From 6e2b99438bae0267355cb7740ef1d437e606511a Mon Sep 17 00:00:00 2001
From: 997146918 <997146918@qq.com>
Date: Sat, 23 Aug 2025 17:10:23 +0800
Subject: [PATCH] =?UTF-8?q?=E6=B7=BB=E5=8A=A0=E5=AF=B9=E8=AF=9D=E5=86=85?=
 =?UTF-8?q?=E5=AE=B9=20ai=E6=89=93=E5=88=86=E5=8A=9F=E8=83=BD?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 AITrain/conversation_data/conversations.db    | Bin 102400 -> 36864 bytes
 .../conversation_data/demo_conversations.db   | Bin 118784 -> 0 bytes
 AITrain/dual_ai_dialogue_system.py            | 121 ++-
 AITrain/main_controller.py                    | 980 +++++++++++++++++-
 4 files changed, 1083 insertions(+), 18 deletions(-)
 delete mode 100644 AITrain/conversation_data/demo_conversations.db

diff --git a/AITrain/conversation_data/conversations.db b/AITrain/conversation_data/conversations.db
index 17da33bf5a56f56f67c6d2ca1a55b386f86653ed..5cd331772eecc6c470bd0435faeeaef9ad132093 100644
GIT binary patch
literal 36864
zcmeHQ?QdJxd6!~4j<a=c`z0UCX)t7?i9v~yNO}WSFH?e`N!qz?v1};}N}^Vyww&rE
z$*^u9MUf&YiIi+nZ=_x=iLxw_)QcsGuRuSy|3q@nx%UDEynPu4?8|=7JtR%N0>Rv3
zOAf)v<h|#f`<&l7_nhbVd!EDm<nP|U+v_xSKIp#h=rvhtzFkvWTk`{xsivmpBL2IC
z|H`jd@Zo&<3;wOG`h42Qi#6AS-~HC*|Ef9vC;K&j_HUQ-e|iZY<VzwT5s(N-1SA3y
z0f~S_Kq4R!_$mne`24GHeE<8kAN4xg?m2sYbq_xLogU|}`kY<uN55Wp`&R4C+pVVC
zH~;4SR@0HPSGyedop0jGYXjZqF1~T?T5aHaMx7mZ9rqsmQ=jurZ(ny;&*5+957l`1
zoayT8rW0OwJ529=c)Ru8)?20@-+J$Zo3}nP{jl{T)6JjW{?U6MqPh=SKfL|siC6YG
zdwT9Z=(>X%ZnysZDTT<$J6(PE+nn9y`<znSCr-yNQRwul+8=cFI=gz$DA9ZOzO$#-
zasQJu${4kL-g~F7$9b0B-OhW?PaWv9J3Z|Wx}7HU)cYqJUP1XqLDS#1zH{@Z@833=
zubWT4$S8ZK!`bV&d#~pVYel(Er?aEY(f-Rb3cmBBTdnWC`=QZyS6}qZH5mR*>#f!g
z-){Yh2~Gdh+1=ylMX&Xks!JW~-)sGUReSM``uf@dt1;*fU3%!tD~E>Mp;M=fxEIzL
z?WrDiCxg}g3rDx3y%(mR0l;Suy6<&-de`~c8D%gwx_j@?C^@6hJx32l$0r>Qknht6
z<9&1;ia3lS$KkWT^6DEO)YsPB?douTUV5?}eZ3Ei?{^NNbf@9a*K?R`UZV-;TllZ`
z@?X^8EngA=iGV~vA|Mfv2uK7Z0ulj<fJ8tdAQAW`BjEmr+SjhQ``){Gv!=13v&Guh
zXlZV;I{vZl<36jiv+3i$W_yDJe;XU^ANMsm?f7gqw^?z-X|{aaXS1|5;B!M$8_sr`
zJMg))&06=S3D?_OES>mnZ=%wcW~!yVfoepFaxHeNv*qKy7ITZe?&r)qqz#4jZMV8I
zrzK~V^n@B2*5du9tG8~xd#%yj*!-5+_Ew|ic0==xM*9tm<$6<t-QHrYK4ZOMw%ssW
zu3Kz2tF_^E^la_rf3Lw?z9a$?0f~S_Kq4R!kO)WwBmxoviGV~vA|Mg?njmob+?7k!
zD#iu0>)hr4pw?Zw{NI=VQ+)iIG*&i4A|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT-(Un@
zqwN2Sw4MLeT9N<%`<lzY{{{mhn=TQM2uK7Z0ulj<fJ8tdAQ6xVNCYGT5`nJ~0^hAY
zcdfSaLlVXhdz?4;|Er(Z;4NPg0f~S_Kq4R!_=6+x=tAwcYpz^+H`s79@STen-96#A
ztJBid>@HiLVg4sdX685hEbY-kTiE=j<VILwzfPYjvXwzLGQ(2i`2IXM%>840XN?bp
znLnlE7g%gfOZl}#=E#OQKDnpP45;I4T5^O>9I#yYOILp#E~cZ=1oH;;@lo8B#Zo-*
zR4WdmF4SI!f0SHKIS8@j0iPdLGEu(T&x+4@ETGQ1^w~K!I?R$;mQGVeqdqPE7!@n&
z5mbTewc@lI9}&m-RFaLZvTR0uIs`MB*US7qx@uuueX_^qGwPyCIT%vX!RNX8y4PQa
zaaul$H;-(y!7Oga7l-xPB^1C_d}2_^?y=<q!XPljhX<K=M+^3|=z<<8qPhCyF56#G
z@-fXdtVUM&RDp%Z)ag96n@}o7wZ&m2n^p1=E#=|!D=aXAx^z!okGZr}x4M{6inH3_
zI@{mE{j|igI={jfeSB_D+uLEw^Xl3Ti)K(AOKmYv7<Ixpz@+y-223oxthwB{kZn%k
z6bvFbcGK#>9x&1XS;C`aH>mv^t1P_F!s|+Ih6NXN*Sxx2z&1y^cJ~R-c*tS{4(5*O
zi$ygs$oyH<Ci+c%QUuoAUqpkcw)IKw-{-L?8y(Y%F}6H_gL<^UQ{&HbGt{42lFE^l
z>GgWM{R+PDg&d!l<UyC_$=843TE!bZrzM`y=<@7n&-RtHtGZE2j;euS)%z=3ZZunO
zG&NptX=-R}G8?-@Ep2A(>9jW5{x>^DQA3lh{j2U7E#KGH=D^);ZEdv2w84sfsZCCs
z6Z=Z7X4q!y?4UiW7W!*#Zor;YYl{v0P_6c6l(E`7a4pKy4pwI;cB6{3wzdxJX0_Yw
zIBPMR8|io_K8yNo=5|=y($a$3nmV1R8GBgqxw8|yWxL)op`5v$&N<0~CTAlo!#O)n
z&_DFo+>DEC7OJM%jt#LbjW#mY+=<%?6PldVAiHs-+2TZX;;hZ)fU&UNIBUgm0baAE
zoi^Y$bmA=PCu5yfGzgn_aiqo4f%<WG7>mDVVG0g4cecY6TSp`Q;xe*PG`QJrL$h0~
zcG%d|*+NCz?0`&kc1t5&hKm8W-9aE8YC&5QT4{AU(6S~11tqQ24NYWYi_>Dee*Jpg
z&;MGVrq|35z9%N-f2nJK@C#=*{nB6E4U-MO{bs7G{B{}r=vU9(9{hq``#<BX-8kFj
zfUobk*M(ow!!fII>~5F%$#42;y$&a?>80{5qPA`)e!cBeaR@i~R}&rso_L0*8k>%J
zWLL>Q!t}!mezuawREe)@D5Ye4YTT`4GD=~UhnDa>Y4L|zU|UNRSYqZ2*Erso0=X;B
zeEa(B^p`F-b}qBzJPSs7B#enyNsp^T8<-a{-6-iSo-0uUO_Rp0mGmatir`*sFoPpU
zCU<&*_qQ-9h>4ax9M)26+#BL9H_NQjOo1t8)X%(oN*eahXr5s{okTI1ghyWu9I%3$
zCbIIqP@$UjVK(LAF!N16&-vME{}-;<SDI+QaLqo?1=!LY%Vh|n*gQ+j@qu~RqvVUC
zfjqj4DGv<?k}$wTv!iPmC5#z*PTNhAi4zByS=F@>EIPEZ?Cns2vaLLfbt?snVc+rD
z3`QvF5E~7t*-b1EN+xt1JIy_gNe)X7&7iRLAPlqkR0t~o^<{C6hgO8qtT3s2*05}G
zkC)KFno^il*WJwPg7r#qk-L*D=>zD>!Ms-RDESaBB(t#~U}eR^8%`?O71U32e1DF{
z5T?>Xfq=lWpFoq>jDg~1i31u3gX_Qq6$%hgn>L*%l#-j;nu~6chQX_}@(fQHM*7&n
z9*?b;Y#X0f7dN<nf^ajcqHgv_;V<AW+Tt39lom*0G~-&`HG`FerF<2LnAfB3E&#yl
z1wzd}!eByqF(RpJF?It7Gzzg+Y%O9~V{EbTw2~QOxdnWsuEoNyB@#5EMQ4=z5B-k;
zA^<J*0YG0I7oejtIxNmVqHz>lAeg-SG}tq%xL*lzAAv=U2eKFqBZJ~e6+?u&aCc1~
z43kZxX_ku!gAD$E!B>N~d`Sc(0ulj<z#kof@y}iX8?OD!;2R}2{OE_(Z1|NUg!ec5
z`ur+)Ke>vPl|4?Wxkuc)ghdo&<eI6D_pepF0pd6L5w^Lixx75!5~PRdPCXWiL(B&2
zS177Sy<n6)wn3hj;V0Cy*kkUW!sYd}gd|d0p130xurFQk0QXh+CKE~UF>;E61Gw}{
z*N{F)&I=ZN@*5Y3HNks?6QvbXU|)h6YD-Ub7aW;QTEH?HT1dTnxRYRDY(K6Q$B1wz
zhSdVO4@%yrWGBECS;2)ju|)96m<#JD>W9~d3bD2enu@iJmb&l=n}$8a8?VAYAr~S9
zAB!;4=gEB%o*bWCfG42AV}Wxr_Jj{_S6DW+`EkEG83C>pE+{-DxLc3{!0Rg<SKC}w
zmklZ$TT|mx_?aG-c?5LSIOz+zdr~VNpuu!s{}|D5!)Y#YN`IEb1cL;lusXlPXN%PA
zbP!$*aA)}lSq1SSy-s~J71ELs21XlN(c-K1Y<U7?nfgf0hFQpmRtvb`4Zt}P$nelG
z4==)1DN%R0D$^P6jpIISWS5+h)GTV{+w*XO$Yf&2ge?lF?q~XDKg+pTe251&d32lD
z34pEUzW8mSh8q^J4x7bhZkF6va^IK4eg9lc$%nbGV7Fg|`wF&vGWQj{xV@VD;sz&i
zUqKj1^-HJ0M8R?C0Wydwo-O_9RLLzjSSHaJFj0dFgC4*M$D?^<K8HMqXH?zV!DPWk
z7I|QbrE;JjY&njp8xNx9@?e(1XC9g6A%k^z_gHjJ-5lqWOAzSzFqpTG3dE91dWDFu
z%XqBAewN0;04ZtlDX?ngAOM3|{s}$P1|bGxBq?=sRa{8C+~CcqmL>f3b@wq8LRR#E
zp@2z}JypzJ%rqbrkX6rE69B?#b0RDOF;lYM3QMkV7R+QUyv;&0`rt^3G#|`Y^ajju
zC<$c5geW9bm`BXWyHe3%O_<rhU#n>}b(lfD)ZI8yRWQ+P<TUDxiJvCvbXH%QLN{X$
zq|1qt5mT&QHY<%3lRY72uxgYSyb~)7QLn)aacfw))sHgl(YAqpsJfH5YZkqZ3w2jk
zD~$29mrrB$gqA$W(~J6IMvZs@y-9Qh?9&E!(dQL{d}N5xZH(~($_x>dZwDaUqHlR<
zT`SCq4oB?*TG%I`1%rm%3j(VKo)E5)4K)%5J;hi+VTj%kbiwf}qD-Yjl<Z^li8@Tu
zf$E#*{kYu_9t*lEft7wu3_SzZb?uqaV}_GLcIKBrpN+O1qy3`Ogx63a|BGu74uk~2
zpSgjG(F~tZC|G!fZH|&?0+GTu2~C_VFjO34P`ott8_&fHQ6u5vyKX#ma6_ntnJAf*
zT@k~Zg`s<ILZ`qq0?VnL)sji214UEe$ti^>o^2(aV##5^BV;Rd3pp$ghxM6JHjzco
z(<#FNIn;hb6o6%rv@voZ{P18N=n@La1AG<y|JOAae*H&xdgToz0ulj<z&93wNbNad
z{qF{^lv#gkHS0g8o)rW~G=c{=5R+ufyO_<@SPtF;3r|6cCwVvsz4{SC;y6V?;nakP
zAdA>BEwaHi@;b40LwGJer3EGv)e>pMDozNmBVJmO6si;flElGe!NV6H^F_bDitxse
z2tyE2&E~<c)wQBnt_A<+o*>vhxj~~@=;v7dh-8*a^4S&mNewpoVcX0)%v93wO;d0j
zn0JGP)8am~cu^2%Q`lrIjxIGj00lser{GD~>)A1NAjP}`EH@9OK+s|^<3SOI63m#C
zlXyxnV=RwoOg&sOo({sn0i5+pW*H6zZU*TALW7z`=x7aradjezTf%a%U-SZ;3m_?i
z(kLwaE71|;Q1`=A3Q)iqE4J=N%wQ66eS&N^%9fwP$>&1>d=QZXbz>Y<UuXm{l7rqG
zaUF6qx3Wqu2S_MZVYo*K2LMVO7eWA30Cg!<5?A)T+FnpNz|`4Q!S_<z%OZ+Cy$@Fz
zUJ3~rD}%xe5#?vm6kGBL#K1C25{aJ>(=ty{El+1yDvUNCkpL_=8XK>J|2J70qy!)(
zfR`l!y#HIge)orz07}91mnQ*KSZ757s1P;GAJy}FNdl+}pwle3Lg7RaG_MYq6U{UN
z<s>dt2geJOl>Y=Fz{n~RWbSz`dV~uy|1{W$Ae%%1_qX(Z4?WLDMigiw%OFdpqz=%0
zVaOB2WFRFQRD$mXSU9L*mN6I_BCAJ8IP_wgcPdQIAX_IhaF*ReRGKDn@XEjl+2Mf_
z;1PKPu%i@^H3Wv^vnCpF;vzwOD;%sQ=jvuaL~>ClpBq3VR;(NlRKcu>u5I=6jYY@|
zWV#WzRYOTwg7OH}Q#6%I8aWU~$XhfXSqr$9$b4p%$lPg!TuCnjKXDeT$P3YCbbu;C
zv#Zh&Ad^>*4xAOE0E-kzo)!-fE{C(=s7y)Z34z{2KY%r2wx{-#0?j0|l=2|T_Z5T^
z6>7yP9$F*MAX5tM(}m$N(0iP%h%-FmB}D+IDp75L1OTY8R}r5FFqN3U$UZv~r57P;
z!Q?CKo$}XK!U%Yag%E2q^2`vdz-ouiVPRvC15Kowh-?LlAUuTK^jtvC$u00y6p{`-
zVo2)nKSj=6HJX&dam*(Y!j^E0%K08I+RS1Z<cbi)<@^om>N9vdOlQMhHcFlqQW|LF
zcz5*igqkU;yGs?`PPiIs#1W4L)?i?hp;aD9<zOWyLq>1XSRjcC$rYtUg-j{A!05W;
z3PR}_HUvV!|Kl|m;@?;|TsB-HAQ6xV{9i_3{AcGyg1~pmQo-eFso?w(U!a^IKq-r1
zigF`GfU*R}3kF#)<&R_Cz&a?h#Sy%LT8*qjkseN*6>ADeVKo<|)yl~7#)1lWkkSzs
zcGaa2)<r!MgU~L>A`w$cu%_@d1ObFDuxhH&w8;KOR35hhD+gVt2z`oDvnd20Nl{fu
z>ke!yxn&YGW+`Q^jv7&<Y(MqsH1rN|Pg?MvrATO4pVSJJ^CptO+2E`e_aRI!Qoea?
