agentic-qe 2.1.2 → 2.2.1

package/dist/learning/algorithms/PPOLearner.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"PPOLearner.js","sourceRoot":"","sources":["../../../src/learning/algorithms/PPOLearner.ts"],"names":[],"mappings":";AAAA;;;;;;;;;;;;;;;;;;;;;GAqBG;;;AA2lBH,wDAqBC;AA9mBD,2DAA0E;AAqD1E;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAmCG;AACH,MAAa,UAAW,SAAQ,qCAAiB;IAQ/C,YAAY,MAAiB;QAC3B,KAAK,CAAC,MAAM,CAAC,CAAC;QACd,IAAI,CAAC,SAAS,GAAG,MAAM,CAAC;QACxB,IAAI,CAAC,WAAW,GAAG,IAAI,GAAG,EAAE,CAAC;QAC7B,IAAI,CAAC,UAAU,GAAG,IAAI,GAAG,EAAE,CAAC;QAC5B,IAAI,CAAC,cAAc,GAAG,IAAI,GAAG,EAAE,CAAC;QAChC,IAAI,CAAC,UAAU,GAAG,EAAE,CAAC;QACrB,IAAI,CAAC,kBAAkB,GAAG,MAAM,CAAC,eAAe,CAAC;QAEjD,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,wBAAwB,EAAE;YACzC,WAAW,EAAE,MAAM,CAAC,WAAW;YAC/B,MAAM,EAAE,MAAM,CAAC,SAAS;YACxB,SAAS,EAAE,MAAM,CAAC,SAAS;YAC3B,YAAY,EAAE,MAAM,CAAC,kBAAkB;SACxC,CAAC,CAAC;IACL,CAAC;IAED;;OAEG;IACM,YAAY,CAAC,KAAgB,EAAE,gBAA+B;QACrE,IAAI,gBAAgB,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YAClC,MAAM,IAAI,KAAK,CAAC,qCAAqC,CAAC,CAAC;QACzD,CAAC;QAED,MAAM,QAAQ,GAAG,IAAI,CAAC,WAAW,CAAC,KAAK,CAAC,CAAC;QACzC,MAAM,KAAK,GAAG,IAAI,CAAC,sBAAsB,CAAC,QAAQ,EAAE,gBAAgB,CAAC,CAAC;QAEtE,2BAA2B;QAC3B,MAAM,MAAM,GAAG,IAAI,CAAC,MAAM,EAAE,CAAC;QAC7B,IAAI,UAAU,GAAG,CAAC,CAAC;QAEnB,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,gBAAgB,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;YACjD,UAAU,IAAI,KAAK,CAAC,CAAC,CAAC,CAAC;YACvB,IAAI,MAAM,IAAI,UAAU,EAAE,CAAC;gBACzB,OAAO,gBAAgB,CAAC,CAAC,CAAC,CAAC;YAC7B,CAAC;QACH,CAAC;QAED,OAAO,gBAAgB,CAAC,gBAAgB,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;IACvD,CAAC;IAED;;OAEG;IACK,sBAAsB,CAAC,QAAgB,EAAE,gBAA+B;QAC9E,MAAM,WAAW,GAAa,EAAE,CAAC;QAEjC,KAAK,MAAM,MAAM,IAAI,gBAAgB,EAAE,CAAC;YACtC,MAAM,SAAS,GAAG,IAAI,CAAC,YAAY,CAAC,MAAM,CAAC,CAAC;YAC5C,MAAM,MAAM,GAAG,IAAI,CAAC,eAAe,CAAC,QAAQ,EAAE,SAAS,CAAC,CAAC;YACzD,WAAW,CAAC,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,CAAC;QACtC,CAAC;QAED,mCAAmC;QACnC,MAAM,OAAO,GAAG,IAAI,CAAC,GAAG,CAAC,GAAG,WAAW,CAAC,CAAC;QACzC,MAAM,QAAQ,GAAG,WAAW,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,GAAG,OAAO,CAAC,CAAC,CAAC;QAC7D,MAAM,MAAM,GAAG,QAAQ,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,CAAC,EAAE,EAAE,CAAC,GAAG,GAAG,CAAC,EAAE,CAAC,CAAC,CAAC;QAEvD,OAAO,QAAQ,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,GAAG,MAAM,CAAC,CAAC;IACvC,CAAC;IAED;;OAEG;IACK,eAAe,CAAC,QAAgB,EAAE,SAAiB;QACzD,MAAM,WAAW,GAAG,IAAI,CAAC,WAAW,CAAC,GAAG,CAAC,QAAQ,CAAC,CAAC;QACnD,IAAI,CAAC,WAAW,EAAE,CAAC;YACjB,OAAO,EAAE,UAAU,EAAE,CAAC,EAAE,OAAO,EAAE,CAAC,EAAE,WAAW,EAAE,CAAC,EAAE,CAAC;QACvD,CAAC;QACD,OAAO,WAAW,CAAC,GAAG,CAAC,SAAS,CAAC,IAAI,EAAE,UAAU,EAAE,CAAC,EAAE,OAAO,EAAE,CAAC,EAAE,WAAW,EAAE,CAAC,EAAE,CAAC;IACrF,CAAC;IAED;;OAEG;IACK,UAAU,CAAC,QAAgB,EAAE,SAAiB,EAAE,gBAAgC;QACtF,mCAAmC;QACnC,MAAM,MAAM,GAAG,IAAI,CAAC,eAAe,CAAC,QAAQ,EAAE,SAAS,CAAC,CAAC;QAEzD,4DAA4D;QAC5D,IAAI,CAAC,gBAAgB,EAAE,CAAC;YACtB,OAAO,MAAM,CAAC,OAAO,CAAC;QACxB,CAAC;QAED,mCAAmC;QACnC,MAAM,KAAK,GAAa,EAAE,CAAC;QAC3B,IAAI,UAAU,GAAG,MAAM,CAAC,UAAU,CAAC;QAEnC,KAAK,MAAM,MAAM,IAAI,gBAAgB,EAAE,CAAC;YACtC,MAAM,EAAE,GAAG,IAAI,CAAC,YAAY,CAAC,MAAM,CAAC,CAAC;YACrC,MAAM,CAAC,GAAG,IAAI,CAAC,eAAe,CAAC,QAAQ,EAAE,EAAE,CAAC,CAAC;YAC7C,KAAK,CAAC,IAAI,CAAC,CAAC,CAAC,UAAU,CAAC,CAAC;YACzB,IAAI,EAAE,KAAK,SAAS,EAAE,CAAC;gBACrB,UAAU,GAAG,CAAC,CAAC,UAAU,CAAC;YAC5B,CAAC;QACH,CAAC;QAED,MAAM,OAAO,GAAG,IAAI,CAAC,GAAG,CAAC,GAAG,KAAK,EAAE,UAAU,CAAC,CAAC;QAC/C,MAAM,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,UAAU,GAAG,OAAO,CAAC,CAAC;QACjD,MAAM,MAAM,GAAG,KAAK,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,CAAC,EAAE,EAAE,CAAC,GAAG,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,GAAG,OAAO,CAAC,EAAE,CAAC,CAAC,CAAC;QAExE,OAAO,IAAI,CAAC,GAAG,CAAC,SAAS,GAAG,MAAM,CAAC,CAAC;IACtC,CAAC;IAED;;OAEG;IACM,aAAa,CAAC,KAAgB;QACrC,MAAM,QAAQ,GAAG,IAAI,CAAC,WAAW,CAAC,KAAK,CAAC,CAAC;QACzC,OAAO,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC;IAC5C,CAAC;IAED;;OAEG;IACH,WAAW,CACT,KAAgB,EAChB,MAAmB,EACnB,MAAc,EACd,SAAoB,EACpB,IAAa;QAEb,MAAM,QAAQ,GAAG,IAAI,CAAC,WAAW,CAAC,KAAK,CAAC,CAAC;QACzC,MAAM,SAAS,GAAG,IAAI,CAAC,YAAY,CAAC,MAAM,CAAC,CAAC;QAC5C,MAAM,YAAY,GAAG,IAAI,CAAC,WAAW,CAAC,SAAS,CAAC,CAAC;QAEjD,MAAM,KAAK,GAAG,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC;QACjD,MAAM,OAAO,GAAG,IAAI,CAAC,UAAU,CAAC,QAAQ,EAAE,SAAS,CAAC,CAAC;QAErD,IAAI,CAAC,UAAU,CAAC,IAAI,CAAC;YACnB,KAAK,EAAE,QAAQ;YACf,MAAM,EAAE,SAAS;YACjB,MAAM;YACN,SAAS,EAAE,YAAY;YACvB,IAAI;YACJ,KAAK;YACL,OAAO;YACP,SAAS,EAAE,CAAC,EAAE,iBAAiB;YAC/B,OAAO,EAAE,CAAC,CAAI,iBAAiB;SAChC,CAAC,CAAC;IACL,CAAC;IAED;;OAEG;IACM,MAAM,CAAC,UAA0B,EAAE,UAAwB;QAClE,IAAI,CAAC,SAAS,EAAE,CAAC;QAEjB,MAAM,EAAE,KAAK,EAAE,MAAM,EAAE,MAAM,EAAE,SAAS,EAAE,GAAG,UAAU,CAAC;QACxD,MAAM,IAAI,GAAG,UAAU,CAAC,IAAI,IAAI,KAAK,CAAC;QAEtC,eAAe;QACf,IAAI,CAAC,WAAW,CAAC,KAAK,EAAE,MAAM,EAAE,MAAM,EAAE,SAAS,EAAE,IAAI,CAAC,CAAC;QAEzD,wCAAwC;QACxC,IAAI,IAAI,CAAC,UAAU,CAAC,MAAM,IAAI,IAAI,CAAC,SAAS,CAAC,gBAAgB,EAAE,CAAC;YAC9D,IAAI,CAAC,iBAAiB,EAAE,CAAC;QAC3B,CAAC;IACH,CAAC;IAED;;OAEG;IACH,iBAAiB;QACf,IAAI,IAAI,CAAC,UAAU,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YACjC,OAAO;QACT,CAAC;QAED,+BAA+B;QAC/B,IAAI,CAAC,UAAU,EAAE,CAAC;QAElB,wCAAwC;QACxC,IAAI,CAAC,aAAa,EAAE,CAAC;QAErB,8BAA8B;QAC9B,KAAK,IAAI,KAAK