing any application code. Deployed to Azure Kubernetes Service with GitHub Actions CI/CD.

Installation

dotnet add package TravelAI.Core

Quick start

Zero credentials - try it instantly

builder.Services.AddTravelAI(options => options.UseMock());

No API keys, no Azure account needed. Returns realistic itineraries, destinations, and booking results straight away. Good for local development and testing before committing to a provider.

OpenAI

builder.Services.AddTravelAI(options =>
    options.UseOpenAI("sk-..."));

Anthropic (Claude)

builder.Services.AddTravelAI(options =>
    options.UseAnthropic("sk-ant-..."));

Ollama - free and local

builder.Services.AddTravelAI(options =>
    options.UseOllama("http://localhost:11434", "llama3.2"));

Azure OpenAI

builder.Services.AddTravelAI(options =>
    options.UseAzureOpenAI(
        "https://your-resource.openai.azure.com/",
        "YOUR_KEY",
        "gpt-4o"));

The IItineraryGenerationService, IPriceAnomalyDetector, and IDestinationSearchService interfaces stay identical across all providers. Swap backends without changing any other code.

What it does

Usage

// Generate a personalised itinerary
var itinerary = await generator.GenerateAsync(
    new TravellerProfile
    {
        Name = "Aftab",
        Email = "a@example.com",
        Preferences = ["food", "history", "architecture"],
        Tier = TravelTier.Premium
    },
    destination: "Rome, Italy",
    departure: DateOnly.Parse("2025-08-01"),
    returnDate: DateOnly.Parse("2025-08-08"),
    additionalInstructions: "Avoid tourist traps, focus on local food");

// Semantic destination search
var results = await search.SearchAsync(
    "warm Mediterranean with history and local food", maxResults: 5);

// Detect price anomalies in flight results
await foreach (var flight in detector.AnalyseBatchAsync(flights))
{
    if (flight.PriceAnomaly?.Type == AnomalyType.UnexpectedDeal)
        Console.WriteLine($"Great deal: {flight.Origin}→{flight.Destination} at £{flight.PriceGbp}");
}

// Execute a booking
var result = await bookingService.BookAsync(confirmedItinerary);
Console.WriteLine(result.BookingReference);

Architecture

TravelAI.Core
├── Interfaces/       # IItineraryGenerationService, IPriceAnomalyDetector, ...
├── Models/           # Itinerary, FlightOption, PriceAnomaly, BookingResult, ...
├── Providers/        # ILlmProvider - OpenAI, Anthropic, Ollama, AzureOpenAI, Mock
├── Services/         # Provider-agnostic service implementations
├── Middleware/       # Observability (correlation IDs, structured logging) + rate limiting
├── HealthChecks/     # Azure OpenAI and AI Search health probes
└── Extensions/       # AddTravelAI() DI registration with fluent provider config

samples/TravelAI.Api  # Full ASP.NET Core 10 minimal API
tests/                # xUnit + FluentAssertions - unit and integration tests
deploy/               # Dockerfile, Kubernetes manifests, GitHub Actions CI/CD

Provider architecture:

IItineraryGenerationService
        │
        ▼
  ILlmProvider
        │
   ┌────┴──────────────────────────────────────┐
   │           │            │         │        │
 OpenAI   Anthropic      Ollama   AzureOpenAI Mock

API endpoints

Deployment

Deployed to Azure Kubernetes Service via GitHub Actions. Every push to main builds, tests, pushes the Docker image to GHCR, and deploys with a manual approval gate before production.

# Provision all Azure resources from scratch
bash deploy/scripts/provision-azure.sh

# Push to main to trigger the pipeline
git push origin main

See deploy/DEPLOYMENT.md for the full guide.

Running tests

dotnet test --configuration Release

Tests cover the price anomaly detector, booking orchestration, mock provider, domain model validation, and API integration tests via WebApplicationFactory. No Azure credentials needed - the mock provider and test doubles handle everything.

What's new in v2.0.0

• Multi-provider support - OpenAI, Anthropic (Claude), Ollama, Azure OpenAI
• Mock provider - works with zero credentials, perfect for local development
• Fluent configuration API - options.UseOpenAI(...) replaces config sections
• Provider-agnostic destination search with mock search included
• All existing interfaces unchanged - drop-in upgrade from v1.0.0

Tech stack

.NET 10 · OpenAI · Anthropic · Ollama · Azure OpenAI · Azure AI Search · AKS · Docker · GitHub Actions · xUnit · FluentAssertions

License

MIT