Swap CI vision test image to author-licensed (MIT) Commons file

The previous default (Red_Apple.jpg) was CC-BY 2.0, not CC0 / public
domain as the surrounding comments claimed - it requires attribution
that the test scaffold did not supply.

Replace with File:20200601_135745_Flowers_and_Bees.jpg from Wikimedia
Commons by Bernard Ladenthin (the project copyright holder), which the
same author additionally grants under MIT for use in this project.
Fetch via Special:FilePath?width=640 so CI pulls a ~80 KB thumbnail
instead of the multi-MB original phone photo.

Updates
- publish.yml VISION_IMAGE_URL / _NAME swapped; the comment block above
  the env vars now records the original Commons file, the author, the
  Commons license (CC-BY-4.0), and the MIT grant to this project.
- All four 'Download CC0 / public-domain test image' step names
  renamed to 'Download MIT-granted test image (by project author, see
  env block)'.
- validate-models.sh OPTIONAL_IMAGES entry updated to the new filename.
- MultimodalIntegrationTest javadoc adds an Image source section
  recording the Commons URL + author + CC-BY-4.0 / MIT dual grant.
- docs/history/49be664_open_issues.md recommended-sequencing entry no
  longer calls the image CC0 / PD.

Verified
- mvn test -Dtest='ContentPartTest,MultimodalMessagesTest,MultimodalIntegrationTest'
  BUILD SUCCESS: 25 unit tests pass, integration test self-skips
  (no -D system properties locally).

Author: claude

.github/validate-models.sh

@@ -26,7 +26,7 @@ OPTIONAL_MODELS=(
 )
 
 OPTIONAL_IMAGES=(
-  "models/Red_Apple.jpg"
+  "models/20200601_135745_Flowers_and_Bees.jpg"
 )
 
 validate_gguf() {

.github/workflows/publish.yml

@@ -30,12 +30,16 @@ env:
   VISION_MODEL_NAME: "SmolVLM-500M-Instruct-Q8_0.gguf"
   VISION_MMPROJ_URL: "https://huggingface.co/ggml-org/SmolVLM-500M-Instruct-GGUF/resolve/main/mmproj-SmolVLM-500M-Instruct-Q8_0.gguf"
   VISION_MMPROJ_NAME: "mmproj-SmolVLM-500M-Instruct-Q8_0.gguf"
-  # Small CC0 / public-domain test image from Wikimedia Commons. A simple
-  # subject (red apple, ~12 KB) so a 500M vision model has a fair chance of
-  # producing recognisable output; the test only asserts a non-empty reply
-  # so model accuracy is not the gating signal.
-  VISION_IMAGE_URL: "https://upload.wikimedia.org/wikipedia/commons/1/15/Red_Apple.jpg"
-  VISION_IMAGE_NAME: "Red_Apple.jpg"
+  # Small test image used by MultimodalIntegrationTest. The original photo is
+  # File:20200601_135745_Flowers_and_Bees.jpg on Wikimedia Commons by
+  # Bernard Ladenthin (the project copyright holder), licensed there as
+  # CC-BY-4.0 but granted MIT for use in this project by the same author
+  # (see commit message + SPDX header on .github/workflows/publish.yml).
+  # We fetch a 640px-wide thumbnail via Special:FilePath (~80 KB) so CI
+  # bandwidth stays bounded; the on-disk filename keeps the original stem
+  # for traceability back to the Commons file.
+  VISION_IMAGE_URL: "https://commons.wikimedia.org/wiki/Special:FilePath/20200601_135745_Flowers_and_Bees.jpg?width=640"
+  VISION_IMAGE_NAME: "20200601_135745_Flowers_and_Bees.jpg"
 permissions:
   contents: read
 jobs:
@@ -411,7 +415,7 @@ jobs:
         run: curl -L --fail ${VISION_MODEL_URL} --create-dirs -o models/${VISION_MODEL_NAME}
       - name: Download vision mmproj
         run: curl -L --fail ${VISION_MMPROJ_URL} --create-dirs -o models/${VISION_MMPROJ_NAME}
-      - name: Download CC0 / public-domain test image
+      - name: Download MIT-granted test image (by project author, see env block)
         run: curl -L --fail -A "java-llama.cpp-ci/1.0" "${VISION_IMAGE_URL}" --create-dirs -o models/${VISION_IMAGE_NAME}
       - name: List files in models directory
         run: ls -l models/
@@ -484,7 +488,7 @@ jobs:
         run: curl -L --fail ${VISION_MODEL_URL} --create-dirs -o models/${VISION_MODEL_NAME}
       - name: Download vision mmproj
         run: curl -L --fail ${VISION_MMPROJ_URL} --create-dirs -o models/${VISION_MMPROJ_NAME}
-      - name: Download CC0 / public-domain test image
+      - name: Download MIT-granted test image (by project author, see env block)
         run: curl -L --fail -A "java-llama.cpp-ci/1.0" "${VISION_IMAGE_URL}" --create-dirs -o models/${VISION_IMAGE_NAME}
       - name: List files in models directory
         run: ls -l models/
@@ -547,7 +551,7 @@ jobs:
         run: curl -L --fail ${VISION_MODEL_URL} --create-dirs -o models/${VISION_MODEL_NAME}
       - name: Download vision mmproj
         run: curl -L --fail ${VISION_MMPROJ_URL} --create-dirs -o models/${VISION_MMPROJ_NAME}
-      - name: Download CC0 / public-domain test image
+      - name: Download MIT-granted test image (by project author, see env block)
         run: curl -L --fail -A "java-llama.cpp-ci/1.0" "${VISION_IMAGE_URL}" --create-dirs -o models/${VISION_IMAGE_NAME}
       - name: List files in models directory
         run: ls -l models/
@@ -613,7 +617,7 @@ jobs:
         run: curl -L --fail $env:VISION_MODEL_URL --create-dirs -o models/$env:VISION_MODEL_NAME
       - name: Download vision mmproj
         run: curl -L --fail $env:VISION_MMPROJ_URL --create-dirs -o models/$env:VISION_MMPROJ_NAME
-      - name: Download CC0 / public-domain test image
+      - name: Download MIT-granted test image (by project author, see env block)
         run: curl -L --fail -A "java-llama.cpp-ci/1.0" $env:VISION_IMAGE_URL --create-dirs -o models/$env:VISION_IMAGE_NAME
       - name: List files in models directory
         run: ls -l models/