,GAAG,CAAC,EAAE,KAAK,GAAG,IAAI,CAAC,SAAS,CAAC,SAAS,EAAE,KAAK,EAAE,EAAE,CAAC;YAC9D,IAAI,CAAC,UAAU,EAAE,CAAC;QACpB,CAAC;QAED,mBAAmB;QACnB,IAAI,CAAC,UAAU,GAAG,EAAE,CAAC;QAErB,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,uBAAuB,EAAE;YACxC,MAAM,EAAE,IAAI,CAAC,SAAS,CAAC,SAAS;YAChC,KAAK,EAAE,IAAI,CAAC,SAAS;SACtB,CAAC,CAAC;IACL,CAAC;IAED;;;;;OAKG;IACK,UAAU;QAChB,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,cAAc,CAAC;QACzC,MAAM,MAAM,GAAG,IAAI,CAAC,SAAS,CAAC,SAAS,CAAC;QAExC,IAAI,UAAU,GAAG,CAAC,CAAC;QACnB,MAAM,CAAC,GAAG,IAAI,CAAC,UAAU,CAAC,MAAM,CAAC;QAEjC,2CAA2C;QAC3C,KAAK,IAAI,CAAC,GAAG,CAAC,GAAG,CAAC,EAAE,CAAC,IAAI,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC;YAChC,MAAM,IAAI,GAAG,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC,CAAC;YAEhC,MAAM,SAAS,GAAG,IAAI,CAAC,IAAI;gBACzB,CAAC,CAAC,CAAC;gBACH,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC;YAE1F,WAAW;YACX,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,GAAG,KAAK,GAAG,SAAS,GAAG,IAAI,CAAC,KAAK,CAAC;YAE3D,gBAAgB;YAChB,UAAU,GAAG,IAAI,CAAC,IAAI;gBACpB,CAAC,CAAC,KAAK;gBACP,CAAC,CAAC,KAAK,GAAG,KAAK,GAAG,MAAM,GAAG,UAAU,CAAC;YAExC,IAAI,CAAC,SAAS,GAAG,UAAU,CAAC;YAC5B,IAAI,CAAC,OAAO,GAAG,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC,KAAK,CAAC;QAC7C,CAAC;QAED,uBAAuB;QACvB,MAAM,UAAU,GAAG,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC;QACzD,MAAM,IAAI,GAAG,UAAU,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC,GAAG,UAAU,CAAC,MAAM,CAAC;QACvE,MAAM,QAAQ,GAAG,UAAU,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,GAAG,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,GAAG,UAAU,CAAC,MAAM,CAAC;QACzF,MAAM,GAAG,GAAG,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,GAAG,IAAI,CAAC;QAEvC,KAAK,MAAM,IAAI,IAAI,IAAI,CAAC,UAAU,EAAE,CAAC;YACnC,IAAI,CAAC,SAAS,GAAG,CAAC,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC,GAAG,GAAG,CAAC;QACjD,CAAC;IACH,CAAC;IAED;;OAEG;IACK,aAAa;QACnB,IAAI,CAAC,cAAc,CAAC,KAAK,EAAE,CAAC;QAC5B,KAAK,MAAM,CAAC,KAAK,EAAE,OAAO,CAAC,IAAI,IAAI,CAAC,WAAW,CAAC,OAAO,EAAE,EAAE,CAAC;YAC1D,MAAM,SAAS,GAAG,IAAI,GAAG,EAAwB,CAAC;YAClD,KAAK,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,IAAI,OAAO,CAAC,OAAO,EAAE,EAAE,CAAC;gBACjD,SAAS,CAAC,GAAG,CAAC,MAAM,EAAE,EAAE,GAAG,MAAM,EAAE,CAAC,CAAC;YACvC,CAAC;YACD,IAAI,CAAC,cAAc,CAAC,GAAG,CAAC,KAAK,EAAE,SAAS,CAAC,CAAC;QAC5C,CAAC;IACH,CAAC;IAED;;OAEG;IACK,aAAa,CAAC,QAAgB,EAAE,SAAiB;QACvD,MAAM,WAAW,GAAG,IAAI,CAAC,cAAc,CAAC,GAAG,CAAC,QAAQ,CAAC,CAAC;QACtD,IAAI,CAAC,WAAW,EAAE,CAAC;YACjB,OAAO,CAAC,CAAC;QACX,CAAC;QACD,OAAO,WAAW,CAAC,GAAG,CAAC,SAAS,CAAC,EAAE,OAAO,IAAI,CAAC,CAAC;IAClD,CAAC;IAED;;OAEG;IACK,UAAU;QAChB,qBAAqB;QACrB,MAAM,QAAQ,GAAG,CAAC,GAAG,IAAI,CAAC,UAAU,CAAC,CAAC,IAAI,CAAC,GAAG,EAAE,CAAC,IAAI,CAAC,MAAM,EAAE,GAAG,GAAG,CAAC,CAAC;QAEtE,qBAAqB;QACrB,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,QAAQ,CAAC,MAAM,EAAE,CAAC,IAAI,IAAI,CAAC,SAAS,CAAC,aAAa,EAAE,CAAC;YACvE,MAAM,KAAK,GAAG,QAAQ,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,GAAG,IAAI,CAAC,SAAS,CAAC,aAAa,CAAC,CAAC;YAClE,IAAI,CAAC,cAAc,CAAC,KAAK,CAAC,CAAC;QAC7B,CAAC;IACH,CAAC;IAED;;OAEG;IACK,cAAc,CAAC,KAAuB;QAC5C,KAAK,MAAM,IAAI,IAAI,KAAK,EAAE,CAAC;YACzB,4BAA4B;YAC5B,MAAM,UAAU,GAAG,IAAI,CAAC,UAAU,CAAC,IAAI,CAAC,KAAK,EAAE,IAAI,CAAC,MAAM,CAAC,CAAC;YAC5D,MAAM,UAAU,GAAG,IAAI,CAAC,OAAO,CAAC,CAAC,sBAAsB;YACvD,MAAM,KAAK,GAAG,IAAI,CAAC,GAAG,CAAC,UAAU,GAAG,UAAU,CAAC,CAAC;YAEhD,2CAA2C;YAC3C,MAAM,GAAG,GAAG,IAAI,CAAC,SAAS,CAAC,WAAW,CAAC;YACvC,MAAM,KAAK,GAAG,KAAK,GAAG,IAAI,CAAC,SAAS,CAAC;YACrC,MAAM,KAAK,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,KAAK,EAAE,CAAC,GAAG,GAAG,CAAC,EAAE,CAAC,GAAG,GAAG,CAAC,GAAG,IAAI,CAAC,SAAS,CAAC;YAE3E,qDAAqD;YACrD,MAAM,UAAU,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC;YAE3C,aAAa;YACb,MAAM,WAAW,GAAG,IAAI,CAAC,OAAO,CAAC;YACjC,MAAM,YAAY,GAAG,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;YAC1D,IAAI,SAAS,GAAG,CAAC,YAAY,GAAG,WAAW,CAAC,IAAI,CAAC,CAAC;YAElD,6BAA6B;YAC7B,IAAI,IAAI,CAAC,SAAS,CAAC,aAAa,EAAE,CAAC;gBACjC,MAAM,YAAY,GAAG,IAAI,CAAC,KAAK,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,YAAY,GAAG,IAAI,CAAC,KAAK,EAAE,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,CAAC;gBAC3F,MAAM,gBAAgB,GAAG,CAAC,YAAY,GAAG,WAAW,CAAC,IAAI,CAAC,CAAC;gBAC3D,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,SAAS,EAAE,gBAAgB,CAAC,CAAC;YACpD,CAAC;YAED,gBAAgB;YAChB,MAAM,OAAO,GAAG,IAAI,CAAC,cAAc,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;YAChD,MAAM,WAAW,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,kBAAkB,GAAG,OAAO,CAAC;YAEjE,aAAa;YACb,MAAM,SAAS,GAAG,UAAU,GAAG,IAAI,CAAC,SAAS,CAAC,oBAAoB,GAAG,SAAS,GAAG,WAAW,CAAC;YAE7F,4CAA4C;YAC5C,IAAI,CAAC,YAAY,CAAC,IAAI,CAAC,KAAK,EAAE,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,SAAS,EAAE,KAAK,CAAC,CAAC;YAElE,wBAAwB;YACxB,IAAI,CAAC,WAAW,CAAC,IAAI,CAAC,KAAK,EAAE,WAAW,CAAC,CAAC;QAC5C,CAAC;IACH,CAAC;IAED;;OAEG;IACK,YAAY,CAClB,QAAgB,EAChB,SAAiB,EACjB,SAAiB,EACjB,KAAa;QAEb,IAAI,CAAC,IAAI,CAAC,WAAW,CAAC,GAAG,CAAC,QAAQ,CAAC,EAAE,CAAC;YACpC,IAAI,CAAC,WAAW,CAAC,GAAG,CAAC,QAAQ,EAAE,IAAI,GAAG,EAAE,CAAC,CAAC;QAC5C,CAAC;QACD,MAAM,WAAW,GAAG,IAAI,CAAC,WAAW,CAAC,GAAG,CAAC,QAAQ,CAAE,CAAC;QAEpD,MAAM,OAAO,GAAG,WAAW,CAAC,GAAG,CAAC,SAAS,CAAC,IAAI,EAAE,UAAU,EAAE,CAAC,EAAE,OAAO,EAAE,CAAC,EAAE,WAAW,EAAE,CAAC,EAAE,CAAC;QAE5F,mBAAmB;QACnB,MAAM,GAAG,GAAG,IAAI,CAAC,SAAS,CAAC,WAAW,CAAC;QACvC,IAAI,QAAQ,GAAG,SAAS,CAAC;QACzB,IAAI,CAAC,KAAK,GAAG,CAAC,GAAG,GAAG,IAAI,SAAS,GAAG,CAAC,CAAC,IAAI,CAAC,KAAK,GAAG,CAAC,GAAG,GAAG,IAAI,SAAS,GAAG,CAAC,CAAC,EAAE,CAAC;YAC7E,QAAQ,GAAG,CAAC,CAAC,CAAC,sBAAsB;QACtC,CAAC;QAED,oBAAoB;QACpB,MAAM,aAAa,GAAG,OAAO,CAAC,UAAU,GAAG,IAAI,CAAC,SAAS,CAAC,kBAAkB,GAAG,QAAQ,CAAC;QACxF,MAAM,UAAU,GAAG,IAAI,CAAC,UAAU,CAAC,QAAQ,EAAE,SAAS,CAAC,CAAC;QAExD,WAAW,CAAC,GAAG,CAAC,SAAS,EAAE;YACzB,UAAU,EAAE,aAAa;YACzB,OAAO,EAAE,UAAU;YACnB,WAAW,EAAE,OAAO,CAAC,WAAW,GAAG,CAAC;SACrC,CAAC,CAAC;QAEH,mCAAmC;QACnC,IAAI,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,aAAa,CAAC,CAAC;IACrD,CAAC;IAED;;OAEG;IACK,WAAW,CAAC,QAAgB,EAAE,MAAc;QAClD,MAAM,OAAO,GAAG,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC;QACnD,MAAM,QAAQ,GAAG,OAAO,GAAG,IAAI,CAAC,SAAS,CAAC,iBAAiB,GAAG,CAAC,MAAM,GAAG,OAAO,CAAC,CAAC;QACjF,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,QAAQ,EAAE,QAAQ,CAAC,CAAC;IAC1C,CAAC;IAED;;OAEG;IACK,cAAc,CAAC,QAAgB;QACrC,MAAM,WAAW,GAAG,IAAI,CAAC,WAAW,CAAC,GAAG,CAAC,QAAQ,CAAC,CAAC;QACnD,IAAI,CAAC,WAAW,IAAI,WAAW,CAAC,IAAI,KAAK,CAAC,EAAE,CAAC;YAC3C,OAAO,CAAC,CAAC;QACX,CAAC;QAED,MAAM,KAAK,GAAG,KAAK,CAAC,IAAI,CAAC,WAAW,CAAC,MAAM,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,UAAU,CAAC,CAAC;QACtE,MAAM,OAAO,GAAG,IAAI,CAAC,GAAG,CAAC,GAAG,KAAK,CAAC,CAAC;QACnC,MAAM,QAAQ,GAAG,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,GAAG,OAAO,CAAC,CAAC,CAAC;QACvD,MAAM,MAAM,GAAG,QAAQ,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC,CAAC;QACnD,MAAM,KAAK,GAAG,QAAQ,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,GAAG,MAAM,CAAC,CAAC;QAE5C,IAAI,OAAO,GAAG,CAAC,CAAC;QAChB,KAAK,MAAM,CAAC,IAAI,KAAK,EAAE,CAAC;YACtB,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC;gBACV,OAAO,IAAI,CAAC,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;YAC7B,CAAC;QACH,CAAC;QAED,OAAO,OAAO,CAAC;IACjB,CAAC;IAED;;OAEG;IACO,yBAAyB;QACjC,OAAO,IAAI,CAAC,kBAAkB,CAAC;IACjC,CAAC;IAED;;OAEG;IACH,gBAAgB;QAQd,IAAI,UAAU,GAAG,CAAC,CAAC;QACnB,KAAK,MAAM,CAAC,IAAI,IAAI,CAAC,UAAU,CAAC,MAAM,EAAE,EAAE,CAAC;YACzC,UAAU,IAAI,CAAC,CAAC;QAClB,CAAC;QAED,IAAI,UAAU,GAAG,CAAC,CAAC;QACnB,KAAK,MAAM,WAAW,IAAI,IAAI,CAAC,WAAW,CAAC,MAAM,EAAE,EAAE,CAAC;YACpD,UAAU,IAAI,WAAW,CAAC,IAAI,CAAC;QACjC,CAAC;QAED,MAAM,YAAY,GAAG,IAAI,CAAC,UAAU,CAAC,MAAM,GAAG,CAAC;YAC7C,CAAC,CAAC,IAAI,CAAC,UAAU,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,SAAS,EAAE,CAAC,CAAC,GAAG,IAAI,CAAC,UAAU,CAAC,MAAM;YAC/E,CAAC,CAAC,CAAC,CAAC;QAEN,OAAO;YACL,gBAAgB,EAAE,IAAI,CAAC,UAAU,CAAC,MAAM;YACxC,cAAc,EAAE,IAAI,CAAC,UAAU,CAAC,IAAI;YACpC,eAAe,EAAE,UAAU;YAC3B,QAAQ,EAAE,IAAI,CAAC,UAAU,CAAC,IAAI,GAAG,CAAC,CAAC,CAAC,CAAC,UAAU,GAAG,IAAI,CAAC,UAAU,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;YAC1E,YAAY;YACZ,YAAY,EAAE,CAAC,CAAC,sCAAsC;SACvD,CAAC;IACJ,CAAC;IAED;;OAEG;IACM,KAAK;QACZ,KAAK,CAAC,KAAK,EAAE,CAAC;QACd,IAAI,CAAC,WAAW,CAAC,KAAK,EAAE,CAAC;QACzB,IAAI,CAAC,UAAU,CAAC,KAAK,EAAE,CAAC;QACxB,IAAI,CAAC,cAAc,CAAC,KAAK,EAAE,CAAC;QAC5B,IAAI,CAAC,UAAU,GAAG,EAAE,CAAC;QACrB,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,kBAAkB,CAAC,CAAC;IACvC,CAAC;IAED;;OAEG;IACH,SAAS;QAMP,MAAM,gBAAgB,GAAiD,EAAE,CAAC;QAC1E,KAAK,MAAM,CAAC,KAAK,EAAE,OAAO,CAAC,IAAI,IAAI,CAAC,WAAW,CAAC,OAAO,EAAE,EAAE,CAAC;YAC1D,gBAAgB,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC;YAC7B,KAAK,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,IAAI,OAAO,CAAC,OAAO,EAAE,EAAE,CAAC;gBACjD,gBAAgB,CAAC,KAAK,CAAC,CAAC,MAAM,CAAC,GAAG,MAAM,CAAC;YAC3C,CAAC;QACH,CAAC;QAED,MAAM,eAAe,GAA2B,EAAE,CAAC;QACnD,KAAK,MAAM,CAAC,KAAK,EAAE,KAAK,CAAC,IAAI,IAAI,CAAC,UAAU,CAAC,OAAO,EAAE,EAAE,CAAC;YACvD,eAAe,CAAC,KAAK,CAAC,GAAG,KAAK,CAAC;QACjC,CAAC;QAED,OAAO;YACL,IAAI,EAAE,IAAI,CAAC,MAAM,EAAE;YACnB,WAAW,EAAE,gBAAgB;YAC7B,UAAU,EAAE,eAAe;YAC3B,SAAS,EAAE,EAAE,GAAG,IAAI,CAAC,SAAS,EAAE;SACjC,CAAC;IACJ,CAAC;IAED;;OAEG;IACH,SAAS,CAAC,KAAwC;QAChD,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;QAExB,IAAI,CAAC,WAAW,CAAC,KAAK,EAAE,CAAC;QACzB,KAAK,MAAM,CAAC,QAAQ,EAAE,OAAO,CAAC,IAAI,MAAM,CAAC,OAAO,CAAC,KAAK,CAAC,WAAW,CAAC,EAAE,CAAC;YACpE,MAAM,SAAS,GAAG,IAAI,GAAG,EAAwB,CAAC;YAClD,KAAK,MAAM,CAAC,SAAS,EAAE,MAAM,CAAC,IAAI,MAAM,CAAC,OAAO,CAAC,OAAO,CAAC,EAAE,CAAC;gBAC1D,SAAS,CAAC,GAAG,CAAC,SAAS,EAAE,MAAM,CAAC,CAAC;YACnC,CAAC;YACD,IAAI,CAAC,WAAW,CAAC,GAAG,CAAC,QAAQ,EAAE,SAAS,CAAC,CAAC;QAC5C,CAAC;QAED,IAAI,CAAC,UAAU,CAAC,KAAK,EAAE,CAAC;QACxB,KAAK,MAAM,CAAC,QAAQ,EAAE,KAAK,CAAC,IAAI,MAAM,CAAC,OAAO,CAAC,KAAK,CAAC,UAAU,CAAC,EAAE,CAAC;YACjE,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,QAAQ,EAAE,KAAK,CAAC,CAAC;QACvC,CAAC;QAED,IAAI,CAAC,SAAS,GAAG,EAAE,GAAG,KAAK,CAAC,SAAS,EAAE,CAAC;QAExC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,oBAAoB,EAAE;YACrC,UAAU,EAAE,IAAI,CAAC,WAAW,CAAC,IAAI;YACjC,SAAS,EAAE,IAAI,CAAC,UAAU,CAAC,IAAI;SAChC,CAAC,CAAC;IACL,CAAC;CACF;AA3fD,gCA2fC;AAED;;GAEG;AACH,SAAgB,sBAAsB;IACpC,OAAO;QACL,YAAY,EAAE,MAAM;QACpB,cAAc,EAAE,IAAI;QACpB,eAAe,EAAE,GAAG,EAAE,mCAAmC;QACzD,gBAAgB,EAAE,GAAG;QACrB,kBAAkB,EAAE,GAAG;QACvB,WAAW,EAAE,GAAG;QAChB,SAAS,EAAE,CAAC;QACZ,aAAa,EAAE,EAAE;QACjB,oBAAoB,EAAE,GAAG;QACzB,kBAAkB,EAAE,IAAI;QACxB,SAAS,EAAE,IAAI;QACf,WAAW,EAAE,GAAG;QAChB,aAAa,EAAE,IAAI;QACnB,kBAAkB,EAAE,MAAM;QAC1B,iBAAiB,EAAE,KAAK;QACxB,mBAAmB,EAAE,KAAK,EAAE,gCAAgC;QAC5D,gBAAgB,EAAE,IAAI,EAAM,iCAAiC;QAC7D,SAAS,EAAE,EAAE;KACd,CAAC;AACJ,CAAC"}

