Skip to content

Releases: h2oai/sql-sidekick

v0.2.4

17 Feb 00:56
6edc7c6
Compare
Choose a tag to compare

What's Changed

  • Fix pending UI blocking calls by #44

v0.2.3

08 Feb 23:29
Compare
Choose a tag to compare

What's Changed

  • Pending h2ogpt integration for vulnerability scanning & Q generation #4
  • Fix swapping Q recommender models #4
  • Update system config to support defog/sqlcoder-7b-2
  • Fix pending high-level ui blocking calls #44
  • Update dependencies

v0.2.2

27 Jan 21:15
Compare
Choose a tag to compare
  • Bug fixes when swapping models
  • Notebook updates

v0.2.1

25 Jan 00:15
83dd5b4
Compare
Choose a tag to compare
v0.2.1 Pre-release
Pre-release

What's Changed

  • UI flow improvements
  • Additional bug fixes

v0.2.0

19 Jan 22:42
88973fc
Compare
Choose a tag to compare
v0.2.0 Pre-release
Pre-release

What's Changed

  • Minor fixes
  • Notebook example on using configured guardrails

Full Changelog: v0.1.9...v0.2.0

v0.1.9

19 Jan 10:35
Compare
Choose a tag to compare
v0.1.9 Pre-release
Pre-release

Changed/FIxed/Added

Full Changelog: v0.1.8...v0.1.9

v0.1.8

22 Dec 05:24
Compare
Choose a tag to compare
v0.1.8 Pre-release
Pre-release

Changed/Fixed

  • Pass Db dialect properly #63
  • Add support for sqlcoder-34b-alpha locally #65
  • Self-correction improvements using h2ogpte client (h2oai/h2ogpt-4096-llama2-70b-chat) #68

v0.1.6

27 Nov 20:44
Compare
Choose a tag to compare
v0.1.6 Pre-release
Pre-release

Added:

  1. Enable OpenAI-GPT models for baseline (evaluation/comparison) in eval mode #53
  2. Generating sample Qs for cold start problem (using OSS remote endpoint for h2ogpt) #54
  3. Add introductory data preview #56

Bundle: s3://sql-sidekick/releases/ai.h2o.wave.sql-sidekick.0.1.6.wave

v0.1.5

16 Nov 19:31
Compare
Choose a tag to compare
v0.1.5 Pre-release
Pre-release

Bundle: s3://sql-sidekick/releases/ai.h2o.wave.sql-sidekick.0.1.5.wave

v0.1.4

06 Nov 21:31
Compare
Choose a tag to compare
v0.1.4 Pre-release
Pre-release

Added:

  • Playground/toolchain in place to evaluate generation consistency using zero-shot, 1-shot, and few-shot settings (via CLI / thin web client using wave)
  • 2 SOTA LLM models focused on SQL generation enabled - sqlcoder2 (Default), NumbersStation/nsql-llama-2-7B, GPT-models (not enabled via UI)
  • Basic RAG workflow supported - semantic ranking + relevance ranking to Interact and improve generation
  • Guardrails: Safety checks in place to avoid SELECT type SQL injections
  • Database integration - postgres (briefly disabled via UI)/ SQLite (default) to get to the final answer
  • Auto schema generation on data-upload via UI
  • Different generation modes - Try Again(greedy decoding), Try Harder(Diverse beam search decoding)

Bundle: s3://sql-sidekick/releases/ai.h2o.wave.sql-sidekick.0.1.4.wave