v5.0.5

What's Changed

• Post-release: bump to 5.0.5-SNAPSHOT by @bernardladenthin in #292
• NativeServer mode, GPU/CPU classifier matrix (ROCm/SYCL/OpenVINO/Win-arm64/s390x), llama.cpp b9870 + bump automation by @bernardladenthin in #293
• Upgrade checker and spotless dependencies by @bernardladenthin in #294
• Release version 5.0.5 by @bernardladenthin in #295

Full Changelog: v5.0.4...v5.0.5

v5.0.4

What's Changed

• Post-release: bump to 5.0.4-SNAPSHOT by @bernardladenthin in #282
• Upgrade palantir-java-format from 2.92.0 to 2.94.0 by @bernardladenthin in #283
• Add langchain4j-jllama module: in-process LangChain4j adapters by @vaiju1981 in #284
• ModelMeta: expose chat template, special tokens, and full GGUF metadata by @vaiju1981 in #286
• feat: llama-langchain4j — rename, CI build/test + Central publish, model-backed tests, upfront model cache by @bernardladenthin in #285
• Configurable fine-tuning API (TrainingParameters + Optimizer) by @vaiju1981 in #289
• refactor: move core into llama/ submodule under a Maven reactor (parent + llama + llama-langchain4j) by @bernardladenthin in #288
• Upgrade llama.cpp from b9842 to b9859 by @bernardladenthin in #290
• Release 5.0.4: LangChain4j integration, fine-tuning API, and Maven reactor by @bernardladenthin in #291