package/dist/learning/algorithms/QLearning.d.ts

@@ -0,0 +1,68 @@
+/**
+ * QLearning - Off-policy TD(0) Reinforcement Learning
+ *
+ * Implements standard Q-learning algorithm for reinforcement learning.
+ * Key differences from SARSA:
+ * - Off-policy: learns optimal Q-values regardless of policy being followed
+ * - Uses max Q-value for next state, not actual next action
+ * - Update rule: Q(s,a) ← Q(s,a) + α[r + γ·max(Q(s',a')) - Q(s,a)]
+ * - More aggressive than SARSA, finds optimal policy faster
+ */
+import { AbstractRLLearner, RLConfig } from './AbstractRLLearner';
+import { TaskExperience, AgentAction } from '../types';
+/**
+ * Q-learning configuration (extends base RL config)
+ */
+export interface QLearningConfig extends RLConfig {
+}
+/**
+ * QLearning - Standard Q-learning implementation
+ *
+ * Implements the classic Q-learning algorithm with:
+ * - Epsilon-greedy exploration policy
+ * - Off-policy temporal difference (TD) learning
+ * - Q-table for state-action values
+ * - Optional experience replay for stability
+ *
+ * Update Rule:
+ * Q(s,a) ← Q(s,a) + α[r + γ·max_a'(Q(s',a')) - Q(s,a)]
+ *
+ * Key characteristics:
+ * - Off-policy: learns about optimal policy while following exploration policy
+ * - Uses max Q-value (greedy) for bootstrapping
+ * - Converges to optimal Q* under certain conditions
+ * - More sample-efficient than on-policy methods
+ */
+export declare class QLearning extends AbstractRLLearner {
+    private readonly defaultConfig;
+    constructor(config?: Partial<QLearningConfig>);
+    /**
+     * Update Q-value using Q-learning update rule
+     * Q(s,a) ← Q(s,a) + α[r + γ·max(Q(s',a')) - Q(s,a)]
+     *
+     * @param experience The transition experience (s, a, r, s')
+     * @param nextAction Ignored in Q-learning (uses max Q-value instead)
+     */
+    update(experience: TaskExperience, nextAction?: AgentAction): void;
+    /**
+     * Get the default exploration rate for this algorithm
+     */
+    protected getDefaultExplorationRate(): number;
+    /**
+     * Get algorithm name
+     */
+    getAlgorithmName(): string;
+    /**
+     * Get algorithm type (off-policy)
+     */
+    getAlgorithmType(): 'on-policy' | 'off-policy';
+    /**
+     * Get detailed statistics including Q-learning-specific metrics
+     */
+    getDetailedStatistics(): {
+        algorithm: string;
+        type: 'on-policy' | 'off-policy';
+        stats: ReturnType<AbstractRLLearner['getStatistics']>;
+    };
+}
+//# sourceMappingURL=QLearning.d.ts.map

package/dist/learning/algorithms/QLearning.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"QLearning.d.ts","sourceRoot":"","sources":["../../../src/learning/algorithms/QLearning.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AAEH,OAAO,EAAE,iBAAiB,EAAE,QAAQ,EAAE,MAAM,qBAAqB,CAAC;AAClE,OAAO,EAAE,cAAc,EAAE,WAAW,EAAE,MAAM,UAAU,CAAC;AAEvD;;GAEG;AACH,MAAM,WAAW,eAAgB,SAAQ,QAAQ;CAGhD;AAgBD;;;;;;;;;;;;;;;;;GAiBG;AACH,qBAAa,SAAU,SAAQ,iBAAiB;IAC9C,OAAO,CAAC,QAAQ,CAAC,aAAa,CAAkB;gBAEpC,MAAM,GAAE,OAAO,CAAC,eAAe,CAAM;IAOjD;;;;;;OAMG;IACH,MAAM,CAAC,UAAU,EAAE,cAAc,EAAE,UAAU,CAAC,EAAE,WAAW,GAAG,IAAI;IAiClE;;OAEG;IACH,SAAS,CAAC,yBAAyB,IAAI,MAAM;IAI7C;;OAEG;IACH,gBAAgB,IAAI,MAAM;IAI1B;;OAEG;IACH,gBAAgB,IAAI,WAAW,GAAG,YAAY;IAI9C;;OAEG;IACH,qBAAqB,IAAI;QACvB,SAAS,EAAE,MAAM,CAAC;QAClB,IAAI,EAAE,WAAW,GAAG,YAAY,CAAC;QACjC,KAAK,EAAE,UAAU,CAAC,iBAAiB,CAAC,eAAe,CAAC,CAAC,CAAC;KACvD;CAOF"}

package/dist/learning/algorithms/QLearning.js

@@ -0,0 +1,116 @@
+"use strict";
+/**
+ * QLearning - Off-policy TD(0) Reinforcement Learning
+ *
+ * Implements standard Q-learning algorithm for reinforcement learning.
+ * Key differences from SARSA:
+ * - Off-policy: learns optimal Q-values regardless of policy being followed
+ * - Uses max Q-value for next state, not actual next action
+ * - Update rule: Q(s,a) ← Q(s,a) + α[r + γ·max(Q(s',a')) - Q(s,a)]
+ * - More aggressive than SARSA, finds optimal policy faster
+ */
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.QLearning = void 0;
+const AbstractRLLearner_1 = require("./AbstractRLLearner");
+/**
+ * Default Q-learning configuration
+ */
+const DEFAULT_CONFIG = {
+    learningRate: 0.1,
+    discountFactor: 0.95,
+    explorationRate: 0.3,
+    explorationDecay: 0.995,
+    minExplorationRate: 0.01,
+    useExperienceReplay: true,
+    replayBufferSize: 10000,
+    batchSize: 32
+};
+/**
+ * QLearning - Standard Q-learning implementation
+ *
+ * Implements the classic Q-learning algorithm with:
+ * - Epsilon-greedy exploration policy
+ * - Off-policy temporal difference (TD) learning
+ * - Q-table for state-action values
+ * - Optional experience replay for stability
+ *
+ * Update Rule:
+ * Q(s,a) ← Q(s,a) + α[r + γ·max_a'(Q(s',a')) - Q(s,a)]
+ *
+ * Key characteristics:
+ * - Off-policy: learns about optimal policy while following exploration policy
+ * - Uses max Q-value (greedy) for bootstrapping
+ * - Converges to optimal Q* under certain conditions
+ * - More sample-efficient than on-policy methods
+ */
+class QLearning extends AbstractRLLearner_1.AbstractRLLearner {
+    constructor(config = {}) {
+        const fullConfig = { ...DEFAULT_CONFIG, ...config };
+        super(fullConfig);
+        this.defaultConfig = fullConfig;
+        this.logger.info('QLearning initialized with off-policy TD(0)', { config: fullConfig });
+    }
+    /**
+     * Update Q-value using Q-learning update rule
+     * Q(s,a) ← Q(s,a) + α[r + γ·max(Q(s',a')) - Q(s,a)]
+     *
+     * @param experience The transition experience (s, a, r, s')
+     * @param nextAction Ignored in Q-learning (uses max Q-value instead)
+     */
+    update(experience, nextAction) {
+        const stateKey = this.encodeState(experience.state);
+        const actionKey = this.encodeAction(experience.action);
+        const nextStateKey = this.encodeState(experience.nextState);
+        // Get current Q-value Q(s,a)
+        const stateActions = this.qTable.get(stateKey);
+        const currentQ = stateActions?.get(actionKey)?.value ?? 0;
+        // Q-Learning: Get max Q-value for next state (greedy)
+        // This is the key difference from SARSA (which uses actual next action)
+        const nextStateActions = this.qTable.get(nextStateKey);
+        const maxNextQ = nextStateActions && nextStateActions.size > 0
+            ? Math.max(...Array.from(nextStateActions.values()).map(qv => qv.value))
+            : 0;
+        // Q-learning update rule
+        // Q(s,a) = Q(s,a) + α * [r + γ * max(Q(s',a')) - Q(s,a)]
+        const tdTarget = experience.reward + this.config.discountFactor * maxNextQ;
+        const tdError = tdTarget - currentQ;
+        const newQ = currentQ + this.config.learningRate * tdError;
+        // Update Q-value
+        this.setQValue(stateKey, actionKey, newQ);
+        // Add to experience replay buffer if enabled
+        if (this.replayBuffer) {
+            this.replayBuffer.add(experience, Math.abs(tdError)); // Priority based on TD error
+        }
+        this.stepCount++;
+    }
+    /**
+     * Get the default exploration rate for this algorithm
+     */
+    getDefaultExplorationRate() {
+        return this.defaultConfig.explorationRate;
+    }
+    /**
+     * Get algorithm name
+     */
+    getAlgorithmName() {
+        return 'Q-Learning';
+    }
+    /**
+     * Get algorithm type (off-policy)
+     */
+    getAlgorithmType() {
+        return 'off-policy';
+    }
+    /**
+     * Get detailed statistics including Q-learning-specific metrics
+     */
+    getDetailedStatistics() {
+        return {
+            algorithm: this.getAlgorithmName(),
+            type: this.getAlgorithmType(),
+            stats: this.getStatistics()
+        };
+    }
+}
+exports.QLearning = QLearning;
+//# sourceMappingURL=QLearning.js.map

