Enhanced Graph Search Algorithm with Pruning and Adaptive Depth

[IANTHEREAL]

The current vector-distance-based search combined with 2-degree relationship expansion is insufficient for retrieving relevant results from graph. To address this, we propose:

1. Pruning Mechanism:

• Filter out irrelevant nodes and edges during search to reduce noise.

2. Adaptive Search Depth:

• Dynamically adjust the depth of the search based on query complexity and data relevance.

3. Leverage LLMs for Accuracy:

• Use Large Language Models (LLMs) to refine search paths and improve result precision.

[tomtom215]

@IANTHEREAL this would take this project to a whole different level IMHO

[IANTHEREAL]

The experiment result for the draft version, I collected some cases where tidb.ai performed poorly.

• Question: How to pause the index creation in TiDB?

tidb.ai answer: https://tidb.ai/c/0193e31b-a324-72f0-97c5-8876411f361f

• Question: is TiKV part of TiDB? or can it be used separately? I was searching for some KV storage and found TiKV repo, but nothing much there, so I'm wondering if it's part of TiDB.

tidb.ai answer: https://tidb.ai/c/0193e318-f2ca-7e63-bd75-f84621e450af

• Question: TiDB的最大连接数是多少呀?

tidb.ai answer: https://tidb.ai/c/0193e31f-a540-7b4e-ba4b-fbc8ad2e2e56

• Question: What is the latest long-term version of TiDB?

• Question: I see a lot of errors says throwing pseudo region error due to no replica available when running BR, what does it means?