zRr8Q8UQYZDP^v#jHTletJ%%3g<T!t}Pxm+C>TEm8wjfsowYa;Qa%1VQP;QbgLI^>H
zd1{P}1+`=VbrO=vFxt&yer<;G_t|oc4nm&EM94Q>9zcu9$MgoY;6r_KUG<Q(0J;t=
z59<!pMLTPvA@FgKj!7sOT7#C~fO{?w5J?QBv<9L3%(+FXrIIZwd8Armn?nF8ALO~G
z*VR=wD^dUfVK{`mN~#1BD#VTiAy1IahR-G<YDN{*jqp1~hJnZp5Iu>Y9j-+;a`!0r
z#IZYp%_UF+x>p=WhQJYN0RQKq#_Q&$#^wf@9UwE~Ush)P|HtfrGo%4LLV}{5mJx6y
z9YDmV1+&ESErR(5FQ!BQk{*nBK238~6cdw5nG|kf_Y4m|tWXufBN3CX@MVycN;)*o
zVjBSJuO`>@^>E6Q2!Z4Han3GePQhGW#E#1W&g%4OL93}zMruQas#5whCR6M{$Ytnk
zIhUJQ8BJQJMEg(E8eWiGtd>wgP(uQ5WIM#FxfU98oRSHl0mrLXN<qv-$^cFChhz8<
z8t~R#53&6UXDj&(A_tw?Bh-iNBj^;=q|-EoBg;On6jWM&Ou6bWV(icz4#(KBgg|sR
z?p?&vZ}fZ|DRW|H1d>Q7L|qOpqnJo}IXf?)GSDg;E<}O=3;VHBu;ne<6|%b|wsje;
zIu_m+szGH~(o$r^&)M<_pQTiZie(~<O-GE4ct%TnB{kusi~zzVg(R9O1l|#MfSN+B
zGL(fEIS)|NMV^Dlt7g~v&JyGR6vN<GvL#K0oDM@`V7?L3+VV)lqZCG3nZr|5xV+Qj
jWk?2wECt~mNjtcQ*o-0!!gxDzn}-;p#|Q0c`ojMJ4a=Rk

literal 102400
zcmeHwX>eTEl^(!Vq)0-Wwb<h^P1%-ci|w~BkUbvRra1B<$r3fTIrK2SdK7E%h@@<1
zn{EJ&jo2s=1c-$oNPysqMI^B=Rj#Q?&5ubcOC_nKQkk;)zV42v%9&JUYEpJp@|}AD
zuoreyrXi`@zFkzS@!sv%@7{OLJ^OcF{MI*KJ`&y;dHs!7{YQ4{757wBR#yDs&Ycw%
z6^HTPBlvISwGJOPto(uhS7v-J^zm?oulc_1RsUGAVcQogcKoZVg>CKlAYXC?<O;|Y
zkSicpK(2sX0l5Nl1>_2p-wHgxVbepO`&{LBkN5+xgb)AuE3hOE9S;Bb(eP`*+&?#d
z<%PXZ?ccj||5IQ7#@?MdV;_Ib|7!RNd~<j0E9*Buw0n1D!~Oi4p_l!yy#D>8;X_A`
zzVX`O?7uc-T`~LI&d0azyv6I~(9UO`-M@F=-WPU0|H3oheCmbo?)>`R@9uo+rTx!6
z^DHj+&Ard=f8v(kJRClJ_~qAMJA^Cj-}{{cgXks?y>|4~K=_T7-zjjlABO!uz@WmH
z3cmi@k??Crii|k&@~h#)NBpn;u*evGl^-8Dbo6kz*nPhdekJ^-AJ6R2;o$3Ugm>ae
zedE?Q{`7M%?0sh6vwYPazquy6@s+3dzOeV%uk3wsCw}zJ@EeEyN3dRpcV>>tTiM+;
zU#Q&t&|{BP*4g}VXMHv6k9AoOJL^<|XMOYT_yT1<@LSI<_-Fn%{J|r*^&%_!qu1Yf
zCG_UY;U5(lgQxPwkwf%+iVS+ie;Ci@hao?V@4{s8YWRpBL;QS5{_?r&Ha+ys$0{pc
zek~OK@iI^Oj~;oQ|9vQHl@6&{f3C-7a~Ewm8}VOd)i3Gez2blLQNH8~$Q6()AXh-H
zfLsB&0&)f93dj|ZD<D@uu7F&D-<}n?=f16&%tv?ZtJv|kJN{YKr*^zt)x6_s)%`mb
zcYLSnN|m`|U)8!DeO3Ra>VNF`w>yqi{n?JcuUe@3Usb(T->&*e)ziN{>mX+$S3s_S
zTmiWPas}iH$Q6()AXh-HfLsB&0`GAJ?pgooO2-X2M!?bRh!y%M9WY#7i<lud6!1Hb
z&;iv@EKRqph~YOM<_8o@Q6rA-7^?aZ9q{Xcu;0|emL3W{NC#X!q(oFD64d?9N9ceS
z3OiQVjcA5H^Z*?&%%Bx?9mnsF*!T0VglvB#5HMBOjoe2E6x&g>a5!w5x_U1i@LOs`
ziG=*NY3cXyuV{*+Mcj~O1hgtTpodf|7;%F^zZTp<2XG6T6R|^@soUG>fE_Xd{$K!C
zw*1@ZKqR7v!?vlIj^f-+2f|t~XlRaK3253@I^Z|Ls-qc>9=7~j_=uq4I5zG?akabX
zfNGnn<+^^?aFxw;AQaXkA;*Yl{-Cvq4wyJ&8)`7**CHF~fFH|_`_LRm(KpZm*U%lm
z9t!A&Vz1{1LaL$bnr2zHv#x6Wr?>b+hCgiUR?v*76#u`s>M}j%N2-2V^$%5lU-frY
z$*P}K{r9T>QuQBT`Tup*Usf$u%~sv08W+RfBSpwzas}iH$Q6()AXh-HfLsB&0&)f9
z3dj|ZD<D@uuD~6xz{4-%iw`}Ew+ElW+ef~Nw+Ft2xBIK{cAts2Dg|#lK8LsMpTXO<
zU3k0u_wlxsYWcT3fVaD<@V1#E|C=^$$9rYf-&NqBe90A%D<D@uu7F$txdL(p<O;|Y
zkSicpK(2sX0l5Od1uJl0<@()~*+3cJCi{P5<@YPL@7!u_4sG}WKFZhHuRwj}#=9#%
zUAZsz;!{l@+nkkvk(|5EhI&}z#l&pSPp8;)ZK`d8#g8TCTiN22*E;Go9#72Hr$)~G
zV!p-ei>8*2dworb>1gW4lsDPL+M9nC?M`1k;q{HN>BgT$Po<+L*kWC}|1@hDOhx0#
zo;o((?hQ63<|c84#KIZYaV0f&J+;)pI=X%_KcFd^xkqvKsOEmvt~Q-&-Siozs@mG0
z>^gYV)&t7Hqo%Dn_#BQH`0T2Fyla|!@Td`X@mWy<Hjac9{oqkY52*O88UdV*C?R}~
z@Uw;!Jb2W#^azf)294KEx=v7~Yw|JTI<6hI4j#1>3)dIF>mYv!R}KVdylUfDjIa~N
z7+b-;Igt?kA~JSxho*|_+m?fC8sQ+$8es$9#dx}(a0FKtXPrO@_vSh-&gzPy(eVgA
zi|ac|5O-@KQwiUSgmGou4?agCnEPvccJgt!mJtr&-3b}|_xv_a@F8@WTb!m#YPgIO
zq@}X#u!##9CjBg5U@NTSv!y8*V@7ClrW?eKTPEGOrO|jy+QrX^Sy-BZL3D`jMa4Zk
zuFdaLq5J1&O-;pWyDF^@=1Wf`sMCzZtwciNIr8<nd4=WcE4o-Wx`IJZc6qIIKRum5
zfMy?v;4GdFUu}iPM=X4%EBJiAUEljs`V_r-;%{ys9(a;?;NC1A*pTjCV#5urwTF#^
zc&C0kl^!_n#V)3nPI=v9??#&wGv|||(R6S9yU}Cmh8ECtYIHocus|%)Kb&f~l)Q2!
zG24=yy~G;(y}kjkM`EFswO?k_oyn#dd<7hZZxO3Rdy=smU<|OzyV05)>Nl%R#pf7?
zZky#w{h)O`727qzA+{R=L9IRYizjUAVbGd^#5&LdCtxE?&)>Bz&QHXJmS*EVOid?>
z&^X;`v^OA45ZvtO#0aKJyTS@z20@{b2+^;&vJnXoJLEBl5!Nu$h!DNoChbpVn3xW!
zgG7lD7r$aFHnEIKyMmxxGborVh~E`QTuZ?iP%*|^rp-B7qn(EHjpJCfg2eZr_Xu&A
z8Zp$b;_KpuZR9lKixy4C(oMQmS68zM|GD*BY4iW*xpcq2=-v-vWvypeYabh(T;12%
zY&&c0GO^*OTE|#NYjR}#7xO)<L|urYvCftuDZ21}E-CugqUi6h6I;2NOVaBKZ{=*U
z5u3ahz4Wu_dHjQ3!)7MEcLt%(u=eB0w(-PVG%?f2=4;sWH0x}}=AQ0ZOvm1)0gabQ
zk?B2?yncpV?Ph%!@--K+cdM?Z$i2IW<{~Nm*P`_ARf5vD?2A2=OX)@TZX#~3W|OWy
zJDZL*u+eKEXx24GI!pIhV&)iYYE9RiOH5x%4xdhrbSGwyK}9DjpPNj~PZ!$B6==_n
z+{xum4ozeYbQ59!|4qfVf7rTvbNz<@T=_R^Z~V$Den2Zwzj><=%Rk8VGHfj%mdPNR
zJDzH-_s(@g$7JnOv`LNDvWe4)**0=L%w0+LVDp;vZcMTIUYt$#j$-FxvA%Tw8OUMC
z<kVX=-pMf*pI~pbu-*l4dW^N7V*_n$x+b~M<wg5p(g~TLV{LI?gQ>w#bjx)|6Gnja
zHI$mK;fXEW<oaALBW$^y*!H^HlBX8namyFLf?QQyF>Oh%g$3}feP7ZwR?IcF^DZLS
zCqI!(u6GsKx1mQC-cb$Lc2&8f7O{aOiIO+#*Yn<f4H=_wVtqfCL^l`TQCmmJC3|()
z8yHTF#h^dHsS7&?){a*jPY#Z<(bK}9FR<Slj%t;!>q@Ez!j)TM>OuYHO$8L?O$9bw
zsKV)<j^yll1RmI87X=trh?q3yg}1$~N$Aa~rS4_bx$jcy>}60q=o>bDV*X6JYmP>C
zv>{@_TCb<09o}>YcMsGj`v%0YWLIZuX{Hb-Se9w1QuC0S#~KFZxcy)8LdCY^)<4+X
zvi^VL<8SzC_;?lZ*1o#CNm69u6|B3hq!{h?PR^``_EOjT5w>G-Fx-@ge3PH}!~pTr
z^bq3TnT|8oSPv($h=_~uAB6=+#wm(2bSd3^4dyMrxEdFStxChDFCj{ZF{wfLo{uG_
zo7s517@3@_OU^ZbWfKeaB(tW6RyjUhyCAqcwOE^4m}85hxN>r0j-5V+@xrB@j-K%v
zJMhzNXar|j@3_~5P)!SY!Iv(x@d?@<M$WPNYdN-tvtLyx46mv_)zDm9FI$d(Uqg@x
z`}4pxh3cG05FuzANE;M1<YB=ur5?B<LBMqoxJK}w&PGfM2<nuQ;QA?KE#wnInYd2o
zcm(?>&TiTC1D0*k*OVaM5sP1qZrZ^Azz^GD{FvoBIPTzjE0IOy7+}1O2qBIKC}M5}
zf(Q^IN{S;EAWCpH9D+OA3D6Z>Ac63+l-m$eal59g(L1s!=vK5K1>N}}TeImiP)zVO
zCrDp#bOa}tlNekRGd2wcp-Y-yRy4BVp&i_#P5gOp7Jo|ie9!D@R%@2e0>q3~zOI1x
zl|4=6CZ&TV!9z%lA%>#=ajGPS;xucSF1XRsh?N}90xMcGCHx4c5UB>@ETDeC6u5hS
zmcsugPcAZbN*zE35YB??R}%t|s)QpVEdaj`&cM4OZUsMSMu@8&Ap7vI(LFAw3UG$D
zY?o3IcuI>Ia4DHVq#|fyGoVBQ*H38{Su+S2n57-|V`j_A0}-3^B;Di9JfAGijDLPR
zOsp;YzQ{vO8Mr=0856?GtLqU7EiB#$`IzGw-~z?q0tJazNNp)sfg&?oN-neIGuyb|
zD$mWjt*{w&TL-HNjIFYb;IC|5GWU?Qo|4Hutbg!sZj{9SWv=_`?%QmxeHrnwe#FNT
z)2&5<V{ic=xbF>gApJ0P{f5`oPL{z)3^@R)#u?Ur2^Ihpe-<4`&YgndFCuKLZ7#L6
z;I(!l$d>HwDq#OZpHmFU7}Bk~mPxrH{{KkD_D8q;WYfv@UL~Of$X7*0{A&+zW$Z91
zW5_96#fQAje$kznKAG%np<p^!#=Q0p);La{hnjP2{!Fr|g<J%U$5W$2sj;jeBlI+*
zTERcpr)vO`RLV(D6LRnU%{rDg#e1?T$OdxEY|CQL{hXLxBIZnN)1t(@3uH<TYVIaW
zw)A;7j%A7_(5qf;TGfUPN0p*$W#yWd^jb1D@U7_epWG+ZwR=fj%LKtKo-&_VE`==@
z5k^*j1M!w*A30{Z7x$7Vg%XZpj7hcBvfk_Dm7Q72p>DNWQL8lvejY_N%GPTBASm7S
zTii(!hI>T(o*A&H`Xd;o>IvcRgwqwi6{?QNaJS%VX28Lql~NV*`B+K>UKLR^B8+}<
z3E@bByNJ6<DGFi%mzle5=;wJMip77}w8Gr{&{3P=DjI!OBXnh8PvbolQE@BcPEr1W
z)4-voG7esihId?g*$E_^J^Z4mCBt!3Gw}1vZsWXL7mnoR+NwZ++(iMZk)vs3ZOILf
z-7CT*yJ=6(B=OsZNw6)0a79=>HB!UI1}O2avEDn~A*h`4i}|K4s5~_pM=hX8o+qwY
ztp%KJ_1fM_pBjP-DS2s_=OE8Qak)hWVS7ztx(ydcpfg#=>j-BCSo>HCfspn~t3Q{R
zzmz%`N7#c3w@%Ka`_Hm?Ei(4$M`M$Tg+A6af{TcN3R{Y&2N65zf_g-EJKLK++s}qh
z($uC$S@T7V^u|uG^ELQD5t@;j#C+H4?P1uh*cHl?Vi}rNwkgsd1iMN_>Nbj07N^D!
z+`|JVF$kk<PA#fEt+^~hR8aRJ1He_Ra<UBaD|4llKmiCyh7Tw5kVV|w{$IDfV%x8_
zesOchhU@q!U%#s>5dYL8+?}t|Mwi(&qfmi1G!fMB!6oWV$Gzd69D!_915m(FeU^ge
z6lIl<Dur-KTZi9$SXhVKbLq6Wb-0UA0=QnB?ix*>?S|eeR7>y5J6>O}H`eN%ISvIA
z$63dDB;`TJL_(i;eJ<G_&8#=&o;nqt56ELjRb1(*%SFeK(f{`n{U@t5*Hib|hxiuE
za{w~8;6jDf<VwDAnspGe3(@P?b?;hd{zhz7Yno3%KAB!t%A?$fOL`;z!h_tc{O#O*
zxTs-7Ze>J264SHTeaTb6QH&_Ra{7|knDfaO`Bs__c7v{#6^ex<-%=*u`pS<8!~PJ_
zZ)Q`tqVy}2UXkSsx0YAe4muWYFxJ|coE>9*<7{{VSQj|%)m~%sbwWM%YA(Q0M>YSz
zCK|rR&T&kO__gG<x#Y#k<mCYZx)|yKV39Z4oX-Oa)M=j!fnk+Z5`g4^Qsx1X|Np~^
zZCAJa^``Ude^v3r_t*Gs*z)^acdvalS7sEi%%v>HY%LuQ@6<W+4MT*G*3;ca$v0Qe
zp(b6_)}#i<+580Y;8-V^ON=MnjSli;FUy<S3t&dC>kO18R^OVKeJj~Jof;fw@i=`k
z`*>pUMydyfKnq;g@Pen#$Jk;NRR*cq%iif&`cwzX6WBy2U3{)B6~&&=%}a$^lNYYB
z(H{CDKf*?5yzaMI!)*G*snvt2N!-|RcKssjZ_nd>O|8}qpQ>r9URJ1llJ`rQ_v4R!
