Andrew Park 08fe3a6c98 evals: switch the default LLM judge to claude-haiku-4-5 ...

Replaces gpt-5-mini-2025-08-07 as the default judge everywhere: the 147
eval configs plus the hardcoded defaults in the evals CLI, recipe composer,
comparison config generator, taubench user-simulator, trial runner and
OptimizeConfig.

Haiku is roughly 3x faster on the judging path and we are out of OpenAI
quota, so gpt-5-mini was failing closed on long eval runs.

2026-07-11 11:59:32 -07:00

..

agents

release: v1.0.1 (#357)

2026-05-18 20:19:37 -07:00

intelligence

feat: Algolia DocSearch + learning subsystem reorganization (#43)

2026-03-12 21:34:31 -07:00

routing

feat: query complexity analyzer for local/cloud routing (#102)

2026-03-24 15:27:23 -07:00

spec_search

spec_search: drop CLI stubs, rename, and ship paper-aligned tutorial + configs (#346)

2026-05-15 14:17:23 -07:00

training

fix(channels): wire channel→agent handler and fix Telegram send pipeline (#94)

2026-03-20 18:35:18 -07:00

__init__.py

init commit

2026-03-12 17:29:39 +00:00

test_agent_evolver.py

feat: Algolia DocSearch + learning subsystem reorganization (#43)

2026-03-12 21:34:31 -07:00

test_device_selection.py

feat: Algolia DocSearch + learning subsystem reorganization (#43)

2026-03-12 21:34:31 -07:00

test_feedback_collector.py

refactor: remove dead shims, stale files, and test duplication (#243)

2026-04-13 15:03:03 -07:00

test_heuristic_policy.py

feat: Algolia DocSearch + learning subsystem reorganization (#43)

2026-03-12 21:34:31 -07:00

test_heuristic_reward.py

fix(channels): wire channel→agent handler and fix Telegram send pipeline (#94)

2026-03-20 18:35:18 -07:00

test_learning_api.py

feat: Algolia DocSearch + learning subsystem reorganization (#43)

2026-03-12 21:34:31 -07:00

test_learning_orchestrator.py

init commit

2026-03-12 17:29:39 +00:00

test_learning_taxonomy.py

init commit

2026-03-12 17:29:39 +00:00

test_llm_optimizer.py

refactor: remove dead shims, stale files, and test duplication (#243)

2026-04-13 15:03:03 -07:00

test_multi_bench_runner.py

refactor: remove dead shims, stale files, and test duplication (#243)

2026-04-13 15:03:03 -07:00

test_optimize_cmd.py

evals: switch the default LLM judge to claude-haiku-4-5

2026-07-11 11:59:32 -07:00

test_optimize_store.py

refactor: remove dead shims, stale files, and test duplication (#243)

2026-04-13 15:03:03 -07:00

test_optimize_types.py

refactor: remove dead shims, stale files, and test duplication (#243)

2026-04-13 15:03:03 -07:00

test_optimizer_engine.py

refactor: remove dead shims, stale files, and test duplication (#243)

2026-04-13 15:03:03 -07:00

test_orchestrator_skill_optimization.py

feat: add skills system with import, learning loop, and benchmark harness (#230)

2026-04-09 14:43:04 -07:00

test_pareto.py

refactor: remove dead shims, stale files, and test duplication (#243)

2026-04-13 15:03:03 -07:00

test_personal_synthesizer.py

refactor: remove dead shims, stale files, and test duplication (#243)

2026-04-13 15:03:03 -07:00

test_router_stubs.py

fix(channels): wire channel→agent handler and fix Telegram send pipeline (#94)

2026-03-20 18:35:18 -07:00

test_router.py

feat: query complexity analyzer for local/cloud routing (#102)

2026-03-24 15:27:23 -07:00

test_routing_models.py

feat: query complexity analyzer for local/cloud routing (#102)

2026-03-24 15:27:23 -07:00

test_search_space.py

refactor: remove dead shims, stale files, and test duplication (#243)

2026-04-13 15:03:03 -07:00

test_skill_discovery.py

fix(channels): wire channel→agent handler and fix Telegram send pipeline (#94)

2026-03-20 18:35:18 -07:00

test_skill_optimizer.py

feat: add skills system with import, learning loop, and benchmark harness (#230)

2026-04-09 14:43:04 -07:00

test_stubs.py

init commit

2026-03-12 17:29:39 +00:00

test_system_learning.py

feat: Algolia DocSearch + learning subsystem reorganization (#43)

2026-03-12 21:34:31 -07:00

test_trace_judge.py

refactor: remove dead shims, stale files, and test duplication (#243)

2026-04-13 15:03:03 -07:00

test_trial_runner.py

evals: switch the default LLM judge to claude-haiku-4-5

2026-07-11 11:59:32 -07:00