Full Changelog: v5.0.3...v5.0.4

v5.0.3

What's Changed

• Claude/elegant wright w6f vx by @bernardladenthin in #218
• docs: move jqwik version warning to troubleshooting section by @bernardladenthin in #219
• Upgrade spotbugs and pitest-maven to latest patch versions by @bernardladenthin in #220
• chore(deps): bump com.google.errorprone:error_prone_core from 2.49.0 to 2.50.0 by @dependabot[bot] in #223
• chore(deps): bump com.github.spotbugs:spotbugs-maven-plugin from 4.9.8.4 to 4.10.2.0 by @dependabot[bot] in #222
• chore(deps-dev): bump reactor.version from 3.8.5 to 3.8.6 by @dependabot[bot] in #221
• Upgrade palantir-java-format from 2.91.0 to 2.92.0 by @bernardladenthin in #224
• Upgrade llama.cpp from b9555 to b9621 by @bernardladenthin in #225
• Add SonarQube code quality analysis workflow by @bernardladenthin in #226
• Upgrade Android minimum API level from 24 to 28 by @bernardladenthin in #227
• Skip reranking tests when model file is unavailable by @bernardladenthin in #228
• Fix Android API gating: use weak symbols instead of ANDROID_API by @bernardladenthin in #229
• Replace synchronized method with private lock object in LlamaLoader by @bernardladenthin in #230
• chore(deps): bump actions/cache from 4.3.0 to 5.0.5 by @dependabot[bot] in #232
• chore(deps): bump actions/checkout from 4.3.1 to 6.0.3 by @dependabot[bot] in #231
• chore(deps): bump actions/setup-java from 4.8.0 to 5.2.0 by @dependabot[bot] in #233
• Upgrade llama.cpp to b9637 and update GitHub Actions versions by @bernardladenthin in #234
• Add log_helpers.hpp with pure log-formatting functions by @bernardladenthin in #235
• Extract SessionState: testable concurrency contract for Session by @bernardladenthin in #236
• Upgrade llama.cpp from b9637 to b9642 by @bernardladenthin in #237
• Add Spotless formatting policy documentation by @bernardladenthin in #238
• Upgrade llama.cpp to b9682 and improve CI test diagnostics by @bernardladenthin in #239
• Add manual publish control to CI workflow by @bernardladenthin in #241
• Add OpenAI-compatible HTTP server (LlamaServer) by @bernardladenthin in #242
• Add OpenAI-compatible HTTP endpoint for local model serving by @bernardladenthin in #240
• feat: parallel_tool_calls + tool-calling tests by @vaiju1981 in #244
• Consolidate OpenAI server: unify implementations, add multi-protocol support by @bernardladenthin in #243
• CI build-cache rollout (sccache Phase 2 + GGUF models) + CUDA fast-build, dep bumps, NativeServer scaffold by @bernardladenthin in #245
• Upgrade actions/checkout from v6 to v7 by @bernardladenthin in #246
• Upgrade llama.cpp from b9682 to b9739 by @bernardladenthin in #247
• Add Windows Ninja Multi-Config build with sccache caching by @bernardladenthin in #248
• feat: wire vision/multimodal input end to end by @vaiju1981 in #249
• Add Windows model caching and fix issue reference formatting by @bernardladenthin in #250
• Fix fork PR sccache failures with token-gated install and build retry by @bernardladenthin in #252
• feat: pin sessions to their slot and expose KV-cache usage + metrics by @vaiju1981 in #251
• Validate slotId and handle next_token array in SlotMetrics by @bernardladenthin in #253
• Enable sccache wrapping of nvcc for full-arch CUDA builds by @bernardladenthin in #254
• CUDA CI: drop sccache debug diagnostics now that nvcc caching is proven by @bernardladenthin in #255
• docs: add JNI safety and server hardening issues to TODO by @bernardladenthin in #256
• Fix flaky reasoning budget tests on Metal by using greedy sampling by @bernardladenthin in #257
• build(deps): bump actions/setup-python from 5 to 6 by @dependabot[bot] in #259
• Feature/todo fixes by @vaiju1981 in #258
• Add tests for ChatTranscript.removePendingUserTurn and JNI exception handling by @bernardladenthin in #261
• fix(loader): make native-lib extraction race-safe (atomic write + con… by @vaiju1981 in #260
• test(loader): cover LlamaLoader.resourceMatchesFile match + content-mismatch branches by @bernardladenthin in #262
• chore: make SpotBugs (effort=Max, threshold=Low) clean again by @vaiju1981 in #263
• fix(loader): finish SpotBugs cleanup (LlamaLoader AFBR + PRMC) + docs(todo): mark audit findings resolved by @bernardladenthin in #264
• docs: Add AI-assisted development note to README by @bernardladenthin in #265
• feat(server): stream POST /v1/completions token-by-token (no new nati… by @vaiju1981 in #266
• feat(multimodal): audio input via OpenAI input_audio content parts by @vaiju1981 in #267
• docs: document PIT mutation testing policy and audioFile hermiticity issue by @bernardladenthin in #269
• feat(tts): text-to-speech via the OuteTTS + WavTokenizer pipeline by @vaiju1981 in #268
• Harden TTS WAV validation and upstream enum pinning by @bernardladenthin in #270
• Upgrade llama.cpp from b9739 to b9789 and refresh Windows patch by @bernardladenthin in #271
• Replace --skip-download with --offline flag (llama.cpp b9803) by @bernardladenthin in #272
• Add SpotBugs check to CI and expand PATH_TRAVERSAL_IN suppression by @bernardladenthin in #273
• Upgrade llama.cpp to b9829; add DRY sampling parameters by @bernardladenthin in #274
• Add ModelParameters.enableSwaFull() for SWA KV cache control by @bernardladenthin in #275
• build(deps): bump ch.qos.logback:logback-classic from 1.5.34 to 1.5.37 by @dependabot[bot] in #278
• build(deps-dev): bump org.junit.jupiter:junit-jupiter from 6.1.0 to 6.1.1 by @dependabot[bot] in #277
• build(deps): bump actions/cache from 5 to 6 by @dependabot[bot] in #279
• Windows: flip default to Ninja, add MSVC classifier, add GPU classifiers by @bernardladenthin in #276
• Upgrade llama.cpp from b9829 to b9840 by @bernardladenthin in #280
• Release 5.0.3: OpenAI-compatible server, multimodal, and llama.cpp b9842 by @bernardladenthin in #281

