AI Powered On Site Search Platform

The vendor required to provide AI powered on-site search platform solutions (google custom search, Drupal search API, and Drupal search API + SOLR) that cannot search across sites.
- Provide AI-powered federated search across all university web properties, with capabilities for results curation, AI-generated answers based on university content, and integration with existing university data sources.
- Core search capabilities
•    Federated search across all UO domains and subdomains
•    AI-powered answer generation (similar to google AI summaries and perplexity) that:
o    Appears above traditional search results
o    Generates answers in real-time based solely on indexed UO content
o    Supports bring-your-own-LLM model with UO-managed API keys
o    Allows system prompt customization for control and accuracy
o    Provides mandatory citations with links to source pages
o    Returns single answer format (not conversational chat)
•    Results curation and manual ranking control
•    Current-site content prioritization (automatically boost results from the site where search is performed)
•    Query rules and redirect capabilities
•    Promoted results and pinning for specific queries.
- Search quality and relevance:
•    Result highlighting showing where search terms appear
•    Stemming (searching "running" finds "run", "runs", "runner")
•    "Did you mean?" suggestions for misspellings
•    Related searches or "people also searched for"
•    Zero-results handling with suggestions
- Scale and coverage:
•    Coverage of all domains: uoregon.edu, oregon.edu, goducks.com, uoalumni.com, uofoundation.org and all their subdomains
•    Approximately 250,000 pages across 675+ subdomains
•    Platform-agnostic content indexing (Drupal, WordPress, any web-based content)
•    State historical news archives (13m+ pages) are excluded from this contract
- Tiered indexing schedule:
•    Priority sites: daily or more frequent indexing
•    All other sites: monthly scanning
•    Administrator control over indexing frequency per site
•    Automatic detection and reindexing of changed pages
•    Support for optional webhook triggers from content management systems
- Content control and exclusions:
•    Ability to exclude specific pages, sections, or content types from indexing
•    Manual deindexing capability for outdated content
•    Include and exclude URL pattern rules
- Indexing and metadata:
•    Metadata weighting (prioritize title matches over body text)
•    Content freshness signals (favor recently updated content)
•    Configurable index update frequency
•    Index health monitoring
- API integration:
•    Find people (API endpoint returning json)
•    Find departments (API endpoint returning json)
•    Find majors and degrees (API endpoint returning json)
•    Find labs (API endpoint returning json)
- User access and permissions:
•    Support for 5-10 concurrent users
•    Centralized management (no individual site manager access)
•    No per-user fees
- API capabilities:
•    API-first architecture for UO to build custom search interfaces
•    Support for rest and GraphQL
•    Adequate rate limiting for production search volume
•    Response format customization
•    Webhooks for notifications (indexing completion, errors)
•    Multiple API keys for different environments (dev, staging, production)
•    API versioning for stability
•    Comprehensive API documentation
- Analytics and reporting:
a. Real-time dashboards showing:
•    Search query analytics (what users search, frequency, zero-result queries)
•    Click-through data (which results clicked, position)
•    AI answer usage metrics
•    Performance metrics (response times, uptime, index health)
b. Automated alerts for:
•    Indexing failures
•    High zero-result rates
•    Performance issues
- Performance:
•    Search response time approximately 200ms or better
•    Expected search volume: 50,000 queries per month
•    High platform uptime (not mission-critical, but reliability is expected)
•    Index size support for approximately 250,000 pages
- Testing and quality assurance:
•    Sandbox and staging environment for testing changes before production deployment
- Authentication:
•    Preferred: single sign-on (SSO) via entraid (microsoft azure active directory)
•    Also supported: shibboleth SSO
- implementation and support:
•    Selection by end of 2025, implementation beginning January 2026
•    Full implementation by July 2026 including:
o    Phased rollout starting with one pilot site
o    All sites configured with appropriate indexing schedules
o    User access and roles configured
o    Dashboards and analytics operational
o    API fully documented and accessible
•    Administrator training covering:
o    Platform administration
o    Results curation and query rules
o    AI answer configuration and system prompts
o    Content exclusion and indexing controls
o    Analytics and reporting
o    API usage (documentation-based, supplemented by training)
•    Business hours (8:00am-5:00pm pacific time) support with standard response times (1 business day).

- Contract Period/Term: 9 years
- Voluntary Pre-Bid Meeting Date: November 26, 2025