package/dist/learning/algorithms/QLearning.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"QLearning.js","sourceRoot":"","sources":["../../../src/learning/algorithms/QLearning.ts"],"names":[],"mappings":";AAAA;;;;;;;;;GASG;;;AAEH,2DAAkE;AAWlE;;GAEG;AACH,MAAM,cAAc,GAAoB;IACtC,YAAY,EAAE,GAAG;IACjB,cAAc,EAAE,IAAI;IACpB,eAAe,EAAE,GAAG;IACpB,gBAAgB,EAAE,KAAK;IACvB,kBAAkB,EAAE,IAAI;IACxB,mBAAmB,EAAE,IAAI;IACzB,gBAAgB,EAAE,KAAK;IACvB,SAAS,EAAE,EAAE;CACd,CAAC;AAEF;;;;;;;;;;;;;;;;;GAiBG;AACH,MAAa,SAAU,SAAQ,qCAAiB;IAG9C,YAAY,SAAmC,EAAE;QAC/C,MAAM,UAAU,GAAG,EAAE,GAAG,cAAc,EAAE,GAAG,MAAM,EAAE,CAAC;QACpD,KAAK,CAAC,UAAU,CAAC,CAAC;QAClB,IAAI,CAAC,aAAa,GAAG,UAAU,CAAC;QAChC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,6CAA6C,EAAE,EAAE,MAAM,EAAE,UAAU,EAAE,CAAC,CAAC;IAC1F,CAAC;IAED;;;;;;OAMG;IACH,MAAM,CAAC,UAA0B,EAAE,UAAwB;QACzD,MAAM,QAAQ,GAAG,IAAI,CAAC,WAAW,CAAC,UAAU,CAAC,KAAK,CAAC,CAAC;QACpD,MAAM,SAAS,GAAG,IAAI,CAAC,YAAY,CAAC,UAAU,CAAC,MAAM,CAAC,CAAC;QACvD,MAAM,YAAY,GAAG,IAAI,CAAC,WAAW,CAAC,UAAU,CAAC,SAAS,CAAC,CAAC;QAE5D,6BAA6B;QAC7B,MAAM,YAAY,GAAG,IAAI,CAAC,MAAM,CAAC,GAAG,CAAC,QAAQ,CAAC,CAAC;QAC/C,MAAM,QAAQ,GAAG,YAAY,EAAE,GAAG,CAAC,SAAS,CAAC,EAAE,KAAK,IAAI,CAAC,CAAC;QAE1D,sDAAsD;QACtD,wEAAwE;QACxE,MAAM,gBAAgB,GAAG,IAAI,CAAC,MAAM,CAAC,GAAG,CAAC,YAAY,CAAC,CAAC;QACvD,MAAM,QAAQ,GAAG,gBAAgB,IAAI,gBAAgB,CAAC,IAAI,GAAG,CAAC;YAC5D,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,GAAG,KAAK,CAAC,IAAI,CAAC,gBAAgB,CAAC,MAAM,EAAE,CAAC,CAAC,GAAG,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,KAAK,CAAC,CAAC;YACxE,CAAC,CAAC,CAAC,CAAC;QAEN,yBAAyB;QACzB,yDAAyD;QACzD,MAAM,QAAQ,GAAG,UAAU,CAAC,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,cAAc,GAAG,QAAQ,CAAC;QAC3E,MAAM,OAAO,GAAG,QAAQ,GAAG,QAAQ,CAAC;QACpC,MAAM,IAAI,GAAG,QAAQ,GAAG,IAAI,CAAC,MAAM,CAAC,YAAY,GAAG,OAAO,CAAC;QAE3D,iBAAiB;QACjB,IAAI,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,IAAI,CAAC,CAAC;QAE1C,6CAA6C;QAC7C,IAAI,IAAI,CAAC,YAAY,EAAE,CAAC;YACtB,IAAI,CAAC,YAAY,CAAC,GAAG,CAAC,UAAU,EAAE,IAAI,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,6BAA6B;QACrF,CAAC;QAED,IAAI,CAAC,SAAS,EAAE,CAAC;IACnB,CAAC;IAED;;OAEG;IACO,yBAAyB;QACjC,OAAO,IAAI,CAAC,aAAa,CAAC,eAAe,CAAC;IAC5C,CAAC;IAED;;OAEG;IACH,gBAAgB;QACd,OAAO,YAAY,CAAC;IACtB,CAAC;IAED;;OAEG;IACH,gBAAgB;QACd,OAAO,YAAY,CAAC;IACtB,CAAC;IAED;;OAEG;IACH,qBAAqB;QAKnB,OAAO;YACL,SAAS,EAAE,IAAI,CAAC,gBAAgB,EAAE;YAClC,IAAI,EAAE,IAAI,CAAC,gBAAgB,EAAE;YAC7B,KAAK,EAAE,IAAI,CAAC,aAAa,EAAE;SAC5B,CAAC;IACJ,CAAC;CACF;AArFD,8BAqFC"}

package/dist/learning/algorithms/SARSALearner.d.ts

@@ -0,0 +1,107 @@
+/**
+ * SARSALearner - On-policy TD(0) Reinforcement Learning
+ *
+ * Implements SARSA (State-Action-Reward-State-Action) algorithm.
+ * Key differences from Q-Learning:
+ * - On-policy: learns Q-values for the policy being followed (epsilon-greedy)
+ * - Uses actual next action taken, not the max Q-value
+ * - Update rule: Q(s,a) ← Q(s,a) + α[r + γQ(s',a') - Q(s,a)]
+ * - More conservative than Q-Learning, safer for exploration
+ */
+import { AbstractRLLearner, RLConfig } from './AbstractRLLearner';
+import { TaskExperience, AgentAction, TaskState } from '../types';
+/**
+ * SARSA configuration (same as base RL config)
+ */
+export type SARSAConfig = RLConfig;
+/**
+ * SARSALearner - On-policy Temporal Difference Learning
+ *
+ * SARSA is an on-policy TD control algorithm that learns the Q-values
+ * for the policy being followed (typically epsilon-greedy).
+ *
+ * Key Characteristics:
+ * - Updates based on (State, Action, Reward, next State, next Action)
+ * - Learns Q-values for the actual policy (including exploration)
+ * - More conservative than Q-Learning
+ * - Better for tasks where exploration is risky
+ * - Converges to optimal policy under certain conditions
+ *
+ * Update Rule:
+ * Q(s,a) ← Q(s,a) + α[r + γQ(s',a') - Q(s,a)]
+ * where a' is the action actually taken in state s' (not necessarily greedy)
+ */
+export declare class SARSALearner extends AbstractRLLearner {
+    private readonly defaultConfig;
+    private lastStateAction?;
+    constructor(config?: Partial<RLConfig>);
+    /**
+     * Update Q-value using SARSA on-policy update rule
+     * Q(s,a) ← Q(s,a) + α[r + γQ(s',a') - Q(s,a)]
+     *
+     * @param experience The transition experience (s, a, r, s')
+     * @param nextAction The actual action taken in next state (SARSA requires this!)
+     *                   If not provided, selects action using current policy (epsilon-greedy)
+     */
+    update(experience: TaskExperience, nextAction?: AgentAction): void;
+    /**
+     * Calculate expected value for next state under current epsilon-greedy policy
+     * This is used when we don't have the actual next action (e.g., in batch updates)
+     *
+     * Expected SARSA: E[Q(s',a')] = ε/|A| * Σ Q(s',a) + (1-ε) * max Q(s',a)
+     */
+    private getExpectedValue;
+    /**
+     * Select next action and update with SARSA
+     * This is the typical SARSA flow: select action, observe reward, select next action, update
+     *
+     * @param currentState Current state
+     * @param currentAction Action taken in current state
+     * @param reward Reward received
+     * @param nextState Next state observed
+     * @param availableActions Actions available in next state
+     * @returns Next action selected (for continued learning)
+     */
+    selectAndUpdate(currentState: TaskState, currentAction: AgentAction, reward: number, nextState: TaskState, availableActions: AgentAction[]): AgentAction;
+    /**
+     * Learn from a complete episode trajectory
+     * Updates all state-action pairs in the trajectory using SARSA
+     *
+     * @param trajectory Array of (state, action, reward) tuples
+     */
+    learnFromEpisode(trajectory: Array<{
+        state: TaskState;
+        action: AgentAction;
+        reward: number;
+    }>): void;
+    /**
+     * Get the default exploration rate for this algorithm
+     */
+    protected getDefaultExplorationRate(): number;
+    /**
+     * Get algorithm name
+     */
+    getAlgorithmName(): string;
+    /**
+     * Get algorithm type (on-policy)
+     */
+    getAlgorithmType(): 'on-policy' | 'off-policy';
+    /**
+     * Get detailed statistics including SARSA-specific metrics
+     */
+    getDetailedStatistics(): {
+        algorithm: string;
+        type: 'on-policy' | 'off-policy';
+        stats: ReturnType<AbstractRLLearner['getStatistics']>;
+    };
+    /**
+     * Compare performance with expected convergence
+     * SARSA typically converges slower but more safely than Q-Learning
+     */
+    getConvergenceMetrics(): {
+        isConverging: boolean;
+        convergenceRate: number;
+        stability: number;
+    };
+}
+//# sourceMappingURL=SARSALearner.d.ts.map

