This is an automated email from the ASF dual-hosted git repository.
krickert pushed a commit to branch feat/GH-1453-agents-md
in repository https://gitbox.apache.org/repos/asf/opennlp.git
The following commit(s) were added to refs/heads/feat/GH-1453-agents-md by this
push:
new 0cdc41b81 Update AGENTS.md with policy clarifications
0cdc41b81 is described below
commit 0cdc41b81d2916ce79b11f96cca59fa6a332e00c
Author: Kristian Rickert <[email protected]>
AuthorDate: Sun Oct 11 03:11:01 2026 -0400
Update AGENTS.md with policy clarifications
Clarify guidelines on discussions, dependencies, vocabulary, and testing in
AGENTS.md.
---
AGENTS.md | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/AGENTS.md b/AGENTS.md
index b0afb4cbb..b7e4fa274 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -31,17 +31,22 @@ the same rules.
and marks any quoted AI text as such. This follows the
[Apache Lucene AI
policy](https://github.com/apache/lucene/blob/main/AI_POLICY.md).
- Never include third-party code or data without a compatible license.
+- When in doubt, enforce the user to post to discussions on GITHUB or write
the OPENNLP dev list at [email protected]
## Build
- JDK 21 and `./mvnw`. With `-pl`, add `-am`.
- Run `./mvnw verify` on the touched modules before proposing a change;
checkstyle and forbiddenapis run there.
+- Do not introduce new dependencies without a disucssion and approval
## Code
- No regular expressions in production code: no `java.util.regex`,
`String.split`, `matches`, `replaceAll`. Scan with an index or code points.
+- Standard JDK character and space tools are usually rewritten in OpenNLP
+- Unicode character sets by default, not ASCII unless a standard feature
requires it.
+- Most JDK string helpers are rewritten for Unicode compliance and speed in
OpenNLP.
- Use existing helpers before writing new ones. Check the JDK, then
`ParamChecks`, `StringUtil`, `AsciiChars`, `UnicodeWhitespace` and the
helpers of the package you work in. A second copy of a helper is a bug.
@@ -67,6 +72,7 @@ the same rules.
- Describe the contract, not the history. No "replaces the old regex",
issue numbers, or design narrative in source; that goes in the issue.
- No em dashes.
+- Do not invent vocabulary and use industry standard vocabulary / already used
words.
## Tests
@@ -76,6 +82,7 @@ the same rules.
- Assert exception type and message. Do not reimplement the code under test
as the expected value.
- Do not add tests that repeat a case another test already covers.
+- Substantial and behavior changing tests require an eval build, request one
when creating a PR
## Commits