zoZA|9E-MzQI?iXpJgZn!s&(6E0j1H9nwqqV?<6Pw7FKcn&fnuwqvm6|q*}cCIG<qI
z)H@6nTeF=)93WPSEKDH_*w{?+GNDAKYcJp|c4kNxBnBj}oM#<CR_$hII#V4BDC|l$
zT}({R!BdnPJB@sMAf@F0P!U_)t=28r;-+Cqr>JC+HDZyt|6{}=PsVsYL8j-ZNJ%pp
z9359gWpX|Vzjt2AQ&?1&>K9$pG^D>N4?0*nNAJ)%TL1h<g&5#Ss+lytv8ctKPug(t
zBN)JXdjT1Y(qh&=l#0e^>qm|MNVh2c<@LZ^-5;%`Q|C{n&(<S(A4SFRUE!MPw<pt=
zj>C+HB!IL)z@OV8W2vP+T+q8Zhx%k3PY%}z%q(0a**xh*V^oAZJ&my0s$r1sX(s>y
zkx?@=;kAwAvy@Y<SUz0=JhT)WQf#b|*dY6VL*-b-ws*Je-uPpDkgrl+frh7d@g1-Z
zFg)2I`t7&d0f9pR$0UNB*p;why&H|u7Iv=9>$%Q%I?`w1o}|69uQ4$@N`H>cu-OyX
zwgl?5BES>KKN%f?`vQ=?1lHROEZ!F3fq`ojfPrwr2oDSp0@1F30;#>zG$txiaHFiT
z0T2q<gS}a(-)%V_QEZ%v*+JS+`B^qRla6&IJAuGJ%DLAy41kEtv&c*Y%pe;)P2lmE
z<a9rVrD7-GR$U!NKHJzNoHiogkaptPrR4Rq7$6{7+0+PE@(G!E6{V2Jc{QWj)O?QP
zx~dGF$$fW?_T2{GC%L*__mx~mE>t=T#=HX-h6$UPMhV9yk<?p^qZ3H!<xCBWc$Kfg
zI0#{YP56awZRakV(tS-xlVD?jVQFAIs6outdj`4^9J3rpv<@82hOQIu^R5k;l7K$S
zw5jR0;5+Sr2Q>~0Bstf<VrApX^{l-K0|aam+`j-!%2SlUumES!b@UvNB`K0&*NS27
zokH?HlgqHTt02A;wq52{-t0l4;uK*Stjgo@2}n<{WOAgHwKfwD5;rV}HJX4-HcTJ5
zEQXPRBz-hbL1+@<KY8MU*EIrDBQbZ5_JW~{i3OCs#OMmK8zeVJhmr%sC`e9rP|u-T
zawvwes~6c-sy2XeBUqa)oP}kBOCcJPm|IAm=t(tPOHB;2#TfXVs;1*pY^cFIHJXmK
zB4(PHY0BqZ9rces$JQ<B8IxXvI|5X>mGk(cpWvc4-+5dF6+1WB_~E;aHLY5Ae7515
zmW%;Pv9?BHP1yhOifzx`^`nj7u8fzm@%vkQZw=3VM(7ql$<;0HeIKlPuF3OcJKiaf
zq`A;DemBi~ho=@gh9%8=sdKEE&hhb23!USWxg32@VPve-WHZ^tlg%u;h+H6e#=xrH
znJMDX#v$+IRp{F|f{GjBRPL~%BT?!nl;Z*OI^)o>SDj`8S6Ha)Fy~<f2-^Wgx@^8(
zl-mk>Azvp)Pf){W>OirS2^~`HSR=Kg;iXS;i5K6P%O6$mm&AK-PO#q{w>HX`Odk+$
zbz6GLBww!;U;7^AaxdSly`zBKlWy(z>ejZZRe)O<==vt(|B_+Xh+#$k|JjOd|8>jg
zrpEOLE6$dhl^wS~$BWGV;;F_%o86G@j{rDbb=}DA6onjAWN6XpvvsKpl+i;`t-68K
z2y}v0K=h+17;1JtY_fr@`)UIX!4=0<%L)MM{e{@jfQNPPD82uY?|C1PlycGPaU>6{
zo)D7+wl?7~nHJ!dOuqswC@UBsBqy6N>n(-j`3DKT8vslMdyj7Qz+18deuB>i5<bAA
zG{P2lg9Pg@CJLl|4hcoj|Dxu=j1Z`l>GHnah}`4S7Igs!nm5K+Ccgp)8($vDdzlLq
zsAX&?AV1;fUB!?46+g2A0OjH)M3Zv?1_cBw;8g*CDjR9Pe)ng&l3TMUSIPaz?P&K7
zBE1OnE@Lirpm1gKrXoAdb1w!jph8A;T>`3-F!o>y@(4|r2L!a{Gi<2U>!*sE<v)AJ
z(SJG5Ak;Y~nySJNU0y7KHj^t4as_P6!GlC&2vCQh8bL>{;|lWQeI6o2N2q5YaQYx%
z#;jrhv@*35OL<$=ds@((y7&4=cMFF}9<MyGri$YS$R!A6y0)#DWd%hhE4E6xVyk}l
z;{_zj{bejsG#4<kK25Wo^6G3Zd8MRz<<Z9qNR<1^QlfxMbk*mms#RXyvL%<4GM8}s
zf8F7V?O)&8vH8Dkd;%ZkORm8CxdL%Z;brG9kyWgP0E`@VBbwn46@j=~_SSHSE3fSW
zkmyAX6%v6{ljrhmRMVy424WjZd10l>y|R?|%J>tX=blK8myk)gMRjlTH5P56fcg>#
z{~+33Lz1h^L0PLDl(^>O`)r=$ybzizLUHM<C%nEfHVwFgQ*6AQjZls^kOdR-^NE=@
z(X)np=HtgjvKVSK@*qHTY`!jm)tGj91*+cHhM3Z4{+@j1ac2(^?Y=ysEe`FvR-|iJ
zYr4-yFB`MGa;zoke)!>@xc0=Y_uE2xEr@naVw4sYsG6Fs+U4bHmPGsE6D`^Q_f@=E
zQT6DKr?<bnt>x~qt^a<@|Jm~JT~BWQ(@n9BLmU3}`v19p`#PoaMGRWASM0!jcYQi`
z^qHrgs>qD!^R^aY>d><sxYaZ?>M+98KnKBK^inp%3I*cyz-lnqh)}0@GZL_OJ+X5a
zWQ?U#M`(09#=CCfn<j7SLnBtM;@V*f2rHJm>w7GIEOo3s{SKADrN;V2_dfKR+4=Yj
zPwm^CT?(%21o^GFhGrW%XO?vyTUX84fqPdbmkCi)>XV~pZq#1{lS8{h&D^6ndsJQl
zzZ~(^bj!iy?pc{!CSYo*Pfp?OuXP<NY0i^V1*|G0K25G_WpbGvA4`66dbMi!G(&SN
zJNwBkcazi+gfwCYcC1V;vjbphPYyI`k}F4Xs2f=3<ZiYeG!z?8Zu`pQ&=E{?A~yQD
zFLydG*$o9#09vAn{us86#Fq@7%Q|Dn4s2VQoE<U({$RjSG|OM=PtMTzhN4@h=44Oq
z<{90I9k_dCa*>E04x{m@iF~Wlo}3B`w_3rIb8J|-nUl*rV}a+ku1qeh1%u#FzY@@d
zX;HFdi7A-XD%_{0r5bw9<W|pEdhEcKmC5;yu<B@rqlZhKECHm;^JrBCHRf47mvshY
zdDqJ1!a?NR+Q=1C+>(EC$X5he8c<Ih;Dp^2;)EZ(a>m7z+q^P46=`>t>-t?|xr<2&
zlcnlVGN$WbgUHfQ)asjOTsL-L)5_#ByDFCYhN5siM91Sm_rk0vx6*`E1B=BDY+RX~
zi6>?oYB1#2BBeb!G_FG{DAkAj4i!_&m>kY<#Y{C!*TinQVP$fD$0a$Zp>dR6>XXB6
z2_f$4CRi?eat5EAYpS}74Q2hx<Xl5{{CX&$LvtzN$yxh}tpMa{`5fIw{om^3a?e0R
zS+_d55QK)VX_mF@jVR3xg-=1Zt5ppfiq26Lmv_sYGi1+Iu1wA!GW=n{lLXC(D4HtG
z$(j4ncnv(KXg&?frE^n=i!%-ptcm<x-6Z?}t}PGe+5c;1>dOns6_6_+S3s_STmiWP
zas|p{1!4z2x}x;K6mg7zquCKFRKn^yT=j@vL$%X=x(b-EEL#M9c;O^gd~l-bt`a-&
z$jamlSJxtD$PESjCGK;x0e@a?00GT5jBMuyw0HmqqJjVpE8U76cz9)UD)OliwXh<F
zU$jgsO?4el&Ll$^1w*D~*jAS3#;R7}4R(E|tC<cAt%p`8r=VIf;^;7{OW79TlQZy$
zRG*62oRj4W#<bOH#rEkg%zu#OgDaEs>w&P})WVh?Dsi5J@vl+%1s*rs%;Gt63sf8E
zx(Xax6O;SM%H&)<q#(~R64Xnb=gew2pU`<su^f2uSM7iPOtICFHxAKI8*ZrwRwkF3
z7g^f7CGsw@EjkFlz;k0}P7Y^Kl4Mhl9#w<=fB(wl3^Qm2UB~hJBc<%Q!Sg}gx?Zhe
z3aEX~_H$>Qv3dS~WmU!Y=eM5Pd}6~qKK?Gh8a`|CqM^DcDLFlpQ41wa3P3L8L;{41
znv2g(vgkZe=F^wES^PRc2Sm*^Pot~{<nD6ibyJjMSkDqBK8GY=t!QC5hCYHQ{U*S#
z;RSE9(Th=sa`Bbq98me&00azpU+;K*s%`=w(0~x=`9J|iUlIWE(-$YA>3+hsM~5mL
z0g`WOX@QLq6#nWFukGwI_I>*$bTRbWTD<l_pne0mKRGyxix4hY$8mgxjZcVQ6u9}U
z_q5kc;KBqt)^#S`7foJmVvBLscsVgUOJ8j7NY}rWlaG&>5yg~z;4`>@`KF);6eMP#
z!k)^7?=;~9S1=q@WdC|8A1lA>K;uCwNgzNeBD}ua2s>e<h1d%2&548vT|m~1s|r96
zA`n+trAW}k${?~%M-*3}IiN`1bRE1~z*a!EPQ;?20T)Xt9x$LF@CAGsnLDmOghgK?
z?2`CP2Cs&VaS*XgW-Xp9>YbTEuO4s}u}fpWcXjv{@Or$HF9xK$uBXOZz3aqGd33En
zfM`C|F-%kT<dv8_Yed)e>vS%JVn4_gLWMF(6AOK4vjFkKI$BfFT6CXcSI@K2N$k`h
z^VMeSLXe<AE*q*Ne0TJQn41*x36!2%YDiu<mY61(Nwj@QjSUlwNaI|x=`1>~(BSE4
zvSucq31F6~K1I{fnObh{7P1eqUHT8cwuiuLBp}Eg0)oU1jTc+(%jJQh#a6pO;$AeG
zJb4_-ROZ(Gi}~2DOcfuN4ICo2Pql!5C~<7$-n~Y9cYKe^>sRt>1aG5~yoVyeWs_8K
zWs#sP5?m`Cn%n;sEfw1z-}>mL|Fr(!;p2z)s{e8|*XL^v<?i`gif{Vp+KEEMU2OOo
znsEx9M+gBZEvXw*-eePNZ-!EYrZ;FMms*-3OLedj%1^qX1>=RFNDab5o=w$K?<u-s
zS38D@=5<~+KALE5lWe*{S8g09%NgzI=6i6d%!9i`b<ZvqZ$y{qJiQ5hstgM~avjSl
zuVOQ46qm43T))mOU_9Ja0QTlXH6ApTir0X+r%2C`lxk}w)#|_I2ubpKt|Zx9RFY5`
zF_$G&{$66zB)?iTPof}S6Jb!KbL)o4nw4%=X^A@=<<<@uD5bnJXwG$Z0g1Ay6eWrZ
zca-i!g(hO@(j6t;QJO>HoZplQ;xAcTqU6UKHWrmAYZq&v92e6EZ<X#yf0guCt&zV<
z#Q)D%Z2RTbz~=D{Kd(Ihp_%{khYPMi{g=Na^jYp1LG9vd=2vmIB6LOpXh8Zb_d<|A
z1EHSU1=e_x+JH4&Lcd<}b2QDQu3vxyBQ;W&jxBj@P08MAM1g4V^#NoQu!RZsb}Q@X
z0^ZYBcD}|t-3J#t94zRZKrWP;bI27)jUkDl4LyNZQwl^%gaGSf^RptkL10ID2vME7
zkl%ITW>T1KESGRgJyz<mzqtti@*i+~o>$2x%;cd$QB`4avH3bU7r_T!R~#<F<dqYM
zhJv5KNnik=hOprV*4c~zQ+9lbYWr0rbfPd&%4;dF*GOK+KmSD@7X4x_5f{}$$**w>
zWWRxIiluT@2RQ7yYLyp|lypd!utPfj;1`xD7X6zt^SW<Gu{qdiw^9_A6Rc6>T4suZ
zgg*Q5q0j2qJz2o7wY3!eS}XbXsw*4u%1FstMM{MIf3jlRueMfij%@g+l_v`tq4LBB
zwgT~QeU)bs{RP+ZGOE~$I~#AQ=Aq_*tn`yJFya$40|JYjjrPEzL4iU(dOixy(BTc4
zaqyq(vNf=@T-M5R`N>y=xczP}Q5Pz$0XehiTdboQwI!<#Ecluz$8!ZyA1w^MYn?a<
z(oT&|lVfXolr>)jW;yblQ`h@}@5$ohY;nqKd&?UbUiE4vXO~a|lM_Pav=2xa!mN{B
zcBGVDD=8cQ)|YR+pWoJfn_~nC9DdL`CGp2W`WI+#zcPr;T;%cRTS=g4h0+2X+eR^?