package/dist/learning/algorithms/SARSALearner.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"SARSALearner.d.ts","sourceRoot":"","sources":["../../../src/learning/algorithms/SARSALearner.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AAEH,OAAO,EAAE,iBAAiB,EAAE,QAAQ,EAAE,MAAM,qBAAqB,CAAC;AAClE,OAAO,EAAE,cAAc,EAAE,WAAW,EAAE,SAAS,EAAE,MAAM,UAAU,CAAC;AAElE;;GAEG;AACH,MAAM,MAAM,WAAW,GAAG,QAAQ,CAAC;AAgBnC;;;;;;;;;;;;;;;;GAgBG;AACH,qBAAa,YAAa,SAAQ,iBAAiB;IACjD,OAAO,CAAC,QAAQ,CAAC,aAAa,CAAW;IACzC,OAAO,CAAC,eAAe,CAAC,CAAoC;gBAEhD,MAAM,GAAE,OAAO,CAAC,QAAQ,CAAM;IAO1C;;;;;;;OAOG;IACH,MAAM,CAAC,UAAU,EAAE,cAAc,EAAE,UAAU,CAAC,EAAE,WAAW,GAAG,IAAI;IA8ClE;;;;;OAKG;IACH,OAAO,CAAC,gBAAgB;IAwBxB;;;;;;;;;;OAUG;IACH,eAAe,CACb,YAAY,EAAE,SAAS,EACvB,aAAa,EAAE,WAAW,EAC1B,MAAM,EAAE,MAAM,EACd,SAAS,EAAE,SAAS,EACpB,gBAAgB,EAAE,WAAW,EAAE,GAC9B,WAAW;IAsBd;;;;;OAKG;IACH,gBAAgB,CACd,UAAU,EAAE,KAAK,CAAC;QAChB,KAAK,EAAE,SAAS,CAAC;QACjB,MAAM,EAAE,WAAW,CAAC;QACpB,MAAM,EAAE,MAAM,CAAC;KAChB,CAAC,GACD,IAAI;IA0CP;;OAEG;IACH,SAAS,CAAC,yBAAyB,IAAI,MAAM;IAI7C;;OAEG;IACH,gBAAgB,IAAI,MAAM;IAI1B;;OAEG;IACH,gBAAgB,IAAI,WAAW,GAAG,YAAY;IAI9C;;OAEG;IACH,qBAAqB,IAAI;QACvB,SAAS,EAAE,MAAM,CAAC;QAClB,IAAI,EAAE,WAAW,GAAG,YAAY,CAAC;QACjC,KAAK,EAAE,UAAU,CAAC,iBAAiB,CAAC,eAAe,CAAC,CAAC,CAAC;KACvD;IAQD;;;OAGG;IACH,qBAAqB,IAAI;QACvB,YAAY,EAAE,OAAO,CAAC;QACtB,eAAe,EAAE,MAAM,CAAC;QACxB,SAAS,EAAE,MAAM,CAAC;KACnB;CA+BF"}

package/dist/learning/algorithms/SARSALearner.js

@@ -0,0 +1,252 @@
+"use strict";
+/**
+ * SARSALearner - On-policy TD(0) Reinforcement Learning
+ *
+ * Implements SARSA (State-Action-Reward-State-Action) algorithm.
+ * Key differences from Q-Learning:
+ * - On-policy: learns Q-values for the policy being followed (epsilon-greedy)
+ * - Uses actual next action taken, not the max Q-value
+ * - Update rule: Q(s,a) ← Q(s,a) + α[r + γQ(s',a') - Q(s,a)]
+ * - More conservative than Q-Learning, safer for exploration
+ */
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.SARSALearner = void 0;
+const AbstractRLLearner_1 = require("./AbstractRLLearner");
+/**
+ * Default SARSA configuration
+ */
+const DEFAULT_SARSA_CONFIG = {
+    learningRate: 0.1,
+    discountFactor: 0.95,
+    explorationRate: 0.3,
+    explorationDecay: 0.995,
+    minExplorationRate: 0.01,
+    useExperienceReplay: true,
+    replayBufferSize: 10000,
+    batchSize: 32
+};
+/**
+ * SARSALearner - On-policy Temporal Difference Learning
+ *
+ * SARSA is an on-policy TD control algorithm that learns the Q-values
+ * for the policy being followed (typically epsilon-greedy).
+ *
+ * Key Characteristics:
+ * - Updates based on (State, Action, Reward, next State, next Action)
+ * - Learns Q-values for the actual policy (including exploration)
+ * - More conservative than Q-Learning
+ * - Better for tasks where exploration is risky
+ * - Converges to optimal policy under certain conditions
+ *
+ * Update Rule:
+ * Q(s,a) ← Q(s,a) + α[r + γQ(s',a') - Q(s,a)]
+ * where a' is the action actually taken in state s' (not necessarily greedy)
+ */
+class SARSALearner extends AbstractRLLearner_1.AbstractRLLearner {
+    constructor(config = {}) {
+        const fullConfig = { ...DEFAULT_SARSA_CONFIG, ...config };
+        super(fullConfig);
+        this.defaultConfig = fullConfig;
+        this.logger.info('SARSALearner initialized with on-policy TD(0)', { config: fullConfig });
+    }
+    /**
+     * Update Q-value using SARSA on-policy update rule
+     * Q(s,a) ← Q(s,a) + α[r + γQ(s',a') - Q(s,a)]
+     *
+     * @param experience The transition experience (s, a, r, s')
+     * @param nextAction The actual action taken in next state (SARSA requires this!)
+     *                   If not provided, selects action using current policy (epsilon-greedy)
+     */
+    update(experience, nextAction) {
+        const stateKey = this.encodeState(experience.state);
+        const actionKey = this.encodeAction(experience.action);
+        const nextStateKey = this.encodeState(experience.nextState);
+        // Get current Q-value Q(s,a)
+        const stateActions = this.qTable.get(stateKey);
+        const currentQ = stateActions?.get(actionKey)?.value ?? 0;
+        // SARSA: Get Q-value for next action that will actually be taken
+        // This is the key difference from Q-Learning (which uses max Q-value)
+        let nextQ = 0;
+        if (nextAction) {
+            // Use provided next action (typical in online learning)
+            const nextActionKey = this.encodeAction(nextAction);
+            const nextStateActions = this.qTable.get(nextStateKey);
+            nextQ = nextStateActions?.get(nextActionKey)?.value ?? 0;
+        }
+        else {
+            // If no next action provided, we need to select one using epsilon-greedy
+            // This happens in batch updates from experience replay
+            // We approximate by using a greedy action (conservative estimate)
+            const nextStateActions = this.qTable.get(nextStateKey);
+            if (nextStateActions && nextStateActions.size > 0) {
+                // Use expected SARSA approximation: average over all actions weighted by policy
+                nextQ = this.getExpectedValue(experience.nextState, nextStateActions);
+            }
+        }
+        // SARSA update rule
+        // Q(s,a) = Q(s,a) + α * [r + γ * Q(s',a') - Q(s,a)]
+        const tdTarget = experience.reward + this.config.discountFactor * nextQ;
+        const tdError = tdTarget - currentQ;
+        const newQ = currentQ + this.config.learningRate * tdError;
+        // Update Q-value
+        this.setQValue(stateKey, actionKey, newQ);
+        // Add to experience replay buffer if enabled
+        if (this.replayBuffer) {
+            this.replayBuffer.add(experience, Math.abs(tdError)); // Priority based on TD error
+        }
+        this.stepCount++;
+    }
+    /**
+     * Calculate expected value for next state under current epsilon-greedy policy
+     * This is used when we don't have the actual next action (e.g., in batch updates)
+     *
+     * Expected SARSA: E[Q(s',a')] = ε/|A| * Σ Q(s',a) + (1-ε) * max Q(s',a)
+     */
+    getExpectedValue(nextState, nextStateActions) {
+        if (nextStateActions.size === 0) {
+            return 0;
+        }
+        const epsilon = this.config.explorationRate;
+        const numActions = nextStateActions.size;
+        // Calculate average Q-value (for random exploration)
+        let sumQ = 0;
+        let maxQ = -Infinity;
+        for (const qValue of nextStateActions.values()) {
+            sumQ += qValue.value;
+            maxQ = Math.max(maxQ, qValue.value);
+        }
+        const avgQ = sumQ / numActions;
+        // Expected value under epsilon-greedy policy
+        // ε * (average of all actions) + (1-ε) * (max action)
+        return epsilon * avgQ + (1 - epsilon) * maxQ;
+    }
+    /**
+     * Select next action and update with SARSA
+     * This is the typical SARSA flow: select action, observe reward, select next action, update
+     *
+     * @param currentState Current state
+     * @param currentAction Action taken in current state
+     * @param reward Reward received
+     * @param nextState Next state observed
+     * @param availableActions Actions available in next state
+     * @returns Next action selected (for continued learning)
+     */
+    selectAndUpdate(currentState, currentAction, reward, nextState, availableActions) {
+        // Select next action using epsilon-greedy policy
+        const nextAction = this.selectAction(nextState, availableActions);
+        // Create experience
+        const experience = {
+            taskId: `sarsa-${Date.now()}`,
+            taskType: 'online-learning',
+            state: currentState,
+            action: currentAction,
+            reward,
+            nextState,
+            timestamp: new Date(),
+            agentId: 'sarsa-learner'
+        };
+        // Update Q-value using SARSA rule with actual next action
+        this.update(experience, nextAction);
+        return nextAction;
+    }
+    /**
+     * Learn from a complete episode trajectory
+     * Updates all state-action pairs in the trajectory using SARSA
+     *
+     * @param trajectory Array of (state, action, reward) tuples
+     */
+    learnFromEpisode(trajectory) {
+        // SARSA updates each transition with the next action in the trajectory
+        for (let i = 0; i < trajectory.length - 1; i++) {
+            const current = trajectory[i];
+            const next = trajectory[i + 1];
+            const experience = {
+                taskId: `episode-${Date.now()}-${i}`,
+                taskType: 'episode-learning',
+                state: current.state,
+                action: current.action,
+                reward: current.reward,
+                nextState: next.state,
+                timestamp: new Date(),
+                agentId: 'sarsa-learner'
+            };
+            // Update with the actual next action from trajectory
+            this.update(experience, next.action);
+        }
+        // Handle terminal state (last transition)
+        if (trajectory.length > 0) {
+            const last = trajectory[trajectory.length - 1];
+            const terminalExperience = {
+                taskId: `episode-${Date.now()}-terminal`,
+                taskType: 'episode-learning',
+                state: last.state,
+                action: last.action,
+                reward: last.reward,
+                nextState: last.state, // Terminal state transitions to itself
+                timestamp: new Date(),
+                agentId: 'sarsa-learner'
+            };
+            // Terminal state has no next action, Q(terminal, any) = 0
+            this.update(terminalExperience);
+        }
+        this.endEpisode();
+    }
+    /**
+     * Get the default exploration rate for this algorithm
+     */
+    getDefaultExplorationRate() {
+        return this.defaultConfig.explorationRate;
+    }
+    /**
+     * Get algorithm name
+     */
+    getAlgorithmName() {
+        return 'SARSA';
+    }
+    /**
+     * Get algorithm type (on-policy)
+     */
+    getAlgorithmType() {
+        return 'on-policy';
+    }
+    /**
+     * Get detailed statistics including SARSA-specific metrics
+     */
+    getDetailedStatistics() {
+        return {
+            algorithm: this.getAlgorithmName(),
+            type: this.getAlgorithmType(),
+            stats: this.getStatistics()
+        };
+    }
+    /**
+     * Compare performance with expected convergence
+     * SARSA typically converges slower but more safely than Q-Learning
+     */
+    getConvergenceMetrics() {
+        const stats = this.getStatistics();
+        // Check if Q-values are stabilizing
+        const avgQValue = stats.avgQValue;
+        const qValueRange = stats.maxQValue - stats.minQValue;
+        // Convergence indicators:
+        // 1. Low exploration rate (mostly exploiting)
+        // 2. Reasonable Q-value range (not diverging)
+        // 3. Sufficient episodes for learning
+        const isConverging = stats.explorationRate < 0.1 && // Low exploration
+            qValueRange < 10 && // Bounded Q-values
+            stats.episodes > 20; // Sufficient training
+        const convergenceRate = stats.episodes > 0
+            ? Math.min(1.0, stats.episodes / 100)
+            : 0;
+        const stability = qValueRange > 0
+            ? 1.0 - Math.min(1.0, qValueRange / 20)
+            : 0.5;
+        return {
+            isConverging,
+            convergenceRate,
+            stability
+        };
+    }
+}
+exports.SARSALearner = SARSALearner;
+//# sourceMappingURL=SARSALearner.js.map

