v0.76.0-rc8

What's Changed

• Fix crates.io publishing for new workspace crates, and make dev builds identify themselves by @michaelneale in #1223
• Split Skippy by functional boundary by @i386 in #1194
• feat: pass reasoning effort through Skippy templates by @i386 in #1211
• chore: update pinned llama.cpp revision by @i386 in #1216
• fix: IPC idle timeout for external-relay plugins by @MahdiHedhli in #1191
• Publish Rust crate API docs with the website by @i386 in #1230
• feat(mesh): expose model-bound tokenizer capability by @i386 in #1227
• fix: repair the Fly console image build by @michaelneale in #1241
• chore(llama): regenerate patch queue for latest upstream by @i386 in #1232
• feat(catalog): add Muse Glimmer 30B to the model catalog by @michaelneale in #1240
• feat(bench): run SWE-Gym through Harbor by @i386 in #1239
• fix: Skippy Nemotron MTP loading by @ndizazzo in #1245
• fix: complete remaining Skippy MTP runtime callers by @i386 in #1247
• chore(deps): pin iroh to 1.0.3 explicitly by @michaelneale in #1257
• feat(logging): add durable lifecycle logging foundation by @ndizazzo in #1174
• feat(logging): expose audited local log APIs and runtime telemetry by @ndizazzo in #1175
• feat(console): add typed logs ledger and maintenance UI by @ndizazzo in #1176
• feat(openai): integrate exact lifecycle and trusted sessions by @ndizazzo in #1258
• feat: make plugin-only Serve mode nodes discoverable/reachable by other mesh peers by @MahdiHedhli in #1192
• Fix install one-liner 404 on NVIDIA hosts without a CUDA toolkit by @michaelneale in #1267
• ci: reshape PR and main CI around composable slices, prepping for Depot workers by @ndizazzo in #1244
• ci: distinguish skipped bootstrap routes by @ndizazzo in #1272
• ci: fail fast on topic lane projections by @ndizazzo in #1274
• ci: flatten paginated lane checks by @ndizazzo in #1276
• Retry transient model route misses in console chat by @i386 in #1277
• fix(ci): make CUDA release checks hermetic by @ndizazzo in #1270
• fix: CI smoke matrix routing by @ndizazzo in #1284
• fix(moa): restore single-model degrade for model=mesh on host ingress by @michaelneale in #1291
• ci: split planning and platform workflow graphs by @ndizazzo in #1282
• fix(ci): keep main planning exhaustive by @ndizazzo in #1295
• fix(cache): certify Muse-Glimmer family for the resident-KV prefix cache by @michaelneale in #1294
• Fix split serving: stages above layer 0 fail to load by @michaelneale in #1290
• fix(logging): harden lifecycle capture and console recovery by @ndizazzo in #1268
• chore: update mesh-llm-ui console dependencies by @ndizazzo in #1286
• fix: move model lifecycle controls off peer-reachable ingress by @ndizazzo in #1279
• fix: name a local GGUF with --model instead of trying to resolve it by @michaelneale in #1269
• fix: cross-platform product bundle tree hashes by @michaelneale in #1271
• Keep degraded MoA requests on the resolved pipeline model by @michaelneale in #1292
• ci: prepare protected Depot PR runners by @ndizazzo in #1302
• ci: add bounded Depot PR canary gate by @ndizazzo in #1306
• ci: isolate native caches on Depot PR runners by @ndizazzo in #1310
• ci: keep Depot cache namespace inert by @ndizazzo in #1317
• CI: pin Depot audit to merged policy by @ndizazzo in #1319
• Durable KV prefix cache: agent prefixes survive eviction, restart, and cold nodes by @michaelneale in #1228
• feat(topology): route Qwen3.8 identities to the qwen35 recurrent family by @michaelneale in #1283
• Make the optional GitHub star step reliable in setup by @michaelneale in #1278
• Fix/windows skippy package large gguf by @wangwenjunfromlanzhou in #1307
• CI: identify failing Depot canary endpoint by @ndizazzo in #1321
• CI: classify rejected Depot cache endpoint by @ndizazzo in #1323
• CI: add protected Depot authority sentinel by @ndizazzo in #1324
• CI: prove Depot sentinel cache writes by @ndizazzo in #1326
• Docs: record unsafe Depot PR authority by @ndizazzo in #1328
• Docs: define Depot PR cache isolation contract by @ndizazzo in #1329
• Allow bounded Depot caching for approved PRs by @ndizazzo in #1333