zjEKsJ=o&>tzw{K*^poGu-Oh`+Q*(k5qRJ#WFr1FPjRHdtS(ZaRtYajnr33;3$bHs)
z78D`bAyDqhYqcxaSN<V1k}dmUBC3|5W*2o$Q5Dt9G1k;d+i=HmHavrlRH@buqyba3
zYqpmy)uOrz6<CRx^T~-;_ErmN9r0@vL+GzzO9-K%r%FA(NIH0XE7(BP52F=%a)fGt
z^LKF$5CME{Ls0>KA$M`Pi?7WtF7p5Xs$#ou%m3T-x9k5!#b143M(!H^P+Wh8%cAE<
z7G*-(6p}^Y;na=Otb=+RL4!g8y(o#K!uR?19IF7xGfuUJ`gL7HL+5hhJrhXp$oeyR
z=X)6?`S7D8KA&&b_r8=qMX#QE{~vMrQ~QZrvb~-B0q_EF8+dNTLTc=Ma;_edBy|J0
z4W}SmSWDbHb_E44to=Ca8%ND9&L-Q&*xA~5qp@V$IP0559Za(GRI=$JdpkyQsW#5)
zZ;)I9rUYtbj$dHC*AolvUgH#(QRfniy>w-f_tD3bL%2Xojm2P{<CNDI1!e{8Z|~R=
z)x^x5Ow3)O@Ai$e{;1b@5<&D->_l>|uMm#`y-ES-xTcqvl_Gfza7Mi58CwkjMSR$J
z5!pQUd;7T8sfP1dCLma$av1VD4P8oiUxUlZJ2TJ5Pk6oaKb`VUqAgx$asgJ#VsfsX
z#d=UL4-SO!&Ei+seAJsqV#qthB=fVW@mge7v8f3RNW!Y=@9|m$yDpd8|6;yJWK1E2
z1iByC1}b7=X)rN2msscoN9Bm1RTMA>V8zmimrmI=a>~{}@iei=z8XE(iob24%&Kda
z#7BEaYIKw>Ho`B9h{CF*Sj{%{8ar6?0Er3sZsD?p2?Y-5Kh3761yB-wrG3n+?Zc^L
z?<lM*@0~$@wQFpMQWld}XOd@Pti2{N-6paQaWFAE<aNKDZ!Ljcv}(f!Zn<MPGKL}9
zWsTS+{@h+JO+*zyrZg#33PT~e?20n*3iKloF9bkh`ciVPE;-i#*ey1CS|Eps(&e0d
zLQc`Tf~cI0K1k(cgGdam63~eVb4#JVc&cM&=|>{|f2CsEf86qaH~-0o+R7_yXqL-M
zt?3HH_dm;pb!>gEy;xjWFI#O8tUUQ&$YOXag*b*F_D%p)up6!rQ%QFcDjGok((x-f
zIw)bdR_hR|XkSuZB}mdlD`gWc{?a#vP`x*otP2TM%DeBW6XeS}uAu6Ku*)Xq5FI7r
zZEsI4jpHnQ>_UlT@u|eZ+c=e2JjSke+={LVo)}2Vc3IU;OS&#;y8hTVc+mg;T(T~n
z#KpSCGSl+0>mqz_03riunq_7AOX@6X>io{v#jg8ME_oJ8#{yvrOPg9Iw@rAli<uA^
zVjqSMK_0DzR@N|^YI=uu<F*S3h9;)p=5Ul*6(@+RplvcTQ!L$D?)X_Ig?^H>OPRFm
zcYlrR$u)a&NxOKu6<kj_ffY8s01{petOy{n<-khIEE^aQffCXzc&ubKj3PoXS)mrH
zv88AWDJDF~0!CpS=i%d8%?=R36iV&t%XI+r7$jj0Tj-!5#Q=JG5Y%)SAqSZc^K0hA
z<lFxh2P?K6+;Vo)Kd%2dKCaoTVcUyb#MJUM_RKaqg+vT_hdFvHxy|Nai8iq5eoFlv
zp8&UbeV4ti;q=)$Am#(OFUOHa?&D}DyV?!@0Yd@_nO&t!8#Y=C+`rV|g=Bw}$966z
zV~a@OLFF#4j7p!>Tc?nsgHztrl=n`HcVmivv2_rUABw|{P;8Y#t>mkFn{_n8l)z<E
z6N79qhHE0wldPLepRLDsfRTuU!lMldnC|BA%ed{XGhXL75CD_08!SG7q=RH<yI0Fm
z{*jwMF_>qMpjWSH_;k%Njj}>imSVA_#p2c%xLAyFP_#^TEG`y_X{IiS9Q<_j1Y4{l
z_L^;Hja?3^ic_r+W39=Nad0RfI?P(mfP;wj_#hi6R@3fNF@S9Lfc?NxG@|`7BB{iC
zjorZR#Mknf0qu2kAG*Tps!~pVZ^;a$%nb3zzQr@fI8JjWFBHlcTjhde+c7qKl@oZ5
zHKU7ONhq8uFGO)ksU=OR_Vb|BE&IMn#!M!q7EKb{B`|gf7MY5fo6kUDV!eyVV#$w|
zBk{_uMr_eUyLGp$kRK%FmNMlweDXQ2|JUu#wZjYLUcto#r2tj{n|cRHR*AVbk!eqQ
zHSbvkDi-A;2w#c#Ir0mUP>2D^=E>z0MpOc}O`)EI#?)S5jrF3|kJJs`?PnF73Z?<6
zGRTsg=u9o3lxhZs0Bi4NEhhnk#TJjz&9n_CrjL33!>jjzd@bQ7VD+u6wq4vIZU#VD
zskR00t<zku;cdF&XTA7YG~e=?7QMb%fe?<b19*##UW2vb)i!%g<46Y}>2jts)v@4>
zwF<)|PY+t@aj98S53;2Ff9L(Q^6dW&8-8A~{fVt_ZeG9#`I0LjS3s`7Z{G^UpZhjf
zM?~^orY2oT9U-aQHbAcZ@#EwZ6iv}u8b}p@PvgoJ*4vw$z2tRW^mt|s2D6Dc2ImRy
z<?h>>g(kYvi6Nc$ciap!OL)hvKjs+*&*Ykh#q+@gS;HYMoEW5TBZ+3I;Z~NRf&SdU
z!!TXTEGsm1X&IKXWf<3A;@&r2ubH_M7s~h&Bn$HmBulnl(~K7#!5*BLiza3obMOMt
zlocs6D)I_6$CXqoZKXA`mEzyp&$rvWaAxs@E}~r?KcS#w+c8ok#%2Vx_OeDH0w+)&
lMvi%X=<+?vH|8unVH=sZ)U4jsY)dPviWa#Om-0^h{{bprXW9S&

diff --git a/AITrain/conversation_data/demo_conversations.db b/AITrain/conversation_data/demo_conversations.db
deleted file mode 100644
index b64a0fd3de3749b02c688af7631865d2195af117..0000000000000000000000000000000000000000
GIT binary patch
literal 0
HcmV?d00001

literal 118784
zcmeIb33walc_&Wlq9{?Me9E_FIkw|<%m5giR2++OY{juHJ3gBDltB=*73+#9`AAZS
zLlC?#os>vQA|+DP#X}M$9^jE?({|HsceDQ<Pnzf7#+EPxVAF2WcAMQcPuu;y-v<l_
zGngR-+ll0WRahG5_r3T1KKHdJH*EK>EZeeelXv?{A!~V7c6Qd9l`FHdvaij`%DNr@
zOZ?2k9}5!yf&ZVa`rF`7d`4mYwaW_rZ`OjmhqChDFPO}0!5^8Q3<NR|$Uq<ifeZvP
z5Xe9v1Az<#G7z}>An?qBMJpb7Ap6Db-ja>}@^5W~C$YZV|E(SV%|7jaE_~$KM~k0-
zbmjBKUwZP<m73RncC&Yr|8scf{eg}1b5`7ce|F7v>@!O@csFi&V~2nJ_8r?cmnZ+%
zf}~GOes1Mwm#j?jv!Qh5<LjP(bnT<hu6*X%$Db;G_QjP?Jo@6w;uoHO`tf!6xThXn
z_x$HlzO&q4UcO<==Joi5=O6us!HejS>o@P%RN~*3_#T5#+v@keg%=rrlyA%C?f%W%
zO<u8m!zO?EcJHRGCa+<i^6vKaJIeiL=icVu=zrS_oUJeSZQ15u2~0hi`rwZ}{p_QU
zuU*HW_SqSj+>f_B_UN;Zu6yLs=T_oN-}Y}S_ihKg%2%pirGxDKm0!rtS#jTe*;Ouv
z?xeRS{l~l{WG6jkfa{slF#xF%p89USH@(}uzU?@=2}Ix7vTbAO+Z+7vn7js9*|vQ>
z;m71f8@=Vg%hpmaoNr??*yP{t#Y?>GC3^6A<}F(B)P31m8#b5v-;MK>cgOZE?7!D1
zL211;=|AT~*eoW2vk?EwF8DS5c?17{^k?QL1Az<#G7!i>AOnF61TqlFKp+Ew3<NR|
z$Uq<ifeZvbJrKD5+9kNm?fKVd<=<2AX2E9)uFL=L1@Gne7t|O0EdL)090mWaV6xzs
z1@rP7^1qSaneWWMK7Xp<>-ih<Kgj=f!Rdm>3Oe%FetMvhxk&~B83<$`kbyu30vQNo
zAdrDT1_BufWFYXn27%@C@5y#KEIxOsugv9lxo@WjEM+#Y!{xO4ZGQi4^nlau^*Nn_
z#p(08Zlwpx{C=mstjts5_X@Yr1CA1JS&5_6<8|0=H`4=Np{%6LRpN4bt<Ib1fzlG2
zV0BsruhZsT!QNu^yWKvI;Bl4NZlnjCg1gjJY8T3E*0LMu0h`-pvpQUU!R>QhPY-z9
zHn-F5_qnV#|8?|$6$i0+>^@Ja-+3)P;4QP_41(9{u(+;apHb!(%G`dB*Xs5yrw8nA
zkKgBWdR!L4TEITT>l3V{0<I%i^VtI)x6grpo-&Jd89gAlN_<YgzqCxS+w<swGN<2a
zwYZ$PuWu<m;I)-nOY9|%GP}pVgnfp^<M5VPtUjB^ol6hc93?)V-R^hTT_uaz10K7}
zDfpc>i^Y>e58$9Uh0AFXoX$n`fZyj7>{c&cVskH~2kaJ?;4LXB^#Ozn=mEiN@%ueK
zkJaTZnNJVcyf%+haM;`)t7l%p{CjfkR=;5NSxRg^uaolsHxx7ziQiuE>w;eu{6oRt
z6+{Xy7W_@YPYV7V{{N2(epoP7Fj{aH{{FFou7dW0J^barYh01JZw3Mx2xK6Tfj|ZV
z83<$`kbyu30vQNoAdrDT1_B==1a1?z;L~mu-@xx%#MkiqX7TIzeUrEjzgLJ~#P1tL
z8-8CW-htoOih1~b&3Ex@`8)VkunoWR-@>nDKK#mi8NZf3fnQ5j<5%wI@N4l+_?5F9
zzZNZChTqu*KhMIy%ufaa83<$`kbyu30vQNoAdrDT1_BufWFU}%Kn4OC2weRTxGsDC
z{qvINWMusl`TvWvMzZp~d8e2BEcgCJf4kt){M+!rmHksyvUq9MJ=tpm_Y?=0=Gata
z6ANPdr={K+si|EW+9OSk{9q(@_?Rr7pa+hglA8wP+MUt9N@?)$FD4qLp1xS%xKvv$
z5A6(3w8a9Gk%{r}SW9@kI@)(MI(;BK-Vq*eh}4~kv{yv~lTv+$w5w6BYLTm^@%C_N
zUpRDH?iiN#b;_rY$o0LE@zL;TeXQ~*-WZ(>$p@$8&bsJ@qj<Y?{7iVP6|aejyV-kt
zrIvcRwneJzi`6vZ1(Dh_av&)04dNT|0(?8(E7hI|k2gn-pOXgKBRx~$iS|g}32D6L
zQbncBVskubaX)Akp0|36td1h9y%3L~1m)#bz8%}PVSdQ^@(tT}ut^++iAOL7!so{@
zl~tcxxk@1h|GO9@f)Or-TQUA(%VzJ!<lzyiU%Sm;?)Pr<y}7=0L%DAUn~%aiwS24J
zk4Lw9w_}*b=JLYrTQ>XY6B3`jVf%Z9<=b(*H@3gY-hlZdd|Jqx8@84gmU%aA*!Uj3
zA^Gw0?K?_wCH4;WN4<ktDw{E$#cy<_ZClFt^cGxe1IDV*ln=}%p>HVh`rdkD+m;=h
zOACE(ZrE7*=9Vp`t6qKhga7e|d4Kw+e-`+1&JrTBuN4RHNusg&k<hTz)6Q_y{(})|
zG!Sh*BLy466HTDFRiH3ws5(44E>|9pPM(%_)`v$cKoIhUT3H-Sg*l6(&@H&^7UAVo
zIJw*l;9@j51eYz_f_ufsnyd$|)^FWHV`VT^hi;!Bi5Vh!V~cmA0T_MW@*NtwpwG~g
zkblz_0xw>>Y0IVr!n8C|&ggCZ);G6#<A~F~;_WRyZ^@30-ffH&REXN-{pOZ!>-}#h
zK3Yu+G`?p2_HEt`AQPbRr{Zow5CZQmVzNS9TO7P036Jy4Wd+bUI3{;iM5lL3EjuH7
z4il619*Xqs2Zx437o@3Lx$3YK90r4fVF~S{jdEv4IMf`T>;nISucfIrsdgyRF)EGk
z!Q17Qmgw|QN)oUX3Qmv5W1B1Usz5SByD2e}fRLW4wLnj>D&Y+X#3Y;Te#Wci<)Eh0
z4L<MoE!$G?teQ73<6I_${gDV5s1N^sD>PWuR~7>GxuhnlUQ9y0DF=UwwS>LM;V6WP
zuv#oHYXt0!;588p30fi{uI9G6${Zu=P(7|?aQ$f#kEnH->I0J4A@SN7A-B!%19`6}
zE%5e+?Hjn>*#4#;^R#fKO`9<0s~oTJZriYhq{xR)eu*X}1%w4m8F-7U%5&6ZnX?!i
zOnhhr2jl|Di_SXvLLH0%2#a4#9M-6FTam+6h*DRZ%{o_!V8qSFdUTG!F@o*vVnR(W
zNjhJzHxe(^OR2;&M03g7Uj2-k8ee_*Z;!t6=Op{TIzNuJ;9W^rGgsjkEB1#+_eJ(L
zUaUAE1#2Pt<M`S)1XN0Wd*r(PQfpgEtXT^k4!6q+Bg_bFDei6}3H7fwP{pdRo1KLe
zO|K;yy=E@)mT&NJW5T-?lk2w8sJP3}{c{rK+GD8HKcu<%#@7xUXb;%tF=Reo9L!BZ
z<}7;LScB^ov8~8%DYRSMHX*ewW@sPlSt;>|T1x4WX=FvKu$d&!^q=6{wxN7O`FecN
zR!rRE;!F$9gkm?urTXi<+x-4*>q}u`ZQQbzc$|=JEJW0ARU6UE&;6>btSoR_HjtLP
z_G!|<YPo5yfopfl?Ui!d1^H|@$;swkxvqiir|w!9Ua8P#vlTjR4#5WdDfy@wwPvQ)
zxc@(xm4DY#Pwu-p;YD{ZcxGN>*5DQD{s&%K&P2fXii3A2*>icbneuY~QK_a{>ZyS{
z79FgE&W-f%0XdOi2aN^J!A!@4($E?CU@!dtixtQ4PotEbMOIg#%_<01PfAbiBT9#q
zoRrd}PsQ_k$zUv73?#bx04b$_kz7g0x>>9&_2;Wa#gFYjB!9cOfFbvx;-HX(+@+?n
zf&wM|N9Fy4aKM1a=rIw3;ujO59(UG43%YeIz+K9Nq=fX=k3%)<QIpb3o|C3D6yaLj
zq|)R{xZ^S`gt_1rI(neHj&$@h`3#3|6bJ81!r>BA1_kGZN1F(TV8r?^DcB2SBDO{%
zzxhaH_=q&r5(!nMw7Hx`7H6T=?y))H(&!JHc*I-{CoD<5^{L055nE{ykWjkG`bh07
zC%#n|64QTaDvnpbJ=rkUzu%0GQ*r}Px}*IM2&G?K#!&jGHVl?)%9a!cL%32Jnv|wb
zMM4KixVALKj@9b5H!uj+w7LY_Z=N;h$aCp1jvfgqxjsRlv&c?0vemv0q*iK?00eID
z|1r6)*W@t-y0k%%#il$tgTMNY=ukjD+^ff&qsZ<lbXz=D%bbJGDq_<LAtMHwBZO3+
z>@VGcSXDgAo6LV+pSRS%2_B7a!!{pc-elaW-swxJ_td~o@?*Dd^OyM%%f`U-Ld3y0