package/dist/learning/algorithms/SARSALearner.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"SARSALearner.js","sourceRoot":"","sources":["../../../src/learning/algorithms/SARSALearner.ts"],"names":[],"mappings":";AAAA;;;;;;;;;GASG;;;AAEH,2DAAkE;AAQlE;;GAEG;AACH,MAAM,oBAAoB,GAAa;IACrC,YAAY,EAAE,GAAG;IACjB,cAAc,EAAE,IAAI;IACpB,eAAe,EAAE,GAAG;IACpB,gBAAgB,EAAE,KAAK;IACvB,kBAAkB,EAAE,IAAI;IACxB,mBAAmB,EAAE,IAAI;IACzB,gBAAgB,EAAE,KAAK;IACvB,SAAS,EAAE,EAAE;CACd,CAAC;AAEF;;;;;;;;;;;;;;;;GAgBG;AACH,MAAa,YAAa,SAAQ,qCAAiB;IAIjD,YAAY,SAA4B,EAAE;QACxC,MAAM,UAAU,GAAG,EAAE,GAAG,oBAAoB,EAAE,GAAG,MAAM,EAAE,CAAC;QAC1D,KAAK,CAAC,UAAU,CAAC,CAAC;QAClB,IAAI,CAAC,aAAa,GAAG,UAAU,CAAC;QAChC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,+CAA+C,EAAE,EAAE,MAAM,EAAE,UAAU,EAAE,CAAC,CAAC;IAC5F,CAAC;IAED;;;;;;;OAOG;IACH,MAAM,CAAC,UAA0B,EAAE,UAAwB;QACzD,MAAM,QAAQ,GAAG,IAAI,CAAC,WAAW,CAAC,UAAU,CAAC,KAAK,CAAC,CAAC;QACpD,MAAM,SAAS,GAAG,IAAI,CAAC,YAAY,CAAC,UAAU,CAAC,MAAM,CAAC,CAAC;QACvD,MAAM,YAAY,GAAG,IAAI,CAAC,WAAW,CAAC,UAAU,CAAC,SAAS,CAAC,CAAC;QAE5D,6BAA6B;QAC7B,MAAM,YAAY,GAAG,IAAI,CAAC,MAAM,CAAC,GAAG,CAAC,QAAQ,CAAC,CAAC;QAC/C,MAAM,QAAQ,GAAG,YAAY,EAAE,GAAG,CAAC,SAAS,CAAC,EAAE,KAAK,IAAI,CAAC,CAAC;QAE1D,iEAAiE;QACjE,sEAAsE;QACtE,IAAI,KAAK,GAAG,CAAC,CAAC;QAEd,IAAI,UAAU,EAAE,CAAC;YACf,wDAAwD;YACxD,MAAM,aAAa,GAAG,IAAI,CAAC,YAAY,CAAC,UAAU,CAAC,CAAC;YACpD,MAAM,gBAAgB,GAAG,IAAI,CAAC,MAAM,CAAC,GAAG,CAAC,YAAY,CAAC,CAAC;YACvD,KAAK,GAAG,gBAAgB,EAAE,GAAG,CAAC,aAAa,CAAC,EAAE,KAAK,IAAI,CAAC,CAAC;QAC3D,CAAC;aAAM,CAAC;YACN,yEAAyE;YACzE,uDAAuD;YACvD,kEAAkE;YAClE,MAAM,gBAAgB,GAAG,IAAI,CAAC,MAAM,CAAC,GAAG,CAAC,YAAY,CAAC,CAAC;YACvD,IAAI,gBAAgB,IAAI,gBAAgB,CAAC,IAAI,GAAG,CAAC,EAAE,CAAC;gBAClD,gFAAgF;gBAChF,KAAK,GAAG,IAAI,CAAC,gBAAgB,CAAC,UAAU,CAAC,SAAS,EAAE,gBAAgB,CAAC,CAAC;YACxE,CAAC;QACH,CAAC;QAED,oBAAoB;QACpB,oDAAoD;QACpD,MAAM,QAAQ,GAAG,UAAU,CAAC,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,cAAc,GAAG,KAAK,CAAC;QACxE,MAAM,OAAO,GAAG,QAAQ,GAAG,QAAQ,CAAC;QACpC,MAAM,IAAI,GAAG,QAAQ,GAAG,IAAI,CAAC,MAAM,CAAC,YAAY,GAAG,OAAO,CAAC;QAE3D,iBAAiB;QACjB,IAAI,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,IAAI,CAAC,CAAC;QAE1C,6CAA6C;QAC7C,IAAI,IAAI,CAAC,YAAY,EAAE,CAAC;YACtB,IAAI,CAAC,YAAY,CAAC,GAAG,CAAC,UAAU,EAAE,IAAI,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,6BAA6B;QACrF,CAAC;QAED,IAAI,CAAC,SAAS,EAAE,CAAC;IACnB,CAAC;IAED;;;;;OAKG;IACK,gBAAgB,CAAC,SAAoB,EAAE,gBAAkC;QAC/E,IAAI,gBAAgB,CAAC,IAAI,KAAK,CAAC,EAAE,CAAC;YAChC,OAAO,CAAC,CAAC;QACX,CAAC;QAED,MAAM,OAAO,GAAG,IAAI,CAAC,MAAM,CAAC,eAAe,CAAC;QAC5C,MAAM,UAAU,GAAG,gBAAgB,CAAC,IAAI,CAAC;QAEzC,qDAAqD;QACrD,IAAI,IAAI,GAAG,CAAC,CAAC;QACb,IAAI,IAAI,GAAG,CAAC,QAAQ,CAAC;QAErB,KAAK,MAAM,MAAM,IAAI,gBAAgB,CAAC,MAAM,EAAE,EAAE,CAAC;YAC/C,IAAI,IAAI,MAAM,CAAC,KAAK,CAAC;YACrB,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE,MAAM,CAAC,KAAK,CAAC,CAAC;QACtC,CAAC;QAED,MAAM,IAAI,GAAG,IAAI,GAAG,UAAU,CAAC;QAE/B,6CAA6C;QAC7C,sDAAsD;QACtD,OAAO,OAAO,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,OAAO,CAAC,GAAG,IAAI,CAAC;IAC/C,CAAC;IAED;;;;;;;;;;OAUG;IACH,eAAe,CACb,YAAuB,EACvB,aAA0B,EAC1B,MAAc,EACd,SAAoB,EACpB,gBAA+B;QAE/B,iDAAiD;QACjD,MAAM,UAAU,GAAG,IAAI,CAAC,YAAY,CAAC,SAAS,EAAE,gBAAgB,CAAC,CAAC;QAElE,oBAAoB;QACpB,MAAM,UAAU,GAAmB;YACjC,MAAM,EAAE,SAAS,IAAI,CAAC,GAAG,EAAE,EAAE;YAC7B,QAAQ,EAAE,iBAAiB;YAC3B,KAAK,EAAE,YAAY;YACnB,MAAM,EAAE,aAAa;YACrB,MAAM;YACN,SAAS;YACT,SAAS,EAAE,IAAI,IAAI,EAAE;YACrB,OAAO,EAAE,eAAe;SACzB,CAAC;QAEF,0DAA0D;QAC1D,IAAI,CAAC,MAAM,CAAC,UAAU,EAAE,UAAU,CAAC,CAAC;QAEpC,OAAO,UAAU,CAAC;IACpB,CAAC;IAED;;;;;OAKG;IACH,gBAAgB,CACd,UAIE;QAEF,uEAAuE;QACvE,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,UAAU,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC;YAC/C,MAAM,OAAO,GAAG,UAAU,CAAC,CAAC,CAAC,CAAC;YAC9B,MAAM,IAAI,GAAG,UAAU,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC;YAE/B,MAAM,UAAU,GAAmB;gBACjC,MAAM,EAAE,WAAW,IAAI,CAAC,GAAG,EAAE,IAAI,CAAC,EAAE;gBACpC,QAAQ,EAAE,kBAAkB;gBAC5B,KAAK,EAAE,OAAO,CAAC,KAAK;gBACpB,MAAM,EAAE,OAAO,CAAC,MAAM;gBACtB,MAAM,EAAE,OAAO,CAAC,MAAM;gBACtB,SAAS,EAAE,IAAI,CAAC,KAAK;gBACrB,SAAS,EAAE,IAAI,IAAI,EAAE;gBACrB,OAAO,EAAE,eAAe;aACzB,CAAC;YAEF,qDAAqD;YACrD,IAAI,CAAC,MAAM,CAAC,UAAU,EAAE,IAAI,CAAC,MAAM,CAAC,CAAC;QACvC,CAAC;QAED,0CAA0C;QAC1C,IAAI,UAAU,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YAC1B,MAAM,IAAI,GAAG,UAAU,CAAC,UAAU,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;YAC/C,MAAM,kBAAkB,GAAmB;gBACzC,MAAM,EAAE,WAAW,IAAI,CAAC,GAAG,EAAE,WAAW;gBACxC,QAAQ,EAAE,kBAAkB;gBAC5B,KAAK,EAAE,IAAI,CAAC,KAAK;gBACjB,MAAM,EAAE,IAAI,CAAC,MAAM;gBACnB,MAAM,EAAE,IAAI,CAAC,MAAM;gBACnB,SAAS,EAAE,IAAI,CAAC,KAAK,EAAE,uCAAuC;gBAC9D,SAAS,EAAE,IAAI,IAAI,EAAE;gBACrB,OAAO,EAAE,eAAe;aACzB,CAAC;YAEF,0DAA0D;YAC1D,IAAI,CAAC,MAAM,CAAC,kBAAkB,CAAC,CAAC;QAClC,CAAC;QAED,IAAI,CAAC,UAAU,EAAE,CAAC;IACpB,CAAC;IAED;;OAEG;IACO,yBAAyB;QACjC,OAAO,IAAI,CAAC,aAAa,CAAC,eAAe,CAAC;IAC5C,CAAC;IAED;;OAEG;IACH,gBAAgB;QACd,OAAO,OAAO,CAAC;IACjB,CAAC;IAED;;OAEG;IACH,gBAAgB;QACd,OAAO,WAAW,CAAC;IACrB,CAAC;IAED;;OAEG;IACH,qBAAqB;QAKnB,OAAO;YACL,SAAS,EAAE,IAAI,CAAC,gBAAgB,EAAE;YAClC,IAAI,EAAE,IAAI,CAAC,gBAAgB,EAAE;YAC7B,KAAK,EAAE,IAAI,CAAC,aAAa,EAAE;SAC5B,CAAC;IACJ,CAAC;IAED;;;OAGG;IACH,qBAAqB;QAKnB,MAAM,KAAK,GAAG,IAAI,CAAC,aAAa,EAAE,CAAC;QAEnC,oCAAoC;QACpC,MAAM,SAAS,GAAG,KAAK,CAAC,SAAS,CAAC;QAClC,MAAM,WAAW,GAAG,KAAK,CAAC,SAAS,GAAG,KAAK,CAAC,SAAS,CAAC;QAEtD,0BAA0B;QAC1B,8CAA8C;QAC9C,8CAA8C;QAC9C,sCAAsC;QAEtC,MAAM,YAAY,GAChB,KAAK,CAAC,eAAe,GAAG,GAAG,IAAI,kBAAkB;YACjD,WAAW,GAAG,EAAE,IAAI,mBAAmB;YACvC,KAAK,CAAC,QAAQ,GAAG,EAAE,CAAC,CAAC,sBAAsB;QAE7C,MAAM,eAAe,GAAG,KAAK,CAAC,QAAQ,GAAG,CAAC;YACxC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,GAAG,EAAE,KAAK,CAAC,QAAQ,GAAG,GAAG,CAAC;YACrC,CAAC,CAAC,CAAC,CAAC;QAEN,MAAM,SAAS,GAAG,WAAW,GAAG,CAAC;YAC/B,CAAC,CAAC,GAAG,GAAG,IAAI,CAAC,GAAG,CAAC,GAAG,EAAE,WAAW,GAAG,EAAE,CAAC;YACvC,CAAC,CAAC,GAAG,CAAC;QAER,OAAO;YACL,YAAY;YACZ,eAAe;YACf,SAAS;SACV,CAAC;IACJ,CAAC;CACF;AAvQD,oCAuQC"}