• Guard approved Depot PR routing by @ndizazzo in #1334
• ci: repin Depot PR isolation audit by @ndizazzo in #1336
• ci: fingerprint Depot macOS toolchains by @ndizazzo in #1337
• chore: sync source versions during GitHub releases by @ndizazzo in #1330
• docs: record Depot PR rollout evidence by @ndizazzo in #1341
• fix(skippy): restore recurrent shared prefixes by @michaelneale in #1342
• feat(catalog): recommend Qwen3.8 27B Q4_K_M by @michaelneale in #1343
• fix(moa): preserve caller tool schemas by @michaelneale in #1344
• fix(split): let Iroh own transport path selection by @michaelneale in #1345
• Consolidate model=auto and model=mesh into one automatic-routing directive by @michaelneale in #1309
• Return a visible error when streaming prompts exceed context by @michaelneale in #1352
• Treat max_tokens as a ceiling, not a context reservation by @michaelneale in #1354
• Stop aborting healthy long prefills at five minutes by @michaelneale in #1355
• Return IDs for OpenAI tool calls on local-model-only serving by @michaelneale in #1318
• Cap remotely advertised model lists to protect the console from a hostile peer by @michaelneale in #1357
• Reject non-protobuf content types on OTLP/HTTP ingestion by @michaelneale in #1356
• Add mode-aware management health endpoint by @i386 in #1351
• fix(skippy): serialize native proposal lifecycle by @i386 in #1242
• CI recognizes Apple provider paths by @i386 in #1363
• fix(ui): remove dead ungated ErrorBoundary that leaks stack traces (#39) by @michaelneale in #1361
• Reuse recurrent KV across growing chat turns by @i386 in #1253
• Fix blank responses from thinking models by @michaelneale in #1365
• perf(skippy): coalesce native KV page transfers by @michaelneale in #1368
• task: refine logging console UX and live delivery by @ndizazzo in #1339
• Stream tool calls as they are generated instead of after the turn ends by @michaelneale in #1371
• fix(skippy): stop a stalled SSE consumer pinning a generation worker by @ndizazzo in #1367
• fix(mtp): make native MTP actually work on single-node serving by @michaelneale in #1366
• fix: keep --log-format json parseable and block new raw console prints in CI by @ndizazzo in #1376
• fix(xtask): regenerate console-print ratchet drifted 2 lines by #1376 by @ndizazzo in #1379
• Wire Playwright into CI and fix the 9 pre-existing /logs failures by @ndizazzo in #1377
• fix(ci): select static-abi in pr-draft when its matrix consumers are planned by @ndizazzo in #1383
• test(ui): tolerate hosted-runner frame cadence by @ndizazzo in #1386
• perf(kv): keep exact-state recording off inference path by @michaelneale in #1389
• ci: cancel sibling PR lanes after failure by @ndizazzo in #1388
• fix(tui): make the dashboard immune to stray console output by @ndizazzo in #1382
• ci: bound compiler and local build caches by @ndizazzo in #1390
• Revert "ci: bound compiler and local build caches (#1390)" by @ndizazzo in #1394
• ci: prebuilt runner images for smoke/SDK/web jobs by @ndizazzo in #1380
• ci: bound compiler and local build caches by @ndizazzo in #1395
• fix(moa): prevent reasoning-only model poisoning by @michaelneale in #1391
• docs(skippy): remove stale disk cache defaults by @michaelneale in #1398
• Stop persisting KV cache state to disk by @michaelneale in #1399
• Fix release builds failing in the containerized smoke job by @michaelneale in #1401
• fix(ci): point pnpm at the runner image's baked store instead of the Actions cache by @ndizazzo in #1400
• refactor: complete evidence-backed codebase audit waves by @ndizazzo in #1396
• ci(quality): install just before the CI-contracts python tests run by @ndizazzo in #1407
• fix: close the four rc6 release-validation defects by @ndizazzo in #1405
• Advance llama.cpp and simplify the Skippy patch queue by @i386 in #1402
• skippy: address #1402 review follow-up (dead comparison + Inkling test coverage) by @i386 in #1412
• fix(api): stop leaking plugin endpoint credentials over the management API by @michaelneale in #1364
• fix(build): link libvendor-hash.a and test Rust on llama.cpp pin bumps by @michaelneale in #1419
