# Harvey — How We Made Review Tables Faster and More Reliable

- Company: Harvey (harvey.ai)
- Announced: 2026-10-22T16:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/making-review-tables-faster-and-more-reliable
- Record: https://forck.live/items/26689-how-we-made-review-tables-faster-and-more-reliable
- Subject: Harvey

Harvey engineers redesigned how Review Tables load data, fetching rows page by page so tables spanning thousands of documents open faster and stay responsive. Review Table is Harvey's inference engine for analyzing thousands of legal documents, returning answers in a structured table where each cell carries the model's reasoning and citations from the source document. Previously all table operations ran in the browser, which loaded the entire table from the blob store and did sorting, filtering and aggregation in memory; once tables exceeded thousands of rows, teams saw lag and, in the largest cases, tables could fail to load entirely. The team moved those operations to the server, keeping Postgres as the source of truth while adding Turbopuffer, a vector data store running on existing Azure blobs, as the read interface in a CQRS pattern. Because Turbopuffer does not support pagination out of the box, Harvey used ID-based pagination: the backend returns ordered row ids, and the client requests data for an arbitrary list of ids. To handle columns like currencies and dates, the team dual-projected strict data types alongside a String field, and when schema drift is detected it falls back to Postgres logic, with drifted schemas healed by a backfill job.

## Evidence

Verbatim from https://www.harvey.ai/blog/making-review-tables-faster-and-more-reliable:

> Historically, all table operations were performed on the client side. The browser loaded the entire Review Table from the blob store and performed sorting, filtering, and any sort of aggregation in-memory. While this solution worked for small Review Tables, we saw performance issues once the table exceeded thousands of rows.

---

Record: https://forck.live/items/26689-how-we-made-review-tables-faster-and-more-reliable
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md