zZg`7-V)Az?Jdbcf&9)7Usg^RFyjmPwmLw;0%$bTJ$)hKE+(2sTlA1chqs`I5ptNfj
zH;<+M)|6<n6gsRHhsW{q++xG(3u~yY{H+ZeH<qvWZZ2K#E!lvf?+Bq3CW#mR3N2t2
zg%5!1F`9TpEv8jy(UOW$ye8#mr;PZhK0GzfQ&J-kxcM*d2LglsB`m5OTu~gnGYNx>
z%(;qrqws0Mqdn2{BOH^_3qh%-8`-5;O^e#gS+WH9Cas8<kB>+89G6;pq7`J{$ot0R
zb9<%Q3$cnW%2qK7fGsNxPDE>liApA$q}E2M{}{!s+k$v1T0sdbyf;#NK^|<14)h>J
zLB`YmL22Za+&M_5Fzlzn5fas-1Jd*XB*9__dO%M&$ao_i<ZvrKm<Pe>vun>ty_3<3
zAjQX<hj0)aQ0nTAiEYv8F?=2(U9p2laQx`NC}m8Ln474?5qMG!FNo}el9-?m6wl*b
zk-Z1-Aikdt)p`tRHYD35v7Nwhx|^E<`lySe$mS_@I_*xkd#)CM8vAO()XVGyC+IDM
zlmr3-oINaNARX{t{r4ML8<%kd6B9PUb~DO{M}|P>3r_zv3946enV{NG9DFE=%omyq
zD&#<c!*s=!%~EJ}yImx%^bkpsHFL@60j^qH>5XIk_ZY<4j6lXPKZmrN{)<wY1uFb%
zg`f5^s$dxop+BEq{=qhg7Epkc`~T;&@;!N$B@ZqxUevVUPv(6)>--h!|MO!7zH}4M
z=q)U+dMk-2uQ!#w5O&<lpv>L=W_kY<7ez?BbI$@Mk5t<(*EJywz~cnb=^98Qy+(Bv
zIfO#Qc?G9yt|CW|cVp3_LVi;8FSVsNN2I4<gM^Es{=Qiapk!CWSe7L}6OY4z^1fC>
z-hS0nD;N?h{l$pQt4lbpGnFjB5zomYC`d7zab(&nq^7+*Do@&jE5E_U$h64SOZ3R8
zXw6A!<P0glfo^Hwf_$(-8aRjd((9R0gz}O0pF!SU3LZyINObr}cpSwcli~3ma$d%(
z`5W<i1Q_Hqm3Wst*dmP$%X<dHq5VR5v?f-wGq!gpUM*Ekr!>P{MRsSQ+hMc1A=lI(
zpG5<i$ZxGE(95+?(7`dJMm?z;$Wr|aQ_4c^EmZ>V)rWsp{MwTgJz8@kQ&h+fBv)%(
zYbp~=^80#2!&M}>3hjs?x#8~U&u+}mdYPPf#K7K7;`+qr>cRwCxFpzFe>5X2LjBRn
ze4RMpjAfu#BakQ`0~rH_&R-uOTf=$-L*ZS;!Otg2glkMC!YZC5NF}5NYaxaMS?!1n
z868xAPbD{|2cwzkp$4VKWttNY>c7o^X)ebB(^80vbCQuan~k9nED+;czxz4`7dBka
zbuo~3dlJ%?e+0f;vgBe#Cu!rccm_ynJslk!CO;@t5gw~W;T~KFVEkgmL0z2J?xs38
zw-wHS9;Vy<UhvlDH`LWh`d1}zsW<vnWTXX|o{|!esF^(R*eradMVtoy$;eCXz^aa#
z6c5pUbV6zv`P8pIe6i`s2f%ez)^!}$U)NSr6__(<3TL3DCmcG;{Y388K{q4FD;;Wx
zRi2dh9*~DaQvGqL+epWN+*PCao>EgiyhZLRk~*(DA?+4H2&;5tAUFy=R>32DY`RuY
zCMlKGN7ssT5*I4$Y6($`U_I?<Uy&dIJ#i}T;EN;oBAr{c=vqz!tF``MzA3xIh=%DM
zL-6C67!Nx%1f}UdAQ&M73T_~#abR2y_TgpHz!*p&+P9OetqLlvqi{g*LU^>FsAOz3
zI^7BzOE2)9MK*V#+wF0>k-gX7CGm)w9rZ{x=6(HZjnt|NIZ2va{|Tw$`iXPSA|eb4
z%?P`5Kb4`LLtLPwiRJabd=1mpNY;6h#Qri<Q9xCZfk{H=5UQ*j;CUG&T9MCpI6N-<
z$5ux(N1&#_&_}9#Rl!`#rgLO-gIui^$;p~s`>aL;0F8GRZ-muM{{Qt^C0PY8<gZ^A
z%KJs$6HA|5@=Wf!#dSG9%vrSPp@qc@o}b@6Z+PAv*=1Q5@xsggdH-{{R7CwiadD1S
zH2~2(A4sa7X?!?dYBqN~Z?%KRi>#E&wKy$Q?csf6{ac&2ytC0?`UZ+sQLaMKcogo1
zmprltrxmjwU<a^j4q%?W;&d7b?UoN7W^Uq4eV%d##HgGeo5OB$hKsx3@1ip}{u5&s
zafRj(vu0rAj;TnfMQW*}F#mK-<V=Xdr7VC7-hq6CxWwc5BVH*}+bY#lp)ys<Op%48
z=*7gjEfA4*kJFC2LPOYHTKtC<bnff!XRv!gb8Pb@1!R4+Po&5RD|=VY9I3w`RiD@7
zaFtV0VTBX_XL72G|8sC9o$8yP<<NYpct;XV%@TQMy&rO2MdWZV!HWgrmBMa*RxXmr
z(%j#S*TfDUQsVN`$XUX~<QZvq6OTI};w=pv;)Q}K0qrPsI*`;fCQ?{&hA8;3?{?fZ
zOSIg_Q1G(muIAxaUcn5)MFb!B05wZz4xy2RCmLc$s-?q+Qyv*o*CM!0uwZ)RKl=AI
z3_tg##?LIJvQYDTE&#N=PXb6FEbTuoA84lp!jKs0Yye&1QYg=qJ8Q$^11QgoRiMkK
zip+w}qew*HE#b*(@B&_pkBanGLf7EE`pY7M=PvYM)T|v5JOeD4UH0PN7Y)&6-@Auj
z_AL#9%qvWRk?6n)vTUjjNBioC&_+k;nwnMuuoBviSjiH-7_CM<?YL|tR2{9a#AT(%
zpxkhpIF6@tYR*Z)bFrExUIoLOn^NB0Q7G8#ZW94&diQ@kK;`w~W1r!7U#PkJECrbm
z?nBkQB!IJ>)H)dH8E3fb;x;0W*GbJGX)q+$P4QZ-lozoTx~(pk#aM2bUgXafGbFCQ
z8%&nFwp?=$Q&T29iQMZrE+Dlwg{O|eiy?YN(Zkqb-VX*6lq*jo1SzwgE`!rMJYY8}
z#!s!sOi%w>FR+j$erXj$!{f!`{W<EH!n5euWC5t|Y|$UpQ3SQ&aXVZlYQy-b#?QR^
z@THNH-#mHpl=!8)QeVPcPa{kK;+N}oM+VyDie`icBHaVv1;p;up#iRC>_v7Lg7#LI
z6Llt+4R6y++_L)o3opDNes3lBj*7LYFjw@D21u*Q5Lqp($`$ztD9WkYNZpt`P#-5N
zbQ-pGv7~R0ysw{3=#f6DwKh7vAAiRVjKubzA%QR*jCF@N5cFt-@r3pd4<=H>EB;s^
z@Zsoxtsq)_?@kT@hZX|nHUn2!8s$vkp3w?q$zyF}vC|V!&M5xH9ph1bRnlA<JBv)Z
zxk9!VS;(hw+R_sEX4t=2)c+?$!r#4vi^Fe0jHvn0+%BPR(FF3P&f`*OMD7}sMk`~1
zdZ~x<_2IE@`QT1z{Fo%xpd^}*oJhBm_(UTbsseU0JzqhpUU?*541f#?OL55kq40RO
zo+yMO!CeSn3w;Y|5Cx(`5B{3UjKsTcXO_xCn%mD}sW2%-Doxc{r;faH*2NhuNS)YS
zo;1&x^9NmY#>a2tC|asH<m{#iqeTT#F|j`;){ucU*sYW_aEn}1Ly-#JHW<4QLPB~<
z>%*_Fc!t6I)|BuzH$2F*>gkh@Rl*8JCo<Q+DbI+4diZNXde+g0E!TZOXDq*k!+X63
z-m_W=ocF2LiAhD0fs9fV!1I%%qC~DZNle0m3G_0UNr7|m+o6zUa11Pgav5HS!o$bh
zFiv%AtEbRvb2?pV+U(iOE9lmry_uO4<e#YJfT=mLO5>1d_+ci#nP3RCd%Nv6hbKKE
zy2xz@^8X96I<oS=m-kmo{tx{Ap~VF`k1hJ@!lMiRe*VABe{Nn)_IL60Z{tUNZ8fJI
z2f(OareIzUZVpDG4&GQ2I7g92cpFMip<;r%b%W>RbH^2e(nO~i>;-#~)miAVxLoe^
z$mG&0c`G1Vxwwd3urgP3!MWINj6YDPu#>mMMC#5{DH+R$$E(RYx+249ETn2GG#^I?
zd#UPSG${AB<3aF*G*y9rRA068T;yN{_jTby$+i3N8&P;rpw!wNJxYzsQvZJFCLPBR
zD9|qjABw{4mu*nM`uqGhRss~_<DMD0tj4vu^+%k|i;O}S)Gm~QK6d?T6vD|}Be2V<
z4|{UA)O03x;6$psppe{7tzYTsm`k@`%k<!VZYGx>*I;#46EJ?2(UU+F%Ia_lh%Kpg
zr~afuk;7An$bu_fjU9Q%LQvo2;!uB1bJSUEH71A|0kRUt?pjnkBo#6S1Dy20ZQL)F
z1_G$&j|Lh+P;|SNL2_#uO_3ZQJ&C4zX%IdkJlmR-_a%k{nWd+VE|ydb(R~Fc!;DvR
zU32@4h%C9k1P&>f3Y8)9d8)<aswdT{Q5<b?+8tCJod$1QEZ@F{PW>4NJM|-)Q_sb&
zQiK~nW2)GjAQagJkUh*l<4C5N^?PY;>~%Wkdv;Dt6bMnX@GPbkCni`WEhk}7aS6uF
zT1BOUdN_IU@>dkyAa@AZhbNF!YCyjhkN|}r9YiTmb9lNt92z3VA3KbyQ<8oxl8ks{
zDip!=6|4@AGd&bt`d+~=>F$pS{O%M8QQv)5TMP)|kt%M^C|9N+7W4L~oTz3TeqN-t
z8rP(X+?x877jYC?J=pY<9*107^p)?^MeeXMT61acF{`;oogSkD05&Rkp;A&F&{hcZ
zUGSs_=?{PT&<AwDTdg=??pi@}Ky%ZIX}em$8sW8+8eO2qKDWj0u%@Tc<JMH*UJLhr
z?=G&oAt^L%o*gv_Xaf1HvbtIA+;RmD`$f@y&=;T`J=iVGZ*mDPs)o318Dz}%SLZXY
zJ}f=4i4}{z@_B}+$^tE-%+r@-K5~K8OiiL(cfkUK0wkPJ-XLDA4{Iqpd=Nu{q~H*S
z0`Uk2UIQY89}WWxMTtmz%MJUe*=-0V`+VpVBHMLx#qn5wWrU<#J-$V1X+yCh-l9XV
z&BMld2u`c(<Be!tt?0e<&)-LpLRsM6LIXZD?-TyS_z;M)y3wR42$pp0<f{`M#-~Pd
z_Jz~bE?W8igG_P0pk+gI9}iJq6I7*){kn_n&_50*Vn`W5i0rtu^7a@IK>q*ote<D)
z%gF!#=e$VXy-Np|{KFD&Zu{b$Ikk(vxA2DxZkkty7hL^6mG3^mXzg*LHMNm;tEur|
z5>L>{W5HcyLtB8=VRxbcDWh&5j^4QKD2?Iy{8tlM0S)YLF|8US{}JBTz<GqdV{OCn
z5(kzfrt*SNC_K5B+4~d}hrKU%#(fApi(Eo<Xg7HVs=%@yV=hW;A*<zbrepaTi@2-v
znO7f<T&P(IUJu;=ILFxiT5h@7lv~tEQQn^_jj|aVFr3gX(Q|@YDkj>|yAka>24_!d
zt%~%aQ?UvGnpts$@f$*+&F*oq;boVbFTFZ3VtndXAC`aSry4HN{T1$g-i|UIwP3r+
zl)1oJJl8HC5|I)h=Q2<o>&NJfMm!QbwI7ebFvZjqJc2^(@aP_%T;!29D%PD2=oyM-
ziX9fa#iX_XIio`Z-v^mhE?UcuUg^;seT6BN#j}OYjdDvjos#8CnQ_gAH71J26Sd+e
zm1ZxsEa0U-co$k+aV7Y!JXRH-7t6+ep;d@GU<4hO>ETH<UQMI^w3L9bBdun$)36u=
zU7!NQW2pggqbUxr3P51yAlh?X$g0c<5Raw?#0^(JMF0>&q08fPyNv5S5E8e18=R3P
ze)-Fss~sBFFz<2v4X8|%QJ}-ZX+tNQ8DX#f`;~N!Wsk6k(&L(2%+ln?oNT2y&z#Y4
z9!DW&e>qLm0>JtN?Y_zqSAU6v^={39K1!PyfX)PDlagPd_;Bbn;@sg-Z*<@+hQ6{`
zL5c_sPLC!f)H=J36L}C5U%vYToPI|!1Ln<|vzzxvA)=s{PWW0&p~vBL(KsYy>Gax(
zH3YOTuHittU-QLt@A2W$5@q6nl5HaK&4#*>_--2DArz^ji%&F=CiW5$v{$9NfZ#zu
z(P6clQsGPgNN^Khe3)Ngz2*YujVnsHXQmH{RpN5RqNYkxk;f|KL!AJ2BF9YC?QsoT
zOLw4|3iT)Gm5eM9w?-lXUy<@|&O-2);4l>w!{zL5H-C{4g;#Sm^N=I7KZ=A<zlt(M
z#Gw#BL@FI;qv|!52#xk1lLLEGo(;v<9t4_A3Fm*UuOPTDc!-_Nr8%2<lnrO&#gRM_
zhJ&R#28t*=HiVPS*d|=~gEcrt<%%yb^jF@e2^*V-gUI(lnH3AI0yMlV7zksjtVtT*
z-iF#6ntZ`iiUhvNnQSA)#5=m={vH%(A;`xg{IT{a5(!mjqT{DP6S477ByfxmgoVt2
zvx%>hTUsUbpo^H7G0BJ4rM$nR&}Fq*1Y<Oq*^U?gVCn-vhWr2HS^4Xhy^{CGOaEfY
zK<=+|pICfyac<7$Mc-NY!NM;rI5_WO);L~zh5c0B^D^Odb@kJP(=w~a>kzzFhs9--
ztN6u4Ek^Q7N0A$<Qz(pzb<U3<uM|1cg!ZE7z)*COGOyHrIC4gw>_pfVg@nj70sFuL
zDyc}WcOFLpvea1}9&3<#I+b2E)bIkis0ZTSOl;R-d>x8JA;##vz%MY1s$}4O*I~JL
zg6ATL##)b2>DTV2*zN$ONaI2OYJ3zvlV0C3B@gXEWDB>DrUvLAD5PSF?oxwB_GVfm
zaNr1DiNq?V$jIl8@b)>%<|7L>+KkaH<Y^@a$b>^DrHK|uD-=B7Y@Dj&wYNjZ<zREP
zyAc@$kaSF(mV!-i*CIzyEO8QvHhc@8Rf9L;JQ&R>H`V~QwDDl?S*a!@chlGxI@<t-
zDz@N5NGy(b@GLL;Pe$iM=;Z2@4mU*Gv>78uk>Dw^2$<=KX^ZK^QlRjEa`RTuY{2~z
z=p=V-70XMi=)~NfR2d)?g0JZ!GGW*NRQL?i*fGTJ$+<<b9%u*N5CMh(7!8b4S0zU>
zA+`w}tDQhEeN+<~-~#b*e+6qX7#)rFu@ZJXHI9xBvJ%klhKV->s<Y=Jdszv__+e=x
z!0{X>5J~_wYY<L|J#ZX=mnJbEXox%;bl05;#g0Hwkst-_6DpdAqG!6%OcEYz26Z4&
z5p8IRofV<fqJsx{r=7w9oJ??-IDt%eK$yw>z520WD3Af_++^|F-6*w9Q#i#7n{RuO
zX^%IcJ#3gATjutAyjHi*Tzinx#;{N1U)azx9>2qRfxzhWp-A%}hCfi|r8{X79X~kA
z46P`PcX$L-#rx`yNf?TEeS?m-nhHJC$1_qYbTAgX;<k`9B+6o|)VdoHETBc*rvnfQ
z)XL}~8{62tF_Z5f{zd2ObjF*$&OrEp=8Wd?bvo((=Ewo$6t(SF>}aTxbE9hAl}=6(
zt!E#sBgPcB{vH$HqF-}1qw*&@RYe7ge-ZnSc2`8&YoyVGXaOOvV|7Y2wIw<229k|S
