Kunj SHAHandNandhaKishorM d0aeae3125 fix: harden inference and runtime parity (#319) ...

* fix: harden inference and runtime parity

* fix: harden runtime lifecycle and fallbacks

* fix: make fast-path state safe for test runtimes

---------

Co-authored-by: NandhaKishorM <nandakishor@convaiinnovations.com>

2026-09-24 09:08:03 +05:30

..

test_audit_regressions.py

fix: harden inference and runtime parity (#319)

2026-09-24 09:08:03 +05:30

test_batch.py

feat(agent): report the calibrated confidence alongside the existing one (#126)

2026-09-24 08:47:36 +05:30

test_blank_lang_routing.py

fix(router): blank/whitespace explicit lang falls through to detection (#292)

2026-09-24 08:11:25 +05:30

test_calibration_persistence.py

fix(calibration): clear inherited buckets from per-type exports (#139)

2026-09-23 12:35:10 +05:30

test_cli.py

feat(cli): answer a question preset with --preset (#303)

2026-09-24 08:17:23 +05:30

test_confidence.py

feat(agent): report the calibrated confidence alongside the existing one (#126)

2026-09-24 08:47:36 +05:30

test_context_manager.py

feat(runtime): add context manager support to Agent and Router for deterministic resource cleanup (#150)

2026-09-23 12:36:05 +05:30

test_criteria_normalization.py

fix(agent): normalize boolean criteria keys and document noul criteria (#146)

2026-09-23 12:35:38 +05:30

test_criteria.py

perf(agent): reuse the tokenized state across questions; autocast on MPS (#109)

2026-09-24 08:28:45 +05:30

test_decision_model.py

release: bump version to 0.3.5 and add all test suites to CI/release workflows

2026-09-21 23:47:59 +05:30

test_doc_tables.py

fix(benchmarks): two parity cells did not match the JSON they cite (#299)

2026-09-24 08:13:36 +05:30

test_docker_entrypoint.py

build: add Docker and Compose quickstart with optional CUDA (#45)

2026-09-23 12:44:02 +05:30

test_docker_torch.py

build(docker): add ARM64 and DGX Spark container builds (#263)

2026-09-23 23:16:17 +05:30

test_download.py

fix(tests): normalize expected Hub paths on Windows (#140)

2026-09-23 12:33:46 +05:30

test_email.py

fix(email): closings the case rule does not reach (#245)

2026-09-23 19:47:02 +05:30

test_empty_questions.py

fix(agent): return empty predictions without inference (#144)

2026-09-23 12:34:13 +05:30

test_fast.py

Optional TileLang GPU fast path: 4-5x lower latency per call, identical answers (#25)

2026-09-23 19:01:15 +05:30

test_head_checkpointing.py

fix(training): honor decision-head activation checkpointing (#149)

2026-09-23 18:54:32 +05:30

test_hooks_api.py

feat(hooks): BaseHook and process-wide default hooks (#276)

2026-09-24 08:26:54 +05:30

test_hooks.py

feat(hooks): BaseHook and process-wide default hooks (#276)

2026-09-24 08:26:54 +05:30

test_lang_guess.py

fix(lang): identify plain-ASCII German instead of labelling it English (#130)

2026-09-23 23:17:58 +05:30

test_langchain.py

fix(integrations): include langgraph dependency in langchain and lang… (#257)

2026-09-23 23:15:11 +05:30

test_lazy_import.py

perf: keep torch out of import laya (lazy torch-backed exports) (#108)

2026-09-23 19:48:55 +05:30

test_load_errors.py

test: cover the load-time and budget errors no suite reached (#237)

2026-09-23 18:52:52 +05:30

test_local_e2e.py

feat(agent): add predict_batch for shared-forward-pass batch inference (#47)

2026-09-23 19:21:04 +05:30

test_mcp_local_e2e.py

feat: optional MCP stdio server (laya[mcp] extra)

2026-09-23 10:29:38 +00:00

test_mcp.py

feat: optional MCP stdio server (laya[mcp] extra)

2026-09-23 10:29:38 +00:00

test_onnx.py

feat: add torch.compile support and ONNX Runtime inference engine (#215)

2026-09-23 19:30:16 +05:30

test_packaging.py

build(docker): add ARM64 and DGX Spark container builds (#263)

2026-09-23 23:16:17 +05:30

test_portability.py

perf(agent): reuse the tokenized state across questions; autocast on MPS (#109)

2026-09-24 08:28:45 +05:30

test_predict_batch.py

feat(router): predict_batch — route-then-group multi-state scoring (follow-up to #47, with consistency harness + benchmarks) (#166)

2026-09-24 08:22:23 +05:30

test_router_batch.py

feat(router): predict_batch — route-then-group multi-state scoring (follow-up to #47, with consistency harness + benchmarks) (#166)

2026-09-24 08:22:23 +05:30

test_router_memory.py

perf(router): reclaim VRAM and trigger garbage collection on eviction and unload (#147)

2026-09-23 12:35:52 +05:30

test_router.py

fix(lang): identify plain-ASCII German instead of labelling it English (#130)

2026-09-23 23:17:58 +05:30

test_runtime_fixes.py

perf(agent): reuse the tokenized state across questions; autocast on MPS (#109)

2026-09-24 08:28:45 +05:30

test_serve.py

fix(serve): enforce MAX_BODY_BYTES when the client sends no Content-Length (#296)

2026-09-24 08:12:30 +05:30

test_server_example.py

Add local web GUI + JSON API demo server (#38) (#114)

2026-09-23 23:26:46 +05:30

test_shortlist_cosine.py

perf(shortlist): clip cosine similarities and guard empty document matrices in candidate ranking (#158)

2026-09-23 12:36:19 +05:30

test_shortlist.py

fix(shortlist): refresh embedding device on each call (#145)

2026-09-23 12:35:24 +05:30

test_structured_api.py

feat(structured): schema-driven decisions via decide() (#280)

2026-09-24 08:41:41 +05:30

test_structured.py

feat(structured): schema-driven decisions via decide() (#280)

2026-09-24 08:41:41 +05:30

test_system_one_lang.py

feat: add lang_temperatures and fix shadowing bug (#258)

2026-09-24 08:39:47 +05:30

test_temperature_loading.py

fix(calibration): handle invalid entries when reporting temperatures (#142)

2026-09-23 12:34:00 +05:30

test_tokenizer_cache.py

fix(agent): write tokenizer_config atomically via a temp file + os.replace (#267)

2026-09-23 23:14:01 +05:30

test_training.py

fix: reject targets longer than their own marker count in mixed batches (#311) (#312)

2026-09-24 08:46:04 +05:30

test_truncation_direction.py

perf(agent): reuse the tokenized state across questions; autocast on MPS (#109)

2026-09-24 08:28:45 +05:30