From f7c3f36f30d8c6b27f7dcbc434bef530c5a2a959 Mon Sep 17 00:00:00 2001 From: parkervg Date: Tue, 9 Apr 2024 20:32:40 -0400 Subject: [PATCH] Better intro, explaining 'blend' name --- README.md | 9 +++++++-- docs/index.md | 9 +++++++-- 2 files changed, 14 insertions(+), 4 deletions(-) diff --git a/README.md b/README.md index 13dfe090..820cf3bd 100644 --- a/README.md +++ b/README.md @@ -19,10 +19,15 @@
## Intro -BlendSQL is a *superset of SQLite* for problem decomposition and hybrid question-answering with LLMs. It builds off of the syntax of SQL to create an intermediate representation for tasks requiring complex reasoning over both structured and unstructured data. +BlendSQL is a *superset of SQLite* for problem decomposition and hybrid question-answering with LLMs. + +As a result, we can *Blend* together... +- 🥤 ...operations over heterogeneous data sources (e.g. structured, unstructured, visual) +- 🥤 ...the structured reasoning of SQL with the generalizable reasoning of LLMs It can be viewed as an inversion of the typical text-to-SQL paradigm, where a user calls a LLM, and the LLM calls a SQL program. -Here, the user is given the control to oversee all calls (LLM + SQL) within a unified query language. + +**Now, the user is given the control to oversee all calls (LLM + SQL) within a unified query language.** ![comparison](docs/img/comparison.jpg) diff --git a/docs/index.md b/docs/index.md index d553e175..528a71a2 100644 --- a/docs/index.md +++ b/docs/index.md @@ -25,10 +25,15 @@ pip install blendsql -BlendSQL is a *superset of SQLite* for problem decomposition and hybrid question-answering with LLMs. It builds off of the syntax of SQL to create an intermediate representation for tasks requiring complex reasoning over both structured and unstructured data. +BlendSQL is a *superset of SQLite* for problem decomposition and hybrid question-answering with LLMs. + +As a result, we can *Blend* together... +- 🥤 ...operations over heterogeneous data sources (e.g. structured, unstructured, visual) +- 🥤 ...the structured reasoning of SQL with the generalizable reasoning of LLMs It can be viewed as an inversion of the typical text-to-SQL paradigm, where a user calls a LLM, and the LLM calls a SQL program. -Here, the user is given the control to oversee all calls (LLM + SQL) within a unified query language. + +**Now, the user is given the control to oversee all calls (LLM + SQL) within a unified query language.** As shown in our paper, using BlendSQL as an intermediate representation for complex reasoning tasks can boost performance by 8.63% and use 35% fewer tokens compared to the naive end-to-end approach. For an example of this approach, see [this notebook](reference/examples/teaching-blendsql-via-in-context-learning).