New Contributors

• @vaiju1981 made their first contribution in #244

Full Changelog: v5.0.2...v5.0.3

v5.0.2

What's Changed

• Add integration tests for reasoning budget token enforcement by @bernardladenthin in #140
• Add tracking link for llama.cpp reasoning_budget_tokens bug by @bernardladenthin in #141
• Add attribution to original java-llama.cpp fork by @bernardladenthin in #142
• Upgrade llama.cpp from b9151 to b9172 by @bernardladenthin in #143
• Add start gate abort window to publish workflow by @bernardladenthin in #144
• Reorganize README badges and add coverage/community badges by @bernardladenthin in #146
• docs: OpenSSF Best Practices passing-level artifacts by @bernardladenthin in #145
• docs: OpenSSF Best Practices passing-level artifacts by @bernardladenthin in #147
• docs: consolidate governance, add security workflows, and update baselines by @bernardladenthin in #148
• chore(deps): bump actions/upload-artifact from 4 to 7 by @dependabot[bot] in #150
• chore(deps): bump github/codeql-action from 3 to 4 by @dependabot[bot] in #155
• chore(deps): bump actions/checkout from 4 to 6 by @dependabot[bot] in #153
• chore(deps-dev): bump io.github.git-commit-id:git-commit-id-maven-plugin from 9.0.2 to 10.0.0 by @dependabot[bot] in #154
• chore(deps): bump google/osv-scanner-action/.github/workflows/osv-scanner-reusable.yml from 2.0.2 to 2.3.8 by @dependabot[bot] in #152
• chore(deps): bump google/osv-scanner-action/.github/workflows/osv-scanner-reusable-pr.yml from 2.0.2 to 2.3.8 by @dependabot[bot] in #151
• feat(reuse): add REUSE 3.3 license compliance by @bernardladenthin in #149
• Add .gitattributes for consistent line endings by @bernardladenthin in #158
• chore(deps): bump fsfe/reuse-action from 5 to 6 by @dependabot[bot] in #156
• chore(deps): bump ossf/scorecard-action from 2.4.0 to 2.4.3 by @dependabot[bot] in #157
• Downgrade Java target to 1.8 and update CI Java version by @bernardladenthin in #159
• Refactor GitHub release workflow to attach signed artifacts by @bernardladenthin in #160
• Upgrade llama.cpp from b9172 to b9198 by @bernardladenthin in #161
• Upgrade Java version to 21 and switch to Temurin distribution by @bernardladenthin in #162
• Downgrade jacoco-maven-plugin from 0.8.15 to 0.8.14 by @bernardladenthin in #163
• Increase reasoning token budget to handle slow CI runners by @bernardladenthin in #164
• Always build and publish CUDA artifacts; remove manual workflow inputs by @bernardladenthin in #165
• Update copyright attribution for Konstantin Herud by @bernardladenthin in #166
• Add ContinuationMode enum for channel-specific message continuation by @bernardladenthin in #167
• Upgrade llama.cpp from b9219 to b9222 by @bernardladenthin in #169
• Add NativeLibraryPermissionSetter and improve exception handling by @bernardladenthin in #170
• Add license scan report and status by @fossabot in #168
• Vendor OSInfo from xerial/sqlite-jdbc with Apache-2.0 license by @bernardladenthin in #171
• Fix InterruptedException handling in OSInfo utility methods by @bernardladenthin in #172
• Upgrade llama.cpp from b9222 to b9245 by @bernardladenthin in #173
• Upgrade llama.cpp from b9245 to b9264 by @bernardladenthin in #174
• Upgrade llama.cpp from b9264 to b9279 by @bernardladenthin in #175
• Refactor JNI lifecycle and log formatting into reusable helpers by @bernardladenthin in #176
• Upgrade llama.cpp from b9279 to b9284 by @bernardladenthin in #177
• Refactor parameter builders with putScalar and putEnum helpers by @bernardladenthin in #178
• docs: add baseline open issues snapshot at commit 49be664 by @bernardladenthin in #179
• Refactor CliArg enum tests into shared base class by @bernardladenthin in #180
• docs: Add similar projects section to README by @bernardladenthin in #182
• Add OpenCL/Adreno backend support for Android aarch64 by @bernardladenthin in #181
• docs: add feature investigation from llama-stack-client-kotlin by @bernardladenthin in #183
• docs: add deep-dive analysis and verification plan for open issues by @bernardladenthin in #184
• Add regression tests for issues #80, #95, #98, #102 by @bernardladenthin in #185
• Document classifier selection and mark 4 issues FIXED via PR #185 by @bernardladenthin in #186
• docs: Add Feature Ideas section to README by @bernardladenthin in #187
• Add async APIs, cancellation support, and typed metrics accessors by @bernardladenthin in #188
• Add per-session thread-safety and streaming guard to Session by @bernardladenthin in #189
• docs: add suggested upstream issue comments for resolved issues by @bernardladenthin in #190
• Upgrade llama.cpp from b9284 to b9297 by @bernardladenthin in #191
• Upgrade llama.cpp from b9297 to b9305 by @bernardladenthin in #193
• docs: Add llama.cpp-android-tutorial to related projects by @bernardladenthin in #194
• Upgrade llama.cpp from b9305 to b9333 by @bernardladenthin in #196
• chore(deps-dev): bump org.apache.maven.plugins:maven-surefire-plugin from 3.5.4 to 3.5.5 by @dependabot[bot] in #192
• Migrate test suite from JUnit 4 to JUnit 5 by @bernardladenthin in #195
• Add jqwik, ArchUnit, and SpotBugs by @bernardladenthin in #197
• Upgrade llama.cpp from b9333 to b9354 by @bernardladenthin in #198
• Add concurrency testing with JCStress and Lincheck by @bernardladenthin in #199
• build(pom): collect dep versions in properties, bump dep/plugin versions, make javadoc strict by @bernardladenthin in #200
• Downgrade jqwik to 1.9.3 and document QuickTheories replacement plan by @bernardladenthin in #201
• Fix JNI bridge to skip null sentinel results from llama.cpp b9437 by @bernardladenthin in #202
• Upgrade llama.cpp from b9437 to b9442 by @bernardladenthin in #203
• Upgrade to JSpecify 1.0.0, add nullness checking, JPMS module descriptor by @bernardladenthin in #209
• chore(deps): bump com.fasterxml.jackson.core:jackson-databind from 2.21.3 to 2.22.0 by @dependabot[bot] in #204
• chore(deps): bump org.apache.maven.plugins:maven-surefire-plugin from 3.5.5 to 3.5.6 by @dependabot[bot] in #206
• chore(deps): bump com.diffplug.spotless:spotless-maven-plugin from 3.5.1 to 3.6.0 by @dependabot[bot] in #208
• chore(deps): bump ch.qos.logback:logback-classic from 1.5.32 to 1.5.33 by @dependabot[bot] in #207
• Migra...

