Skip to content

feat: implement Document AI functions & support custom IK dictionaries#1064

Open
breeze-wink wants to merge 17 commits into
oceanbase:vldb_2026from
breeze-wink:vldb_2026
Open

feat: implement Document AI functions & support custom IK dictionaries#1064
breeze-wink wants to merge 17 commits into
oceanbase:vldb_2026from
breeze-wink:vldb_2026

Conversation

@breeze-wink

@breeze-wink breeze-wink commented Jul 14, 2026

Copy link
Copy Markdown

Task Description

2026数据库暑期学校测试:湘桂粉丝队

Release Note

Add LOAD_FILE support for reading BLOB data from local locations and implement AI_SPLIT_DOCUMENT as a table function with text, markdown, sentence, word, max, and overlap handling.

Register the parser, resolver, expression, and execution paths, and add a one-command runner for the official mysqltests.

Tests: ./tools/deploy/run_document_ai_tests.sh --no-build
@LINxiansheng

LINxiansheng commented Jul 14, 2026

Copy link
Copy Markdown
Member

Document AI & IK Custom Dictionary Score

Document AI Functions Score
===========================
score: 100.00 / 100
load_file: 50 / 50
ai_split_document: 50 / 50

IK Custom Dictionary Score
==========================
score: 100.00 / 100
ik_custom_dict: 100 / 100

FTS Large Benchmark Score

FTS Large Benchmark Score
=========================
score: 100.00 / 100
mean_improvement: 55.07%
full_score_improvement: 50.00%

build_improvement: 44.51%
  build_ik_all_sec: baseline=35.2836, current=19.499, improvement=44.74%
  build_ik_content_sec: baseline=28.3764, current=15.281, improvement=46.15%
  build_beng_en_sec: baseline=14.7578, current=8.466, improvement=42.63%
tokenize_improvement: 46.55%
  tokenize_ik_avg_ms: baseline=0.76478, current=0.3415, improvement=55.35%
  tokenize_beng_avg_ms: baseline=0.42262, current=0.2631, improvement=37.75%
query_improvement: 74.15%
  query_cn_avg_ms: baseline=16.6628, current=5.3184, improvement=68.08%
  query_beng_avg_ms: baseline=24.3042, current=7.3896, improvement=69.60%
  query_mixed_avg_ms: baseline=17.5593, current=5.7707, improvement=67.14%
  query_limit_avg_ms: baseline=16.2334, current=1.3352, improvement=91.77%

FTS Large Benchmark Report

========================================
FTS Large Benchmark Report
========================================
timestamp:              2026-07-17 16:28:29
label:                  vldb-ci-29593716819-1
git_head:               e690d7a
git_dirty:              0
rows:                   20000
batch:                  500
rounds:                 3000
query_rounds:           200
samples:                3
warmup:                 30
skip_load:              0
----------------------------------------
select1_avg_ms:         0.2043
select1_stdev_ms:       0.0037
raw_load_sec:           1.464
raw_load_rows_per_sec:  13661.2
build_ik_all_sec:       19.499
build_ik_content_sec:   15.281
build_beng_en_sec:      8.466
build_total_sec:        43.263
----------------------------------------
tokenize_ik_avg_ms:     0.3415
tokenize_ik_median_ms:  0.3617
tokenize_ik_stdev_ms:   0.0293
tokenize_beng_avg_ms:   0.2631
tokenize_beng_median_ms:0.2639
tokenize_beng_stdev_ms: 0.0022
----------------------------------------
query_cn_hits:          8001
query_cn_avg_ms:        5.3184
query_cn_stdev_ms:      0.0010
query_beng_hits:        11000
query_beng_avg_ms:      7.3896
query_beng_stdev_ms:    0.0041
query_mixed_hits:       7332
query_mixed_avg_ms:     5.7707
query_mixed_stdev_ms:   0.0031
query_limit_hits:       20
query_limit_avg_ms:     1.3352
query_limit_stdev_ms:   0.0017
========================================

Workflow run

@breeze-wink breeze-wink reopened this Jul 15, 2026
@breeze-wink breeze-wink changed the title feat: implement Document AI functions feat: implement Document AI functions & support custom IK dictionaries Jul 15, 2026
@breeze-wink

breeze-wink commented Jul 16, 2026

Copy link
Copy Markdown
Author

Hunan-Guangxi Fans Team

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants