host.today/ht-docker-ai
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(tests): switch vision tests to multi-query extraction (count then per-row/field queries) and add logging/summaries

Browse Source
This commit is contained in:
Jürgen Kunz
2026-01-18 11:26:38 +00:00
parent 4c368dfef9
commit 76b21f1f7b
5 changed files with 624 additions and 367 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
changelog.md
View File

@@ -1,5 +1,14 @@
# Changelog
## 2026-01-18 - 1.12.0 - feat(tests)
switch vision tests to multi-query extraction (count then per-row/field queries) and add logging/summaries
- Replace streaming + consensus pipeline with multi-query approach: count rows per page, then query each transaction/field individually (batched parallel queries).
- Introduce unified helpers (queryVision / queryField / getTransaction / countTransactions) and simplify Ollama requests (stream:false, reduced num_predict, /no_think prompts).
- Improve parsing and normalization for amounts (European formats), invoice numbers, dates and currency extraction.
- Adjust model checks to look for generic 'minicpm' and update test names/messages; add pass/fail counters and a summary test output.
- Remove previous consensus voting and streaming JSON accumulation logic, and add immediate per-transaction logging and batching.
## 2026-01-18 - 1.11.0 - feat(vision)
process pages separately and make Qwen3-VL vision extraction more robust; add per-page parsing, safer JSON handling, reduced token usage, and multi-query invoice extraction