v5.0.1

What's Changed

• Bump version to 5.0.1-SNAPSHOT by @bernardladenthin in #127
• Refactor GitHub release workflow to parallelize snapshot and release jobs by @bernardladenthin in #128
• Add comprehensive JavaDoc documentation to public APIs by @bernardladenthin in #129
• Add Maven Central badge to README by @bernardladenthin in #130
• Remove snapshot build documentation and badge by @bernardladenthin in #131
• Upgrade Windows CI to windows-2025 with Visual Studio 2026 by @bernardladenthin in #132
• Use static MSVC runtime (/MT) for Windows builds by @bernardladenthin in #133
• Upgrade llama.cpp from b9106 to b9134 by @bernardladenthin in #134
• Release version 5.0.1 by @bernardladenthin in #135
• Upgrade llama.cpp from b9145 to b9150 by @bernardladenthin in #136
• Refactor CI workflow: add explicit snapshot/tag checks by @bernardladenthin in #137
• fix(publish): quote gate job names to avoid YAML colon-in-scalar error by @bernardladenthin in #138
• Upgrade llama.cpp from b9150 to b9151 by @bernardladenthin in #139

Full Changelog: v5.0.0...v5.0.1

v5.0.0

What's Changed

• Add CLAUDE.md with build commands and architecture overview by @bernardladenthin in #1
• Add comprehensive tests for Pair class by @bernardladenthin in #2
• fix: update deprecated GitHub Actions runners by @bernardladenthin in #3
• Add macOS 15 CI target and fix ARM i8mm feature conflict by @bernardladenthin in #4
• Claude/add getmetadata method x va jm by @bernardladenthin in #5
• fix: add -DGGML_NATIVE=OFF to macos-15-no-metal target and fix artifact name collision by @bernardladenthin in #6
• Update GitHub Actions to Node.js 24-compatible versions by @bernardladenthin in #7
• Configure Maven compiler to target Java 1.8 by @bernardladenthin in #8
• Update GitHub Actions to latest versions by @bernardladenthin in #9
• Update JNA dependency version from 5.13.0 to 5.18.1 by @bernardladenthin in #10
• Fix README badge href for llama.cpp version by @bernardladenthin in #11
• Upgrade llama.cpp from b4916 to b4927 by @bernardladenthin in #12
• Fix GPU hang crash on macOS 14 Metal CI runner by @bernardladenthin in #14
• Upgrade llama.cpp from b6545 to b6714 by @bernardladenthin in #15
• Upgrade llama.cpp from b6714 to b6715 by @bernardladenthin in #16
• Upgrade llama.cpp from b7217 to b7218 by @bernardladenthin in #17
• Upgrade llama.cpp from b7783 to b7788 by @bernardladenthin in #19
• Upgrade llama.cpp from b7858 to b7862 by @bernardladenthin in #20
• Upgrade llama.cpp from b7864 to b7867 by @bernardladenthin in #21
• Upgrade llama.cpp from b7867 to b7885 by @bernardladenthin in #22
• Upgrade llama.cpp from b7904 to b7920 by @bernardladenthin in #23
• Upgrade llama.cpp from b8190 to b8210 by @bernardladenthin in #24
• Upgrade llama.cpp from b8268 to b8495 with compatibility fixes by @bernardladenthin in #25
• Add C++ unit tests for b8576 upgrade (server_tokens, grammar, rerank,… by @bernardladenthin in #26
• Add unit tests for untested Java classes by @bernardladenthin in #18
• Document API compatibility files and known breaking changes for llama.cpp upgrades by @bernardladenthin in #27
• Upgrade llama.cpp from b8576 to b8579 by @bernardladenthin in #28
• Add vocab-only mode for lightweight tokenization without model weights by @bernardladenthin in #29
• Extract oaicompat_finish_reason() helper and add unit tests by @bernardladenthin in #30
• Claude/fix getifaddrs error by @bernardladenthin in #32
• Fix std::filesystem link error on GCC < 9 cross-compilation toolchains by @bernardladenthin in #33
• Fix Android build: disable LLAMA_BUILD_TOOLS on Android by @bernardladenthin in #34
