[const-cloudinary]

Summary

Adds task documentation for AI coding agents that use the SDK, a contributor guide for agents working on this repository, and a restructured README. No source changes.

Item	Purpose
docs/ (16 pages)	Task pages for agents using the SDK: credentials, dependency and first call, configuration, image upload, large/raw upload, signed browser upload, image and video delivery, search and asset management, moderation, structured metadata, Spring Boot, JSP taglib, troubleshooting, platform capabilities
examples/ (4 programs)	Runnable versions of the upload, large-video upload, delivery, and search tasks
AGENTS.md, CLAUDE.md	Contributor guide for agents; CLAUDE.md contains only @AGENTS.md
SECURITY.md	Private reporting via GitHub advisories, plus server-side usage guidance
context7.json	Context7 indexing; same account key as cloudinary_npm and pycloudinary
README.md	Restructured to the same section list as the other Cloudinary SDKs
.gitignore	Artifacts left behind by running the examples

README fixes: the install snippet named cloudinary-http45 with a 2.x version, which has never been published. It now names cloudinary-http5 with an X.Y.Z placeholder and points to Maven Central for the current version. The GitHub issues link pointed at a non-existent org.

Why docs/ is not packaged into a jar

docs/ stays in the repository and is not added to any artifact.

• Of 15 widely used Java libraries on Maven Central (Guava, OkHttp, Jackson, Gson, commons-lang3, SLF4J, Retrofit, AWS SDK v2, google-cloud-storage, Stripe, Twilio, SendGrid, java-jwt, Spring Core, jsoup), only jsoup ships any Markdown: a 3 KB README in META-INF/.
• Embedding the pages would grow cloudinary-core by about 27% (175 KB → 223 KB) for files no runtime reads.
• A resource inside a jar is not a file an agent can open without first locating the jar and unpacking it. Reading the repository at the matching release tag, or Cloudinary's llms.txt / .md documentation URLs, is shorter.

examples/ is not packaged either. Each doc page carries its complete flow inline.

Verification

Check	Result
java snippets compile against the built 2.5.0 jars (fragments wrapped in a harness; Spring and servlet types stubbed)	103/103
Examples compile with --release 8	4/4
Internal links and cross-page anchors	0 errors
External URLs	37/38 return 200; the link to tree/master/docs resolves once this is merged
:cloudinary-http5:test, base (c9920c8) vs branch, each on a fresh product environment	206 tests, 0 failed, 39 skipped on both
:cloudinary-core:test on branch	275 tests, 0 failed

Behaviour described in the pages was observed against live product environments while the pages were being written. The examples were run live, including no-credential and bad-credential runs, before the rebase onto 2.5.0. After the rebase they were compiled, not re-run.

Snippets are reference voice: they show correct usage and state documented behaviour.

Known limitation: the Spring Boot controller snippet uses Map.of, which needs Java 9+. That matches Spring Boot 3 (Java 17).

Needs a maintainer

• Context7: register cloudinary/cloudinary_java at context7.com. context7.json is inert until then.
• Repository metadata: description, homepage, topics, and social preview.
• CONTRIBUTING.md: the old README linked it, but it does not exist. The link was removed, not replaced.
• docs: agent-optimize README + add AGENTS.md/CLAUDE.md and IDE pointer files #388 (outside PR) adds AGENTS.md/CLAUDE.md and IDE pointer files and overlaps with this one. Left open for a maintainer to decide.

🤖 Generated with Claude Code