package/dist/learning/algorithms/index.d.ts

@@ -0,0 +1,32 @@
+/**
+ * Learning Algorithms - Reinforcement Learning Implementations
+ *
+ * This module provides various RL algorithms for agent learning:
+ * - AbstractRLLearner: Base class for all RL algorithms
+ * - QLearning: Off-policy TD(0) algorithm
+ * - SARSALearner: On-policy TD(0) algorithm
+ * - ActorCriticLearner: Advantage Actor-Critic (A2C) algorithm
+ * - PPOLearner: Proximal Policy Optimization (PPO-Clip) algorithm
+ * - MAMLMetaLearner: Model-Agnostic Meta-Learning for fast adaptation
+ */
+import { AbstractRLLearner, RLConfig, QValue } from './AbstractRLLearner';
+import { QLearning, QLearningConfig } from '../QLearning';
+import { SARSALearner, SARSAConfig } from './SARSALearner';
+import { ActorCriticLearner, ActorCriticConfig, createDefaultActorCriticConfig } from './ActorCriticLearner';
+import { PPOLearner, PPOConfig, createDefaultPPOConfig } from './PPOLearner';
+import { MAMLMetaLearner, MAMLConfig, createDefaultMAMLConfig } from './MAMLMetaLearner';
+export { AbstractRLLearner, RLConfig, QValue };
+export { QLearning, QLearningConfig };
+export { SARSALearner, SARSAConfig };
+export { ActorCriticLearner, ActorCriticConfig, createDefaultActorCriticConfig };
+export { PPOLearner, PPOConfig, createDefaultPPOConfig };
+export { MAMLMetaLearner, MAMLConfig, createDefaultMAMLConfig };
+/**
+ * Supported RL algorithm types
+ */
+export type RLAlgorithmType = 'q-learning' | 'sarsa' | 'actor-critic' | 'ppo' | 'maml' | 'legacy';
+/**
+ * Factory function to create RL algorithm instances
+ */
+export declare function createRLAlgorithm(type: RLAlgorithmType, config?: any): AbstractRLLearner;
+//# sourceMappingURL=index.d.ts.map

package/dist/learning/algorithms/index.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/learning/algorithms/index.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;GAUG;AAEH,OAAO,EAAE,iBAAiB,EAAE,QAAQ,EAAE,MAAM,EAAE,MAAM,qBAAqB,CAAC;AAC1E,OAAO,EAAE,SAAS,EAAE,eAAe,EAAE,MAAM,cAAc,CAAC;AAC1D,OAAO,EAAE,YAAY,EAAE,WAAW,EAAE,MAAM,gBAAgB,CAAC;AAC3D,OAAO,EAAE,kBAAkB,EAAE,iBAAiB,EAAE,8BAA8B,EAAE,MAAM,sBAAsB,CAAC;AAC7G,OAAO,EAAE,UAAU,EAAE,SAAS,EAAE,sBAAsB,EAAE,MAAM,cAAc,CAAC;AAC7E,OAAO,EAAE,eAAe,EAAE,UAAU,EAAE,uBAAuB,EAAE,MAAM,mBAAmB,CAAC;AAEzF,OAAO,EAAE,iBAAiB,EAAE,QAAQ,EAAE,MAAM,EAAE,CAAC;AAC/C,OAAO,EAAE,SAAS,EAAE,eAAe,EAAE,CAAC;AACtC,OAAO,EAAE,YAAY,EAAE,WAAW,EAAE,CAAC;AACrC,OAAO,EAAE,kBAAkB,EAAE,iBAAiB,EAAE,8BAA8B,EAAE,CAAC;AACjF,OAAO,EAAE,UAAU,EAAE,SAAS,EAAE,sBAAsB,EAAE,CAAC;AACzD,OAAO,EAAE,eAAe,EAAE,UAAU,EAAE,uBAAuB,EAAE,CAAC;AAEhE;;GAEG;AACH,MAAM,MAAM,eAAe,GAAG,YAAY,GAAG,OAAO,GAAG,cAAc,GAAG,KAAK,GAAG,MAAM,GAAG,QAAQ,CAAC;AAElG;;GAEG;AACH,wBAAgB,iBAAiB,CAC/B,IAAI,EAAE,eAAe,EACrB,MAAM,CAAC,EAAE,GAAG,GACX,iBAAiB,CAenB"}