z>_pKCM2*xq4S`9ctWau`kcTuB1W#25(KC?>8;BALZqP-V1f6@~8HSA)Qewlr%1Lcc
zYAR!a4GfA$P|S&5xb(#2#n1lz2XwSMo@c~NRVHd;HmX{J6yf&cObR;Dldqv<jMPR^
zrQmYA)5*NR^?CXmt?Yo$G1yUQg8G2wwMwZh^`url20@HvvASS7q<O&CZ)eDT@L2{y
zsxDF=(5Siy<gD~KGUxCl53E7h@{K{f7(`Eu0H^?fqR%-^Vw_mKoB(S18UrYGmZ{Hs
zS#?F+SI`tqrXo_*+1PAuOinTMHqE5j#nQj|F^*h$_cKgnR6e0Ovr&za7RGFyQ@n9<
zvKlnZn~`G`?Z_BNwUm*kJOfMaqtu4mYKjnvb{|KZ64##`IdD*bE)=}T6$kM$zW)VZ
zO4V)gp3M$4G0`V51!U6A#0XIk3N5HWOtTv4!-nYh39(zBCMc~HU(`Tp)@mdge#J(u
zBh!LLv+x+38;AxIHiezvNa{I>rxc$(US&1fBkep6DU8`!xI18?ti0h`SU>ueD_mQV
z&5862%KFo*Te`S!`Bu8}wd)w~Y1N#OFQBH2R1Ajhb9k1PbjvG;Rx><od5T${Iy{(L
zo)8Pa1vF6MyW8V%rzujn!4~-ZlLmRd+swn*pM0JVCaJSf5FAdEaFMYD`m{as>cdg{
zx!ubAztvg!i<bUq$-<n!ShQgN&*nL^R)eQz`w6`3W5PJN17b$)fE(Fw<SI5Fhe?Uw
z@SWxE>PY8AY;Px|qsE0qKi`ZCIlVJHS|_)$p6!HN1-F{JK-?vZ4z<dSG`}?N9d*L#
zZh%lngNuB$LOxHez}$7^&Qi3hPO3f&#+N%#N=oBL*wm3aSduVj=saWqA19v;o__aH
zs_+`<;r++d0p8z=$E72fCRi`SJ?|p{k3l675mQR)WO0b%Ha(~dZjMw}MZ|tIzRMNV
zf=tePb3=Ht8Rw53KF0$QGtL;C^tL|S6eHzg@VG02Bt(Pfa4H^`AXO7UMc6nVXPz5^
z50tLsX}QR$2C3<sbY`zKfy}pdUNaZB9-5SD_qaTs^h}+|ojZvv0{50M4O8`~2KDBK
ztf}1l1+v0w^fF&tL8->cjdr&YZmB1WxxiKJ@J0btz?iFNjGu*ji?9Uh=V4Z%oj6i;
zf&0984cu6{YZMc$d0jzdd>Sq=1`Q!}L6v!fVsx+xFQb4Bi_M@Tg!&}VUBOBYIME><
zgU$(xuk@kKo|+HwjWB)Wy+ahp?mo_E2#}wV*dodO(#T0M(lt!L7(W)B8pE0)e3!;(
z2BK&uAR2DdE0>LI$>{>3s#bX!k5;p01Xb|E$gBs2My5lcAgQ?>ZR8-m$W(o#eE^Fs
zrKXYSB<slsREPFau&2IV-d7{_V-XS}YC91m0r8SIHc?5j1;$vI@x#-^NyrsjBF93A
zY$5KUpqR*<#jc`g`QDr0@12OW9hNKllz>(^G##GoP)D=kp)BU}<IHF#2Ry;qN^73N
z)5OgqXNav)65c094xf_`PDe*DrnVjtJ_O`(9%+hNaH9!|qT@9UQ^h~)79yXgwgxYt
z=>0g2+9s)1e_HB6qWmBzfmR0Y!{}q$Fv5w4BI<+G<WJV*iE)q(1&bzn<%v^hyJ40;
z;w?=u4SA>!FTk2e5E3Xk5;{mu0c-S4NhELt?9|hmmQ@}3A3r6lI`Gu%jGKekX}H<k
zwfeZZnS|*_U}zf4*)@uU#wWqk#1;?_!MfNU1aMDCokMcnDZ)04-VX3Y-N(fAu)^#x
zJkzv0qt{ihB&IK1&zYX}xu}@l$P?vEAKM?4&vnGmf&->My$32s?mP|hSL<e}b{9EH
z{d~N49gnTy^*koZ0|IJ=N{l!+q0U3$TzJQ0lZZ+h#pS`TV-2yk;|R@@tZE$u(s2xF
zN|b`ts9r(Lo$E+C0CagmesjIS*qs8Opf@$6Q}U*zHr{tZ$+31dmuob)!r!z56cySk
z=BAIW?B<ys42OXFr?U<2f)3$q&4co!QdsL9sfZBjYnrG@k2OURNJl%Pz0+J{#13Me
zs5VashY?l9PRy8bSR83;iAWw1OrZERuFk)s5dh|)fH?~QWID)RVnJ|95MY3gDhM8)
z7>>2K$me$>$HB!Skj4$UBz=%D2zVe{+F1+cQ|ku2E2Zuyd+AJw!<D2t>pi1{NFu)j
zbnzqzMYj{7?bu1=`$W#x^sccBpl%ch#SH>_%LQtyPZ$mOcA^MaXbD<V+X|eb_3c+2
zhY;dcr1H|K3`fCU)l*%>^hK|7rhiX^L!;0L>8*q{uGrzx!G5Tp$i!|rcmxJ5R2md2
zi8q!GQ5gNBvpB_Fm~p07<oNu^F;SXE5}I~y$9Wfh>idI~IAOwoVUSX;Gh?Ng;e|n2
zps`4EN30_VqlCiORQlN;>8*fK!kq<(E<#gcVpe=jH%s}^d5(8S&S6zE6>;EO;&uk*
zt;VI=5XE@v#}LcGNg|!pNs5mG47o9{*7tg4A8qI*k$~TAT{soyM5BI8QS&$kr%8|&
zhlbV*z(v~j!lK1D>dZv6-(m*2U{Z=>?7PAX?7gj(1i(A5a0}oK4e^*qjAj!6kXc|*
zg=%QtdJl4;l+pySk$xl$C)9&lNI*hRx@~Sk0-vvz6Ri7roa83V%f<jx&yAm~%s7T&
zXYq{KZ%}NYS=oh#wX`O=0Q1LlRs!>s|6iQ#%F4Io6)$O8d~DGN3nKHPcp&p*27y4~
zTTI#pZ=k{!m5F6kMMl!rFilP~ZYgkZ7IVN3kZsx70k<`BvJLZJAX&Ip<0ck2lTsU6
zDm#ZPit!G`e4?_07PxU#)*ubk!f=96V`UA@3gX#Pq<^b-k$hyOYh2Iq>K3Y|XU(X*
z`xKuOIekoO86jtv4GW0ZF>qO#@Udw7dFyF%nkTWLs2?u|5P0B&Jm9Je+@NCGg%t;|
z4qixJUIc?iQV8_$*s)mmdBo&X$v#B%TsBOtG4_5?BmLM9yMW2ScfZLbT=hK~gqfGl
zrBsrnxS0PNsT|{BDj1Xi3eP11e^Lk4Wh4^MbqLZ#D*|{HQ;Vg8$Xmw2CpYhsYTIDQ
zK$}59!F@(ML#zT8b_e%ENehiOU#h5=4hO;$gK|S14TQoX)Z>_a+rV@5GyrQf04>d}
zVVYQZs0zi~G%lz$guY<_j&704MMY^6qp(K#+>m^rM>={+DK?1qRU&K_IfX(jDkR{d
zo8hK5+FlVCrzm~eL(?(2NSEN$_k}K2?7=_ox{@x%^u6TR%Z%O3dH_IM+)%<zspzJy
z4;2U?O(paNWgcl90QKlpAM&19Jm#<%r<BlV_S3&*=zn_yw~n;vH!5J9g<m^qWs%gf
zp9Y2`Y$T@bu%5OTP6HrC`IAI`8Hk7aCd3~;7de7>W_`SX1F=wEQ4{B<Mk@PYWI>}J
zCR%1if9yzA^mql23WD+!rw8X(Ga1+pK`tUnqktzD7?3^t=$!K8fPAKvSBJzzj3PV9
zCxpat30h~i_l(mv+7u22@sGp?i&g*vG`cxq(kN~X56p2f1G5Kan_@64rU0E5Ujajb
zCV;0vJZv7T8x!6QqaMf_|A<X64LtHDw`>Fr-<ubH68agJbBC96InES4S_xAOcT-0L
z7_XnEiIj;rHD*!55)F@{ej*;!#CsFfVA@hG9*kCpvuR#QKW*D1HtK8Q+{P^?bAsSu
zVtf~IQYUdLkBxz!5I>9Uo#r9|vH=nzIyL2X-NVHLG5UGP^G12^Fa{0KfY25U6sv>p
zjB+<#^~3!id^nIw*kxyvf@ctUC#<sywxp`KMlllVmyR5gCdTnCI`&1<9qA^ANks)}
zxJu44_APybvv0Nri$*oBYWAflC2H9cA{vE$I6c=x&`eOeimb-nX4SIv$tqc@Hi7AL
z64op}AE@G9EIu}Fpt7{X4D1-15!*x&9IRH(TU_$Tv7%Cw{J<O~3dit|f~JrAm&s*`
zKrpngbZ!VqdPOY9r;f6+M-$CrLP<CqCkvi-h^nduQ=*x?Te~@$!>=4*XttJdW-ZX5
z*}UM3P{PO-sxe-;whu~$O$t;6?KJ4&IipJzVp6Cfo){8ml0IPE{liH5C=dj=)!79w
zD`L2a1tSEE`_HJiL>vG}LN6i(8JOq8Q3ZdKcQ;T-T2)FKw~{m!o*33N#s(av6s^oT
zl?JdSopwNSj#rSxc{e&e867>zsgOs~iH{R!*0Clk>oHv$1L{l!Y<TI3m02*XHu@QB
z24B^<jz(2*z*aca46ROM3cD~up`Y^;z!U+D5J#UHh_qvrC8ZP;iK=ESwW!Qw{%G3g
zoHBm`^52;FYc<xIrjdf1-(f8JR4HdsZLYw)pinI)e=s7A2BOVpP;wfcXu@m^Wd3-h
zfO7<JuCKM#<DX+5MyE6a>+te2zj!^m8F_|R;q*9JvGgv_>T*^ED{&7@9pmTrj3TvB
z=Z}e3;Si>Mkls!(T|6i-Yp#)KUlm=98%sPvc(gLsegL1TXHpDk!~if`diaz#9tHO@
z|34>N%E~XwTfd}p@f7_3|G<A{em+eQs9v#+mF3i6Um`{ysKOQI&D;>xNCs102=xy=
z!(3*AR(T{t5~(_Z@|U4WxxN=e%3)(cY|q8`Rog>qan3DYd<D08&8=OiFuk3tSh$LL
zt9ZHOpEp)w%$$6vEiPnsPSVv5*6{IS8ec4O`ZSf}V>T}z*fV~L8<DZyomjjZIno@d
zL(O0anjOc8o<9|BIV<lQl<W56H$||;Q?R>98XRp#IEQP0f}Rp4+1(^J_fZ`9JQas<
zA)_`e)Uk1{)^TO2rwNt%ijaXRrUW13rFN~z$7`G>yQRoxMXDGZppCs{WAXoKdj?^l
zvn&3NmIl7|EpEN+&{&*Ct@C7Sn#A!8;sOZmP^P$=@(OAy<fM-X3MWUZiNHKtHl~?7
zB1A6zA!rxSD3}3Wn>>0VB2GjCePmE3T6YwOfFenq<MKcQqGzBLHKo#52QNg%$5_2X
zb4q%}TX<wolR_n5k5?jy7ZAAPrZ7nd0|$j#q*$7br!Q&R15yF30SRNT00GL$^-*UO
zI0uI$ZW%v@RVx(u!5LT*h)?v>X+Ds$6`gNxkISSl2!WsMN7q1x2VUOF?G>K}!R7-7
z2+fHoANL>8&!p#ml6ESg!oYVLxRin;(hx=u?pESvQq2YOP9~?Z#2YV@MtboYVALE9
z+z<eoqv9ToAmQ~zGqPlCA5N(31Ls0Tt{XxGg+vQqHGxvN!>kPv#c{hkA-(9pj3Cs`
z4MZAZ95`E~assI;ECxql5U-5w31QhdcCo=XM4SnPf=?zQNcOSd0P#5rB+uXi_nd*}
zfw#wn(@q+2hBzm`R4PGb#hfOLWONy)LlGCa;~zM7w{Z5|qQS1w)C9t-F6n8^&|2)O
zwFdYp(oO>>^-@}4&6FC1$J7B46RnYmuQ$r=ZOEA+_@;0V1(By|HpdKl(8N1(UZD>>
zMUlzIeLOmcD=RrlpoBY;nwYFQ%)=std$3@H!iimbA}22JJgHii#Dgh}Q1HDoJ}=+b
z-qeCVXMAQV>{?J~?6C+AQ|vyF{{ivtn$4VdAJkyid=!M5cX@<fK8w(&He{z^Tn%bd
z+61_&f>F#WMS%yZ1%VIbs;Q)lcLL955CA=DP{w(P3DQIqu1;++Xc=8&G&J<0k4$PB
zk`GSthEHG%#V5Q!QxU?v!9hMc0MkbvZHrY<F{5<28J&wnH^}azdTOqg1QL%HEHQSO
zMF*tMS-(#x{Kh8Epc^zOG#X+-$eWvf4UhIlD+4@CfQY*Wo$6SNT!n$dJISIPBRh-a
zb88b9M+|>8m<G+YBd|#JEE4`qYT?BIilSUO_@Yx|kqI^%r)wv|LOezeLl{+sNK+vL
z!qEW#l4Fr%AS-l4He^O-rc#_FFcYz*ct(vi?;^&pmH~9r3Yt3{M?A~karE;8D-m0;
zGLOztPHJ#WnhMfageIfIN0D>Ei<5myXjW{<fMRuejGHkadw=SEolyJ8M$Vnu$h!Gt
zk2!Lurr;!H>3{~2gaef_1}@{f0$7PKGF4zx6`IKg;ieYJ2AKMzROt~pJ4KnDo-@E1
z0EA>IdY19HqSBO74M=ZwDwVott114*YtZmsx~VEs3?T_(RYavyB|?T_$1vJbr6a6=
z7@@yo2e<%-{FDdxV6FsC$cYK*RvwL{FJj?Dos30M7!3#a2u_=659?iP2#b{eUzq*t
ztYr@@3FP!H_$&OG`T5;|K=2#yP*h-b^>x~)z#Rt7Iz-Vr>xq|)sk5L)-VSpEkTqpl
zV;1Byad^0GmD=Iam3DKtCp`tsMS|iNndC&kb8*jvWoze>bXA%>^zPHF1|1Ft_erj-
zg2%<BODc(tsix=zbz6)j_NUExH@**wA^+<iky{^p@@*FI3U1Mo_U*GoqT}5UJSa>U
zP-TJ91A$r;jn79T1wybIC}yotYj~oEs}@9;@FAQ$$xz&TU{<3aNLHJFbOwWfkv`C-
zNiioy*;uj>pEhe`4P=f(YSPF0u$rRY)LM;Lyzf}LH61WO&?am_HFH{9hl~m1>U;*|
z{Ev@SW1MyH<sDp7->fBPqqfSFlG@m?rc|58hRz6UNfqn*n1nUjyIlfCH5hwNS0|;W
zb83vBetYsUqV9XPb76g#mb%RcGn>%3QneEgRHQ1&qd1CTG2Toi9wa6_I3ufo%SwF#
zo>F?3W@Q~S?UkS$Hy$)G8(d!>dE<pRA6W{cNfZN--D&L-Nrx7$<o7@d`vVrZ1i@#^
zxfIUP(y7r9YogFZFhgoNk!WddgojBZYt@B=$pMUv@cjgYFB)ncHWzlh7#6i$$z=HP
zo()8OzVx|x|Np{yk7X@;c1d;4e_rr&{F(X5K;SYV5M1?LuEKBCipINV^S8KwOek!1
znL$$WC53~fa;+#@Z<>YY`IM{3wbH2e%Zkujix8o=dN6@KBkKHEbrLDt`y;G63BK?h
zuD_RQiFB1oyepx<l`Js%^E}P3(%-+!Auml|G*S$oP_PShZ7Z}0R_qqYsBqA={9iuz
zIs|oa&9}Mm{koQ}@0wNkHcTUICGVNS{s#==35QtTcc2@S4Ir11L4kk6D+M%BFP<+^
z${Z*%8t6mRi}$H2RS!HMrcODB$NP|pAwkRv5V(KFrL&2f#ZO`qI+F$(r6AR!MQf-C
zJC$oV)5DD72#hDuHa4*W3SKusH7dNQ8f)e2XsC=jL9VXUgx29jWAS_|in&>SmZx*@
zMqP%@;X-wTU^eXi<F)jz<WGZW{`{#66asksJtmqlN=j3pxY9&4&z6c%X9F`H)hFOl
z(cn=FmJg;gOX^Cd%XH2LTD~*%=kTb454_98@q=1gzH?S_oG3X{8cWI8NG`8Z6ZP+!
