Darius Baruo
Jun 15, 2026 15:18
Harvey AI revamped Vault uploads utilizing presigned URLs, boosting pace by as much as 57% for big batches whereas enhancing reliability for authorized groups.
Harvey AI has overhauled its Vault file add system, implementing presigned URLs to streamline efficiency and reliability at scale. With authorized groups more and more counting on Vault for AI-powered file administration, weekly uploads surged from 2.2 million in January to fifteen million by Could 2026. This improve arrives as Vault surpassed 200 million lively information, doubling its rely in simply two months.
The important thing change? Shifting away from server-proxied uploads. Beforehand, information handed via Harvey’s backend earlier than reaching Azure Blob Storage, creating bottlenecks as authorized groups uploaded hundreds of information in giant batches. By adopting presigned URLs, file information now flows straight from the browser to Azure, whereas the backend handles metadata and orchestration. This shift has considerably lowered system pressure, improved efficiency, and enhanced scalability for enterprise use instances.
What Are Presigned URLs?
Presigned URLs are momentary, cryptographically signed hyperlinks that grant time-limited entry to add or obtain information on to object storage techniques like Azure Blob Storage or Amazon S3. This strategy eliminates the necessity for utility servers to deal with file information straight, decreasing latency and backend useful resource utilization.
Harvey AI’s implementation includes a three-step course of:
- Initialization: The backend validates person permissions and generates short-lived add URLs for the browser.
- Add: Information are despatched on to Azure Blob Storage utilizing the URLs, bypassing the backend completely.
- Finalization: The backend verifies the add completion and initiates downstream doc processing.
This structure aligns with greatest practices seen in different enterprise purposes, enhancing each efficiency and reliability.
Efficiency Positive aspects in Actual-World Use
The adoption of presigned URLs has delivered measurable enhancements for Vault uploads. Common latency for uploads dropped by 13% (from 6.7 seconds to five.9 seconds), whereas high-percentile latencies (P99) improved by 27% (from 1m 25s to 1m 2s). For big workloads, comparable to 1,000-file batches, common add length fell by 57%, dropping from 2 minutes 35 seconds to simply 1 minute 6 seconds.
Even large-file uploads noticed good points: common durations decreased by 35%, enhancing from 1 minute 17 seconds to 50 seconds. The outcomes underscore the structure’s scalability, particularly for high-volume enterprise environments like authorized corporations managing huge doc archives.
Enterprise-Prepared Reliability
Past pace, the adjustments tackle the reliability challenges of enterprise deployments. Presigned URLs reduce backend dependency, decreasing dangers from server restarts or timeouts throughout uploads. Nonetheless, Harvey additionally applied fallback mechanisms to deal with restrictive company networks, comparable to VPNs or firewalls, which could block direct-to-storage uploads. In such instances, the system reverts to the legacy server-proxied path seamlessly.
On the backend, the structure now makes use of batched operations for duties like duplicate dealing with and database transactions, guaranteeing clean coordination even with hundreds of information in flight. In the meantime, failed uploads are categorized and retried intelligently, stopping remoted errors from derailing giant batches.
Why It Issues
For authorized groups, the revamped system interprets into quicker uploads, fewer errors, and a extra predictable expertise when managing vital paperwork. As Vault scales to assist even bigger workloads, the presigned URL strategy ensures the system stays environment friendly and resilient.
Harvey AI isn’t alone in adopting this technique. Presigned URLs are a confirmed resolution for large-scale uploads, extensively utilized in techniques like Amazon S3 and Azure Blob Storage. Their capability to dump information switch from backends to storage suppliers makes them indispensable for enterprise purposes dealing with excessive information volumes.
Wanting forward, Harvey AI’s enhancements place Vault to deal with the following wave of progress in enterprise file administration, notably as adoption of AI-powered instruments continues to broaden in industries like authorized and finance.
Picture supply: Shutterstock