• Rescue terminal Buzz replies in MoA by @michaelneale in #1417
• Compact oversized chat context for provider windows by @i386 in #1280
• Unblock release dispatches after the config schema module split by @michaelneale in #1426
• fix(packaging): retry transient HF artifact uploads by @michaelneale in #1424
• Fix chat-chain pipelined splits: verify-history accounting, early-stop replies, session leaks by @danielwinterw in #1408
• ci: skip build slices for draft pull requests by @ndizazzo in #1415
• fix(mesh): close the iroh endpoint on shutdown instead of dropping it by @ndizazzo in #1431
• fix: redact plugin provider health details by @ndizazzo in #1432
• fix(release): compare release bases, not raw tag SHAs, in provenance check by @ndizazzo in #1430
• fix: drain hosted split stages on shutdown by @ndizazzo in #1433
• fix: bound downstream endpoint resolution by @ndizazzo in #1438
• feat: iteration-level scheduler for concurrent staged serving (#1416) by @i386 in #1420
• task: make operational logs durable, attributable, and queryable by @ndizazzo in #1440
• task: add lifecycle timelines and recovery to the logs console by @ndizazzo in #1441
• feat(skippy): replace flat prefix caches with unified radix by @i386 in #1429
• perf(skippy): add scheduler lab and cache-aware radix scheduling by @i386 in #1447
• test(skippy): pin scheduler workload fixtures by @i386 in #1450
• feat(skippy): make resident KV admission capacity aware by @i386 in #1452
• ci(llama): tiered family battery + agent-repaired canary on the family-certify runner by @i386 in #1436
• Run the llama.cpp canary daily and accept split GGUF metadata by @i386 in #1457
• fix(skippy): restore small-context admission and state handoff by @i386 in #1459
• fix(proxy): cancel upstream before response headers by @ndizazzo in #1448
• ci: assign imported Just files to Rust ownership by @ndizazzo in #1464
• refactor(just): split root Justfile into semantic imports by @ndizazzo in #1458
• fix(ci): keep change planning below expression limit by @ndizazzo in #1468
• Send remote mesh inference directly into the serving pipeline by @michaelneale in #1463
• Clarify end-to-end encryption for mesh traffic by @michaelneale in #1470
• Fix llama canary family certification failures by @i386 in #1471
• skippy-quantize: compose-mtp — splice an MTP draft into a sharded target GGUF by @i386 in #1439
• Route repeated prompts to workers with verified radix cache reuse by @i386 in #1449
• fix(skippy): use raw f32 activation transport by @i386 in #1482
• ci(llama): wire the agent repair loop into the upstream canary by @i386 in #1481
• ci(llama): heartbeat, arm64 build guard, and push-permission diagnostics for the canary repair loop by @i386 in #1489
• ci(llama): preflight repair-token push permission before any repair work by @i386 in #1493
• ci(llama): probe actual token write capability in the repair preflight by @i386 in #1496
• ci(llama): prune stale /tmp worktrees and fix the preflight probe delete by @i386 in #1499
• ci(llama): auto-approve agent tool permissions in the repair sandbox by @i386 in #1502
• fix(llama): rebase patch queue onto upstream b19cbe925b by @i386 in #1504
• ci(llama): ensure the repair PR exists before applying the agent's body by @i386 in #1506
• fix(llama): rebase patch queue onto upstream b19cbe925b by @i386 in #1505
• fix(ci): validate CUDA runtime toolkit labels by @ndizazzo in #1465
• fix(ui): render LaTeX in active chat responses by @ndizazzo in #1466
• ci(llama): review certified canary repairs with a fresh agent turn by @i386 in #1507
• skippy: support staged Qwen3.8 Flash-Next inference by @michaelneale in #1509
• Certify Qwen3.8 Flash-Next in the llama family battery by @i386 in #1513
• fix(skippy): address Qwen4 split framing review findings by @i386 in #1516
• fix(release): use dotted workspace version form in skippy-scheduler by @michaelneale in #1518
• ci: regenerate console-print ratchet after #1516 by @i386 in #1525

New Contributors

• @MahdiHedhli made their first contribution in #1191
• @wangwenjunfromlanzhou made their first contribution in #1307

Full Changelog: v0.75.1...v0.76.0-rc8