zQLGt!epyDbr3n3I9;?%l)|lLMDB?<f4~&POw;Y4y<@5iyWi2br+qd*{OB!=yxuuK6
zoQXwe7j9WlH~+DD^Rt_>{#Vwv%XA1$UjCk#my5LXowN|fTITm-tGCZn>US2<2=eAL
zvUq~me(*doSr)yASPceEUDs7R*s;G2OSWAeo6~r5C2CgWAJLJ-Tg9bRv~s7oRP*im
z@{u6=HKWbVRPm#1m>aE7Pw8f{;Cj?qGG}u?U(<>K)KofxR^R?My~1bf<uJnD>NcLx
zgX8V`pReP1<>C?snuj!pTc&^p#x>=H(N&96qK8kUzVnpdicR(wq|wY;<1ekYtiiWd
zJ}l-kpakwj3zr%tc{4z18lZ2it3XjMpnzT7l$2yiPX?3<IucrlC!3<fSk~9VD{nBW
z8|6{)O)i5Iu+5Z$8x+>_w@BlC5NT2u?ePkaV!BGa_KIgt@fCO(ie^yL#Ib^xVSYCX
zxG*&xUrncCGr?&b4{GolgHkJ|__EEs7=REw7u$nwGTIP^7KN06a$rR}CcdXr77s4U
zJy^j(xHvTkmnx`>uj=~LfzVmxz?w5FHo~M6Z2$;wC|Ls#iX~zWXH6OhucpE!GgJs<
zvq41gDELU?_&ohkMR=?>Uh<bfHj(KbO!lLB3wT+yd4wp5Gc*SMVJN|58>Tj+|12`q
zBc1I~6PBJ(oJDp!<}tdQY5ATgN8h+}h!e`9)P$1zJ3uJtKy(XMt1}(<9YXT001-;%
zB5@%DVdaAwLRp+dD4I<Rv!R@cF(%tvS|TBo{E-OZ!U+lp$wEHB0gf_>v91xm@f(DQ
zz9$Gp$RTQa7zjj14x*|Fa+=1UOw}T>o{E^TyU2}^UlwDfU~D~Lw!z@~m4Jj;EH2<I
zbGHT(IWxqRV3}lop~b13U$|ssw{OBdFjbXG6dT!yOLX<`K8es48x5fCD00%OTaO*3
zN9l>qixVZ!&?VN3^Z6y#Yc8=!jZS4!KCv|u^*|2{0`N4pLBv8k(L#cH7}WFfg%B~U
zKxv)Ci<O;Xv@rVgS&Y_^2b<(B?2tGh4^L484pmZ)<BcitWG%E|G`TgM>MAU{_>WtD
zNmtAk=P|v0kLHF8XVAp7DejZ%5LQ>A)9SRD)LWYo%carRaE2`LPBEJuf`&q>#p?nE
z%*tu7)doQt0$wGW)8qTh$lMql2DJ1x81}MS3sJP=OmiW~rSkuugrj(yn8l7r8wu1$
zoUfeB#5BY~L8>WjL}!r&THY#H)9jVK^xQCku=0)XaoN2^bG~^B2;)j#x0n>Q$<g71
zxP}xQVja&k<Os3>;8dHpictkJ0DW`fmd}siHK5m6#U5#@N*YJ)ew!2TjP>t{HqoSL
zniYe|&qLL$xr&D45kE5vOPXklwjQT19z7^^Oh%d;Ah^)G5Ixcl3yZfs#@ewNYJUWS
zKSOK-9J9H&H3xA=MbUL9xG5Jes2;E8Hl3bW9aMYlv0#}-dK`1{M05>N!VT|(5OVLN
z)m~~sn4tp<$sekgh7Lvt_hPtJc#1l$X%5EtbY%EQ%5y<j2{yYuomj=I4}b8@4mwxx
z>ECB@s_Ja5Z~qzNI2CW9<r7tb7e3@5GTay2BO<B_BFH2Ul;I|yWby!$5*=1IT3|ED
z14<tJ+o4C|`TuLPew<ZMQ*cdwWq#hWck=!+Z~oHXTk@B=n-(uzRJ$;H{^5E5Jnxz8
zYP{$w_z`X59V9x|lA3c0?owB&T`03z%S>Z$+?(d+kBN%W$Q7s+D;2C_@ho<>q(N1%
z<l>9Y9usdT)v{Jxh-2H_E*ttE{es))G9SQ+b92Lp6bC%#Rt!w3fD=wZLe?V0MSFan
zL~pHprZRdSlL7Uc&m10F>g&N~fHZBLi@9f*4!=vhjdVC)il(}V;U*UD(6OcfjW*O7
z8gxh-3=J0mo))-23=X$~h{b!wTN#nC&26efniI)op437cT%9)Ku#}nA7AK6#*Th?x
z&G8T^9JMOEUD1U^0Z8JZf=soHNJFX)V7CaVCmU){g>>w7<Fj6!&!GCWzdny90L&L}
zCY^pKSWg4u+h)u<W7PB2t>8ayOM(C55v?P;c@Li^1L(%oZcMMk%A<6=;HwkZ#-|1v
ze*6x?&}EhHi8nDqFV|w>RwF{kk~B~`d8>hQTy#@Q{p9XQs09rOsb+112W&6`uxv_5
z$MH74#?|=@fPd@RdkOfZ;tB@*$F;z}WiBkBlhc5+#Zu^Yx;<%`+gB$vjZaNeNdN8}
zj4i<V#?)}WSt0Vdn;($rF4`S$5$v9fO!rZxmvm(Q2~>K$FWx{l$eqt=@+dbc)V#_B
z%&St$ZC4R?!a3}M%Vj)o3-+A#JFk<0ald#yL$r9G<_IekM9<*}6h&;GqL3ULq)RbI
zgtI$_snc0(hqZ(uF^J#k`45M+7TT>SGck6<5m3GV-`HW-iP!PNeoJ%M8x?pZ8JbA)
zAbQ?6L=m<Aqv0`(KVrjXu&Ab{8-X+>$wM0idTE?2j8SQnI^k%P#AFlRC{^r|T4|LP
z-pj*HSd+tpb_l)dtru4jR-?G=PLD~WBf$UVN_L|=#B2GDuF>4+1_k_58x1if?U4v@
zDZu12ttsBc#->Q40}>X}vyED&Fh`sY)oE_?(H1ZLYZnp7HR3glMBmdK@cJ2$rwR~V
zuC)QM*$c6B(_u~si+}Qc9I9Mg&d#JwKVCNjw7Hp@_*U$(#H2x+v6=<44*&FZ0>x6X
zfPL*I4JfYFfWjytOl*OORNqk^0!m3cwnD2D{qV+?FOJjkt1bqF{FFerMgan7hC5xb
zGQfI#!)z)ZM8Lu0vU%*LJox<N8hq>9;xhKFk7!`9d<F&yS1+F296ibDHRW|JAVjCr
zVJ4|Q{G-V0_+rZcUz_#otb)7qA6vE|uW{+%lK(IFpL1_oygKJKBwBhF{PFx>&A(^f
z^VtEs=+pHhF1$A99>2A?I7{tUy}as`9aySR^2!dkP-1;$htuNptb%UGUv9VKl^u4!
zP)g6ap>K#J_|kOb5Oz*TdwMj+K2)@T@PN%>^Jt!ldt-RUF0Q*~@jc=WI*58SA05Qy
zFSFw~9;+9>Z8i^my$63=EF~^H;<pHiWAY>0%PcsYqpZZmkL+;@W%PbKy;^Y4r}(V&
ziFifgQ#>xeGk(okFGP#oQCe=n9$$y%Le?URuh}pb#wD&>p16>DukaOlAsb>i7>jDL
z+3d*|!UA@-3f}E>I>mJbi3{1?9>33r0V$C5SK@`NOi{wi2gM~{NcD_IT$i7?kk#uG
ztfhjztn3QCkiE#_!iWPWCf8~%H1mv8d}-O@dn%VPX7;#!4*c_!S$+!`Quxg6FPU9H
zSW%T`Ep*!LXtvZ`PI<;9h%e<OE+@E3d``c=v`iq+>NhPQF7I*{SfL0hqBhHINxq!w
z8JGCd(!}M;oPMj-;&PUHyuK^@a(0M8$|<@X7Sx@oFQ<OS1+rX{xSZEkYAvysILhp_
zdi6@ioV`e}6*+7ay`<eL7MtTii~B(<7Rch6_!?!G+iDSC%1vAj)jtkziN)$eviwTF
zoQ*ln9u$39U7E{b`~<=lE{DhA5no!IxSY*V;`7<<euv#va>ZW`m_wN>{60ZQzMRcb
zWC1t2AxOb<Im+c=8#o0&Y-0-o)mJQ83eQ`y>(Gvf2R8mV9mwP-$dW(fDnczfZIiG=
zS}saluFNe!vmsjmWB7`{90WcZlpW~0pAjhha=<BFTd+HExrK?#`F&2oZuJ6nwkw(E
za0MU-<A>d-q!BErV^LvF{fx~fzO*25IlILrcuPu3eNLPA%Ela|3H>Xy*qw+@Bwudk
z8J9zRX@25zg4g2rdwj^Wd9P%F!Z0VWx_LLoJqj}hvgR2(RKdK&<!oLXybg!W?Xh0r
z0)--aMJP$c41EXe<s`D4dB$oLU&>Bg&TjP!R-dKB=JQ_R0tIdWb1|J(w+?eN&)6O0
z|1Zs1m{suk{MVK><xMR8`O;gKJf8dR;=Y`}!o!)L3<NR|$Uq<ifeZvby$}%B-JS@C
zxf~XsyVO_a@>4DIZ`vQ4En<W&6f~4kAEOO>PZg5|t~t7JXqR}Q;ISasBCfkFaUn~Y
z&FgSEt$v%|e<fZB<BBnu2@~N_WS@MY_*E?~m*5uH-I}<N)9&>-omfxi^C7Wv<za|l
z$n8Ou+KiXVFr<105!hQ27eWuD6X91+iQkL4fLGv!1PW4OXg}u3Bx4BCAO!ufecVk!
z|C<vRa+G+>N*svdIqay4yaF$TIp>tBb)gqnb0L0JxH~Rj=%&Plyh2$?nXAO*^jd!l
zX8-I$!N%$$t!@_bPxvvM<l<KqEQtE7Tama>X^Bm+I<12Dx8h#Uejy7SGmOe~xsg3h
z#*hVq&RXcQz-bf2bvGt1<gohPZl6c+xXP|blEZ~SaxMhhv?Le5suMH^v3*10Lh91x
zE0Kq&^4Wp5B!?U2vq>Zuzp5LVGKaYC`ox9QMa@^<h0wQx!c6EP8x~|GiEZj=DYB!*
z96v;I*Cj6G!2)8Z+wXH(ZC9XT#j%kdvle=suoE<yK-DuI^8c^T+M8AI^0G_IzL<AB
z?}nw{Tk?w~U(G$0o0T)T=wBDzvaouAyx@`fHS-tEyB9C~P5cCwi8iX=SzEOh^$=>_
z&NZ*l<~0)7D9)OJ!V_Jwwh*iGspjpE7~Lb)Qfnic$7PWUfRz0T@d|oWRyEaQl?PRL
z^oe}yBwo)4Yx9mtX?Ih2oUNQ}8kSlx;d2V*Sy-on+9s6#;+1mm1U@-BO`DGBD_CuQ
zq_Y7Z7%$#H1=aYe$cbt6Ek^6kVlQ*_NE5y#vM1hF8V>ElgV_2L9qh)J;;CrWsMNoQ
zKCOWs#|7p3`e=8BJko%FcmX=Kq}mGhy_3=N=!Iu}0KMn1yGdWqhUAs4$c7OC4ktPb
z4NK`!A{LYh;o>Wzl@PL8{FVkGiw*jol!7tbSW|1q*9ocH-q@+JX`XflVbc{VuK-SZ
zBVS66bzz~SyaN|2)WXvTsh<G@pL!?h-bcmA9_=m@em^AkJ$8?Iz24va%NpG8ebGYY
zH)|iz+%CtI+)$p!OYV6kvQ7-5V$E*xKrNdA_&@%d0oW~mo&)%P%@GzE16bbQ965j$
z<JEv46(x<FMd`E5dJpj+R&+_@dx&(-SIPSh5a~2y`BwuP1)_t!(MfQ^5Z;AXYD$P0
z@T~4a*w}(ePd_TAp8ogOad>o=7Bc8j0hgLt7n(v(Llw#Na9Ano0)(ZW_LN729|X1O
zH16jD{d!ifq$A!YK1fGgU9LIe0t4t0Y8!)c0~SD1A2{__T);fadg`AX=pmqD?0@s1
z3{J!rnQm=A4F@5%)PcS1#>2i*7;O2y@6-8~i=Sg4d|Gq9S=)cY6YaG9w>A{tXGHsb
zE~Kms<7cu{?}5W%+<O3+y>}~v*(>4$sd?V8l}G6{Z|>y`4A2?PPXY1Lm^eUkFevTX
z#RtcMNNKb~ODpyrfdYuGN1L#kN@@<t%}3DL$z?1b7?P4EuxuO&cFKrt^S-hEt<78B
z+2}8Q!(UkL-@M(wnH33N{QfHdd6sAu{}b+)yH>nL19Wq;MYSPX&Fh+QN?elF07BLu
z;Jvz#*Ydf&sC8A^b#cQOD~yfQjbYV$%0pWU1qT)ynjYFGufd@!pAzq9NMW;D)JCGw
zx^2|Ca`}%ls&~J0!d8fd!~@XRF$EFAs~ZUy(1xRXad4VtfEHq?m{Ho#D7W=udY)2?
zO3G;9JhsDP0uF}?K0Mw(+*A+7=5@VzqttvDyDaIGTL$GoCw<g@(!!XVz@--&nB_rH
zZm#1sxHO}^=Zt}fB7@p(bGeMmE6qysFScG_{c87#pJkj!tzc?+n(wnV7hXC|V8Ng?
