Video Upload Capabilities
Multimodal video processing - August 2025
Video Upload Capabilities
Multimodal video processing - August 2025
We’re expanding our multimodal capabilities to include video uploads, targeting availability by the end of August 2025.This feature will enable:
- Video Content Analysis: Upload and analyze video files directly through the API
- Frame-by-Frame Processing: Extract insights from video content at multiple time points
- Visual Scene Understanding: Analyze visual elements, objects, and activities in video content
- Multimodal Search: Search for information based on both visual and audio elements from your uploaded videos
File Search & Connectors
Expanded data access
File Search & Connectors
Expanded data access
Expanding your ability to access and process information:
- Search across various file types
- Integration with external data sources
- Parity with enterprise connector capabilities
Expanded Structured Outputs Support
Enhanced response formatting
Expanded Structured Outputs Support
Enhanced response formatting
We’re expanding our structured outputs capabilities:
- Universal JSON Support: JSON structured outputs are now available for all models across all tiers
- Complete Regex Support: Extending regex pattern support to all models (currently available for
sonar
andsonar-reasoning
) - Advanced Schema Validation: Enhanced validation options for more reliable structured data
- Output Templates: Pre-defined templates for common use cases
Documentation Overhaul
More comprehensive resources
Documentation Overhaul
More comprehensive resources
We’re committed to making our documentation a truly exceptional resource:
- Improved Layout: Optimized structure for better self-serve workflows
- Consistent Design: Unified branding between our landing page and documentation
- Enhanced Content: Revised explanations for better developer understanding
- Vertical-Specific Guides: Specialized documentation for publishers, merchants, and other industry partners
- High-Fidelity Demos: More comprehensive examples beyond our basic cookbook demos
- Comprehensive Prompt Guide: In-depth resources to help you build effectively with our search + LLM API
- API Selection Guides: Clear guidance on which API is best suited for specific scenarios
Increased Rate Limits
Support for high-volume applications
Increased Rate Limits
Support for high-volume applications
We’re scaling our infrastructure to support significantly higher request volumes:
- Working toward supporting use cases up to 100K RPM
- Designed for high-growth startups and enterprise applications
Better Error Handling
Clearer troubleshooting
Better Error Handling
Clearer troubleshooting
Improving our error framework:
- More descriptive error messages
- Clearer guidance on resolving common issues
Multimedia Capabilities
Beyond text processing
Multimedia Capabilities
Beyond text processing
Broadening the types of content you can work with:
- URL Content Integration: Specify URLs within prompts to search for and analyze content from specific web pages directly
- URL Parsing: Extract and analyze content from web pages
- PDF/DOCX Uploads: Process and analyze PDF and DOCX documents directly via the API
Search as its own API
Standalone search functionality
Search as its own API
Standalone search functionality
We’re working on exposing search results as a standalone API/product, separate from the end-to-end integration with the LLM.This feature will allow developers to:
- Directly access search results without invoking the LLM.
- Build custom workflows and applications leveraging search capabilities independently.
Context Management / Memory
Improved context handling
Context Management / Memory
Improved context handling
We’re addressing the limitations of managing context in API calls by introducing new context management features.Key improvements include:
- Efficient Context Storage: Avoid appending responses from previous API calls to new ones, reducing the risk of exceeding context windows.
- Session-Based Memory: Enable session-based memory to maintain context across multiple API calls without manual intervention.