docs/history/49be664_open_issues.md

@@ -939,8 +939,9 @@ the same pattern as the existing CodeLlama / Jina-Reranker model downloads.
    `InferenceParameters.setMessages(List<ChatMessage>)`) plus a
    `MultimodalIntegrationTest` gated on `net.ladenthin.llama.vision.model`,
    `.vision.mmproj`, and `.vision.image` system properties. CI in
-   `publish.yml` downloads a small vision model + mmproj + a CC0 / PD
-   image and passes the paths on the `mvn test` command line.
+   `publish.yml` downloads a small vision model + mmproj + an author-
+   licensed (MIT-granted) test image from Wikimedia Commons and passes
+   the paths on the `mvn test` command line.
 4. **Fourth PR (CI):** add an Android emulator job to formally close #121 and
    #50.
 

src/test/java/net/ladenthin/llama/MultimodalIntegrationTest.java

@@ -37,13 +37,21 @@
  * files explicitly). See {@code .github/workflows/publish.yml} env vars
  * {@code VISION_MODEL_URL} / {@code VISION_MMPROJ_URL} / {@code VISION_IMAGE_URL}.</p>
  *
+ * <p><b>Image source:</b> the CI default is
+ * <a href="https://commons.wikimedia.org/wiki/File:20200601_135745_Flowers_and_Bees.jpg">
+ * {@code File:20200601_135745_Flowers_and_Bees.jpg}</a> on Wikimedia Commons
+ * by Bernard Ladenthin (the project copyright holder); published there under
+ * CC-BY-4.0 and additionally granted under MIT to this project by the same
+ * author. Any image the test machine can reach works at runtime &#x2014; the
+ * URL is just an env var.</p>
+ *
  * <p>Closes issues #103 and #34.</p>
  */
 @ClaudeGenerated(
-        purpose = "End-to-end vision regression: real vision GGUF + mmproj + CC0 image fed through " +
-                  "the typed ChatMessage(role, List<ContentPart>) API; asserts non-empty reply to "
-                + "prove the OAI multipart content round-trips through the upstream mtmd pipeline. "
-                + "Closes #103 / #34."
+        purpose = "End-to-end vision regression: real vision GGUF + mmproj + author-licensed (MIT) "
+                + "test image fed through the typed ChatMessage(role, List<ContentPart>) API; "
+                + "asserts non-empty reply to prove the OAI multipart content round-trips through "
+                + "the upstream mtmd pipeline. Closes #103 / #34."
 )
 public class MultimodalIntegrationTest {