• fix(android): use CACHE BOOL FORCE to disable server tools on Android by @bernardladenthin in #35
• fix(build): add mtmd include path to jllama target by @bernardladenthin in #36
• fix(build): explicitly build mtmd when consumed via FetchContent by @bernardladenthin in #37
• fix(build): set LLAMA_INSTALL_VERSION fallback for mtmd standalone build by @bernardladenthin in #39
• Claude/fix llama model loading 8 mk bo by @bernardladenthin in #40
• Claude/fix llama model loading 8 mk bo by @bernardladenthin in #41
• Add -q flag to Maven commands to suppress download progress output by @bernardladenthin in #42
• Add model validation before tests; ensure downloads are valid GGUF files by @bernardladenthin in #43
• Enable macOS and Windows test jobs; use Java 8 in all test jobs by @bernardladenthin in #44
• Fix Java 8 JNI compilation error and suppress Maven download progress by @bernardladenthin in #45
• Add reranking model download to macOS and Windows test jobs by @bernardladenthin in #46
• Opt into Node.js 24 for GitHub Actions to fix deprecation warnings by @bernardladenthin in #47
• Skip warmup in RerankingModelTest to fix native crash on Windows/macOS by @bernardladenthin in #48
• Enable CUDA build job and artifact in release pipeline by @bernardladenthin in #49
• Upgrade CUDA from 12.1 to 13.2 by @bernardladenthin in #50
• Add package job for downloadable JARs; update CUDA classifier to 13; … by @bernardladenthin in #51
• Add memory reporting before and after tests on all platforms by @bernardladenthin in #52
• Fix YAML structure: restore test-macos as a separate job by @bernardladenthin in #53
• Disable doclint in maven-javadoc-plugin to fix package job failure by @bernardladenthin in #54
• Merge ci.yml into release.yaml and remove ci.yml by @bernardladenthin in #56
• Add MemoryManagementTest: context shifting and prompt-cache prefix reuse by @bernardladenthin in #55
• Add @param/@return Javadoc tags and remove doclint suppression by @bernardladenthin in #57
• Upgrade llama.cpp from b8579 to b8609 by @bernardladenthin in #58
• Claude/update pooling type bf nxj by @bernardladenthin in #59
• Upgrade llama.cpp from b8609 to b8611 by @bernardladenthin in #60
• Add chatComplete() for OpenAI-compatible chat completions by @bernardladenthin in #61
• Add complex chat scenario tests (ChatScenarioTest) by @bernardladenthin in #62
• Add comprehensive test coverage for untested API surfaces by @bernardladenthin in #63
• Upgrade llama.cpp from b8611 to b8645 by @bernardladenthin in #64
• Fix: Package OpenSSL/BoringSSL DLLs in Windows x64 builds by @bernardladenthin in #65
• Fix: Copy OpenSSL DLLs at build time, not configure time by @bernardladenthin in #67
• Capture core dumps and surefire dumpstream files as CI artifacts on test failure by @bernardladenthin in #66
• Support OpenSSL/BoringSSL libraries across all platforms by @bernardladenthin in #68
• Enable BUILD_SHARED_LIBS for BoringSSL DLL generation by @bernardladenthin in #69
• Cleanup: Remove unnecessary DLL discovery mechanism by @bernardladenthin in #70
• Upgrade llama.cpp b8648 → b8664; expose KV cache idle-eviction API by @bernardladenthin in #71
• Improve testTokenizationUnicode to cover 2-byte and 3-byte UTF-8 sequences; add reserve() to str_to_bytes by @bernardladenthin in #73
• Upgrade llama.cpp from b8664 to b8665 by @bernardladenthin in #74
• refactor: extract get_server_context helper to eliminate JNI boilerplate by @bernardladenthin in #75
• Claude/ident...

Snapshot Build

Snapshot from 53f307a — https://github.com/bernardladenthin/java-llama.cpp/actions/runs/25552571414