v26.06.14

What's Changed

• feat(evaluation): add evaluation subpackage skeleton and pyproject entry point by @miguelgfierro in #268
• feat(evaluation): add matcher primitives and statistics helpers by @miguelgfierro in #269
• feat(evaluation): add corpus loader and registry modules by @miguelgfierro in #270
• feat(evaluation): add G1-G5 gate framework by @miguelgfierro in #271
• feat(evaluation): add scorecard renderer by @miguelgfierro in #272
• feat(evaluation): add LLM-as-judge and judge client by @miguelgfierro in #273
• feat(evaluation): add champion tracking and flyeval CLI by @miguelgfierro in #274
• feat(lab): add retrieval metrics (hit@k, recall@k, MRR, MAP, nDCG) by @miguelgfierro in #275
• feat(examples): add flyradar and flycanon evaluation examples by @miguelgfierro in #276
• test(evaluation): add unit tests for evaluation package and retrieval metrics by @miguelgfierro in #277
• docs(evaluation): add evaluation package documentation by @miguelgfierro in #278
• fix(evaluation): resolve PR gate failures (lint, CI extras, remove flyradar example) by @miguelgfierro in #280
• feat(evaluation): simplify to 3-file metrics-only package by @miguelgfierro in #282
• fix: resolve CI gate failures — lint, typecheck, ruff by @miguelgfierro in #283
• refactor(evaluation): move retrieval_metrics from lab/ to evaluation/ by @miguelgfierro in #284
• refactor(evaluation): replace RetrieverMetrics class with plain functions by @miguelgfierro in #285
• Remove compute_retrieval_metrics() aggregate from evaluation by @miguelgfierro in #286
• Add GitHub Pages site — landing + MkDocs docs (correct install instructions) by @ancongui in #309
• feat(evaluation): migrate evaluation harnesses from playground by @miguelgfierro in #279

Full Changelog: v26.06.13...v26.06.14

v26.06.9

What's Changed

• docs: complete 26.06.x coverage + uniform middleware error lifecycle by @ancongui in #302

Full Changelog: v26.06.8...v26.06.9

v26.06.8

What's Changed

• fix: provider-aware reasoning-token cost (COST-001/002) by @ancongui in #301

Full Changelog: v26.06.7...v26.06.8

v26.06.13

What's Changed

• feat(observability)!: native instrumentation on by default + metrics de-dup by @ancongui in #307

Full Changelog: v26.06.12...v26.06.13

v26.06.12

What's Changed

• feat(observability): SP-10 native pydantic-ai OpenTelemetry instrumentation by @ancongui in #306

Full Changelog: v26.06.11...v26.06.12

v26.06.11

What's Changed

• feat(agents,tools)!: SP-3 native human-in-the-loop tool approval by @ancongui in #304
• style(agents): unquote ApprovalHandler annotation (resolve unused-import nit) by @ancongui in #305

Full Changelog: v26.06.10...v26.06.11

v26.06.10

What's Changed

• feat(reasoning): SP-5 native structured-output modes by @ancongui in #303

Full Changelog: v26.06.9...v26.06.10

v26.06.7

What's Changed

• chore: drop no-op ellipsis from AgentMiddleware.on_error by @ancongui in #299
• fix: audit P2/P3 hardening — provider robustness, durability, thread-safety by @ancongui in #300

Full Changelog: v26.06.6...v26.06.7

v26.06.6

What's Changed

• docs: refresh diagrams for workflows refactor + redesign banner & diagrams by @ancongui in #297
• fix: provider-agnosticism, cohesion & correctness hardening by @ancongui in #298

Full Changelog: v26.06.5...v26.06.6

v26.06.5

What's Changed

• refactor!: align workflows & tools with FireflyAgent and pydantic-ai by @ancongui in #296
• docs: README banner + 8 brand SVG diagrams (Firefly-family identity) by @ancongui in #295

Full Changelog: v26.06.4...v26.06.5