z8`6WuqQLB^msW1+B7(hDypIX7hcriBVj#r0&BL*;{#$=MXi7*@lR{Oq<6ZRrVmTe}
zF7aLlN$v5>Z8l;Z^+cm%wAoD0C01%Ici52MH!f;5J!JK5IAfNWC*H$gNTZI_=r*)t
z&UV9oOL-nwp~GW=|6?d>OwZ$+Waqh7{0sw{_B`g62ghd0Q=uryZgtqxJJpBbZ&VPJ
z3dFk^lxSdu8cK%72!m2eI}?br7uu}gQ)4}C3aA^tvz3l?t+*;Bpk^^L4ABRd&gnpj
zdU}nd7oU-SOsBd_yo&)!D?N<tip!E4R+>2^pfokDJq)nVHvGd%9I|q|xRP|n+JLqv
zdRFT~nFB+LJh8x-i)PFMf#AmDn0$s^L43C{>`Ck#;`Td<`_`&RUsJMW*BG~}CSg}K
z#SWkuxjA|PyAa|VE9qTWeo=oM-y)sa%STyo1&YoRQW$JIGT+aL*F^RWMjB{8UPL?}
z+1nU7*-M53|B%s9I$;Q_A?qV&LYPE`MT*hm6>@EBr1LOdV`L@)?r8O4G4}L$+wEZ_
zdzW};O5!lIpNzXqHSGnk<yhxA>P<XCniTwS*)t#x^8fR){xYlJtNCO3*1QW#ugyKR
z*s<upEqrADAJ6+5p12x+D&G~q$S9umGpd7~=DO&zxYI#HZ|oLpW}|xCtiJJ13y7t1
zoA?m3P<603RI!&s_9y{_XNvv7RNQS~;v1`zPunvXF?D8SA9e7)D}Et0#m{0$TpsLc
zY#FuSWnfR;rXT$E=yT}m4djTc8O5{Kb+xoHG?(CMwO7h*7Z4+)*3QwO+}qBZ*WnNG
zhBAzpGf`2j^}&OBN?SVK&!><LTAyhp;o}pF7#rM4jqGYv%wi_)9|W-V_B7Ig@Xcu$
zO9sqD@gH^xGD(|lq9?VC|5)r^fkM-8^-w~Zk?{oex1NZhs~yDcPEFkAc07-hry_3j
z2D@wyCsqDuRO!PXzTDYK17TK*E;i&%yjv?o3=MbWou+ETW36%(;_vV`Ff8-{7vNIg
z9=UEm`5}|l2(ZORmRBB+gbvW`$wZ7k5~_~YSEj-dLPrkN`J43ee%hqD@qNGw&p)x7
z25Gz_I>8FLYjsAZY3vs6DfFw<dQ7gGR#26|P||!;7A!}Iz%3wmnvR2Ie2J^`8GwG*
zFJFg>6&HvO=G9WGx;ooq6o3ZR^MG<Z;ym7A<ahI;4=2j;jiXmM2X9;#mx*>35TZHq
z>T`Uofm^{==%(zsao`p?obu}dmn=-46PQvIZ`Yj2Fh0wNswWQ%MV=Va0`ddXEIFWJ
zfk{5Vje;-(H1<Q94$@9=aZlu2%FHJkmpv9V0Gplq!_i-@!KumrU!J`pD}P7c|5++7
zIgQ@`I~SMajO5&~=t~RVUvPZ>|DInouLCdm-TDb+iR+kvd6{&8x}eQez<}0~kVBe9
zsy~4RN*HW{IW1D{LAk#dXp0UWiM0*rggjboPytEP1x(U%SEtlS=hQQj=2r13CTS?4
zpf15Pk~B1rrmCHffaWu0QbnWzZdUz=y%WZj55@<_@yfro65Jp<#V45ndP8$Oa{+`y
zhDY}yRH+QUVa7ohw+uifiU+8R`O(Skv>qKnzceg1oc!O`QHVygh)*!UYENz?s6OJE
zEoiYqv0*wc6wdr73y0{;&x&7V#xJC6GBuco^l)^HNtQbYc}`XuYLWU{0dd6^qX>3r
z61xDT?nBbh9-i7{%zBKBGFCgGV_6hjqRrOnbX%?IxF`tCKNjhtTdffv=Q@MteyVBQ
zC|G8M3PiWVW6eA-B-K<)JvEALnW+EJ7&vqX2b)AFq(@B`d-qk)9iA4y!ZEW#a|iR#
z6K<m6){g}(NH*E6mbB)f0WH6(d!4Ryhq#t8$RnB?83lAmT~D+ll!M<$7t;DBR0q^G
z=~$S6ZbOwJ3a`>*hl_i5F?P65e2fVnN<69QW|la|WrX|qQEixDl$KfoFc&<|fawt*
zWx#w-b5z4vLGrE{s-t6BJSyIU`4ie59iXB8@qt{S-SVN%lnjc&8pyb(weA^*-*PWI
z{IlYhQ;QvQpPOeAa7uOZi>fH#^EFdx8&60P4~(3VN9&;s@M3kLNGjC@l)xuJfeQpr
zS_<HzEt^q~AU?uSvPpvyGY1z1k%Plj%ZTy9tgMiaghxFGui(M3H{|9}sgsK6(ID+c
zg^SbJus74V7hhTWOJd8fiC+RTP_d)c0p>;-Bl=GQ7w8=ptPYbQXQr3pqn{B=D`qM{
z=gJzHIhU2CikHh%KYbb$V|@Or^BI^E|9#&FAf?K0i)$EPKc$WFm|JgVLmNT&t2CuG
zSd4)x5sM4+r83Z`NaoN!lB&p<Jj`fbmoYK2Kw++Z#Ec0>v0(@xeln==u~7cs{u#+s
z^8a(Q*Jb6uoA*ykpU(aB#rcbNE%@K_U&TY2pUZ|o)l%{6EOuP^o;Ge^NZF}^Kx7Fs
zt%cuG#e_7~hD~<zx#Lh`<Q}quWIhg)N;DHIUZ^-3D_-!L?pnT{ikIg|EvG09(bFOi
zV<TR}4@RQRr(lbO$J?X3sJ9?Ibqqx_+=GysMx<bGynb`ML%kXXuZ-;<jt({GtS!uo
zf`YeW=#VFqN?^>j{7;{xRKj)Q?=da&JuNkuJFPs7M5x_)JUqS!WybUe9M!RgSlh7N
z+8OJprVPWrG3=rwe5cxO=%g`THn>9=`IJ%qT;^r{)hmRSTf`T5hG&HqFGg;++UDYM
zY_tT#cGXIQW05oG<b8uuQ#bNF($GY_PP+dXG1h2rtfmRvs9)^Gu!A_1152~giXmK`
z2AJ`wVH_-&ypfl(JkOYh8N}+)t9d*@jUvEa?Re=t5&>1I7K{YOXq<H1aN}ECozDRF
z-}{%>q3r_2;&Yrc{94X13Vra6XVJb&Y49)v6_-X*^?4r5i4OFHN5?5VlITj2$EoBK
z-2tc!Ro2vmLnoz)7Q6+G0LmT>BtW=BEY)|!YFhMK2PKd;cwks^n?WMeItl|lo0@?>
zzTglT$m%5Fg{^uS&xSmkFRcDkUOezMRq=pYdKrbgz#Q?afI}iSTSkYD%fV)_k<?Qa
zX=SZ_Xn_h$BB>M((b`Vh^E7xk+J_byS~*Ihu7g$_r{WK!3p`kNlBuI%>i>hle%cNZ
z$QGYr{GrS3naA|73zaWI1>hmqqJ!0Zn}IaB8;M%BeydHdX)H9az=JlxObHpc!8Y%E
zoRF|Ue43E3dXrYs7{(wC!g$PA!Y*>wh`^dOQRbm%2F!LvxDJSLq?7f#I`z8osZqRI
zv-3sZf&BlRtk<*htMdM3i7V&Z3r4d47619M{RD0lUxPqI%$BX!Qcokg{GjcOX)t;+
zZac80rA%5j$-|*kL36u>R22$&oEb65Igl6f1U=%bTva_)Tzh+tdJ2(|s^TU#X`J3e
z*ceMvFLVJ7yGVSYM;qe}z>Bd1J+Rt1rqq2qQL&wpW3~IJjYlba>+O&9BL7O(pPpso
z(}k>_w0i!Gi9hp<pVmWMg(aWJ%Z*kYhj0o$EWSd-yH?Z@uelQc4U@2o&1A#s4V!Vt
z{nd$z%&F63B9L&yrCexU$lsR@<nLvfK%boaMfOT4L<-&M$Yih5L-wbQN#N%We&-Sj
zu!CR3-VSEg>E!R_n{}TY1!EhB4S5MHE6UVxXDFEWseZWTvuy8gj`$*TKv@o29jGy?
zyMUjMs%swC<Rly4I=dUI*gBAx(?p>425o%0q+oMr$T2x!`Hj>8OY>q34ZPJ(i9Qa4
zrZS8w6;QF6@&6g@n}2d01Ub+D|9;l8ALZS)q$c;8oW4bc3(DsGQ}*wFY$1Tm2N^*C
zwO?;A8N^au>ez^3wZLT*aN1Gw;w(hrm)m3n*{97yF}@E%USIj^n}CYSEn*oHH!cc2
zs-Zroto+a<@=<&ztWLrM<jiD7PB-K4k(9ak?>}`SryKaP=;yx9TUwZ3R>nO=SZN3Y
z4J(jhC%xP>Kqac^@ayPOMn_`xaCodMJl%#_Vv)TE_y9zNuz1i(n%Ik&z8;Tu+V*U<
zdOVrFOCB?9`*Iz8*2*1XDPx^Sv}l}5Ze9~OOc|_Y9rqT@b#rGJh#2UDJAB0i_%hMQ
zG=)_Qc*DFr+AdjSmQR<^lQ;|_Om)Zj8b`($Wd_j=-}x>wxX{TZjKm+(LV7OQa}9~>
z%uu^P>l>VGJ3uCHPW1iFmv+Mp4O}C7xq{H8iHwpF$eXLD?;<Hnqhfm}rQRAO08yBw
zRLD`&LI}lkwW!>}ns7Z*P`+qKHINH)^D;=GVuo-3mkBCnxL<sokg`_1U&{-#1ld)C
zV9^Gv6wK+>w^FH!%7^f1f28g_y@n5eZtH_br<Yrpq)!6{?G{I-@`U8prKKw@uxhIW
zaXsUpfX<mP3MZrch(@}04#|Twqz^qvFrp&8tR#54TbkO%$H5}68XXCsXiI8plNvFQ
zwJ&yn>Qd0igc8;8)Imf=h{ccu9NH-lG@w}>&w>Y0oq{>97{C`D+)FddL(K|HqQ)46
zftM*k6*>&2!}c{`m>|ZhQ+Mnx^vLeVL|sZXjx~fwk=~^)DU?E^?;U1dBy<pLj_*Zz
zj5IwN?W?9|JGvsn`ux8g1xq#z4zt@b6@Sd4yYxu^(YXJ=IJ-V8|NXo_T>90;Kh4>+
zP@La_2QojuVF*-Zi(422SF+km^&%=m8;|c7NDSCuOh{!gN?oNia6xWs!F*jXDlZ$S
zI+@UgM5EyeX*cEUQFX~H%;ZK9AA$@$W@MsnQyOoPYTKN|RTJZQ6fZ@WADCG45z^2Z
zk_3$7XPOW&ydOt_blEqilj06G7S`pqI9(Y#pQLy+@7#Q{gEotsnaZscNad=DU^eqJ
zzG_0Mf!vKxfam&_Q-d@%%(p1O2_`GGE{=QdGy}6JW}^nx!wu247QSJE&^K}%@}6Gb
zw_h6Miw3GSlj`tZR3M_;Sv}28k8u>jVS+OPBSJcZfQ{oiSLZXYJ}mvwSt`H^yeV$t
zY_G$(k>v{(1<djL3dS{B8))-vT7CFeCq|7=4I8ccn}~vyRc;VBGT-t6EpyLePh+IO
z_%zlZNUiz`Q#%%;V-T&w?#xs+a?8fQ`U~)0sy2#mF~+J|t%bIcL9FabV50^O#r5fU
z2XWir9_d&mJ%aM<qeGFtp=b@H;~A>GJ~##s2a*y`k@h<iq?y*k2dNV%I3!Jt!~)nT
zv`_gEQo&dNR1tt+mimuDN0WAL+$T3<GE4=n5IS&P8a#;cpm<A6?2m~x^!eDBP_Y-1
zn0MPNpGJQuanD32ARQ82d{gJiWM}p|K8;5|@d>hc0W(pgfrF6rShzqh9t=gBNANB9
zv`Bv){waUx1i9T!owS&pZp(~;<+0It+r>NLH^JQTnRIH+J%@@2-pomQR#jQ44^+5l
zWD9D^GV@5|xqM#A8gPgk7^|ZiP?ORzO6K#nB~UyI2(bk&iEc_hbxoS8z&~z55yy^q
zM5h8|6E~fS9XLUlIt;&dravRr-oi%;M*EKGd4_LowO|)pra_hkTi;zCfVo@wzW63n
zh%adIW}f+1q6Tv1aauUA2T=yZGTJEsQQsR2p4PLDlZMJ-vmd50WMX_2<Kt3`^8X8R
zc4aMlVrg~m>-Zz{lYu}60vQNoAdrDT1_HkY2n4<=mUCnF9j)JLR23FEeHtdJvZRuG
zkxCGlFR?{7g`2T)M-?qt$fF>#%l4;|yN3@MZ^VaC60UAg2?54Oz~GewZTKj8cq-M@
zVTXfty0~3v7s%LG6wv)i>2gp{)hcltVRLQZRxLKotIar0X}Q+^L47bGj^gpd(nKJR
z-|Cs|SPC}z+vBuRc{+N&C04UDcI04qbVzDulcW>l>6I7)4{>0HS9G{PJT^o$lXkv9
z=Lcc)G`7KEvjY~RW~hXy<k)K@VgkkDw>Z!1az94(aWg#6#>?=`Iq`Gn5XO%pz=R+a
zAuo{)QuZgu0>>5lP>T$_lC!#+&r{WCIRqMhYo}i7Oa`9iZnkYchUjk4DsE-09`tBg
zJ&o!(o=yh`Em;yfwh!a1fi&X!R!mKyabC3ZGdRY_dDTtHT@x5W1oSH{w|F_iih9X`
z*Pu2MQAR0QcqC?bHcCUI;mO^!?XQE4`^0zF4JukfZfS{x_Cq=F0j6<f0<xPt5<Sz+
zj*C(b`E)lwE{F<oQbbE>(rIEgH;zCt$AP01MGZ_MsKTo?F&sU1WIrb7hNoD8ZT$)Q
z5bVw9JjvH(KsL#HYn9#Y;jzQ<fyW((u<9BWK=@>XA?prXk;7GJvAG44b=<}w{7>66
MkUW;Zp=nG159Z6!p8x;=

diff --git a/AITrain/dual_ai_dialogue_system.py b/AITrain/dual_ai_dialogue_system.py
index ab4de91..c6dbf2c 100644
--- a/AITrain/dual_ai_dialogue_system.py
+++ b/AITrain/dual_ai_dialogue_system.py
@@ -286,6 +286,9 @@ class ConversationManager:
                     timestamp TEXT,
                     context_used TEXT,
                     relevance_score REAL,
+                    dialogue_score REAL DEFAULT 0.0,
+                    score_details TEXT,
+                    score_feedback TEXT,
                     FOREIGN KEY (session_id) REFERENCES conversations (session_id)
                 )
             ''')
@@ -305,7 +308,9 @@ class ConversationManager:
         print(f"✓ 创建对话会话: {session_id}")
         return session_id
     
-    def add_dialogue_turn(self, session_id: str, speaker: str, content: str, context_used: List[str] = None, relevance_score: float = 0.0):
+    def add_dialogue_turn(self, session_id: str, speaker: str, content: str, context_used: List[str] = None, 
+                         relevance_score: float = 0.0, dialogue_score: float = 0.0, 
+                         score_details: str = None, score_feedback: str = None):
         """添加对话轮次"""
         if context_used is None:
             context_used = []
@@ -318,10 +323,11 @@ class ConversationManager:
             # 插入对话轮次
             conn.execute(
                 """INSERT INTO dialogue_turns 
-                   (session_id, turn_number, speaker, content, timestamp, context_used, relevance_score) 
-                   VALUES (?, ?, ?, ?, ?, ?, ?)""",
+                   (session_id, turn_number, speaker, content, timestamp, context_used, relevance_score, 
+                    dialogue_score, score_details, score_feedback) 
+                   VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)""",
                 (session_id, turn_number, speaker, content, datetime.now().isoformat(), 
-                 json.dumps(context_used), relevance_score)
+                 json.dumps(context_used), relevance_score, dialogue_score, score_details, score_feedback)
             )
             
             # 更新会话最后更新时间
@@ -335,7 +341,7 @@ class ConversationManager:
         """获取对话历史"""
         with sqlite3.connect(self.db_path) as conn:
             cursor = conn.execute(
-                """SELECT speaker, content, timestamp, context_used, relevance_score 
+                """SELECT speaker, content, timestamp, context_used, relevance_score, dialogue_score, score_feedback 
                    FROM dialogue_turns 
                    WHERE session_id = ? 
                    ORDER BY turn_number DESC LIMIT ?""",
@@ -344,14 +350,20 @@ class ConversationManager:
             
             turns = []
             for row in cursor.fetchall():
-                speaker, content, timestamp, context_used, relevance_score = row
-                turns.append(DialogueTurn(
+                speaker, content, timestamp, context_used, relevance_score, dialogue_score, score_feedback = row
+                turn = DialogueTurn(
                     speaker=speaker,
                     content=content,
                     timestamp=timestamp,
                     context_used=json.loads(context_used or "[]"),
                     relevance_score=relevance_score
-                ))
+                )
+                # 添加评分信息到turn对象
+                if dialogue_score:
+                    turn.dialogue_score = dialogue_score
+                if score_feedback:
+                    turn.score_feedback = score_feedback
+                turns.append(turn)
             
             return list(reversed(turns))  # 按时间正序返回
     
@@ -378,10 +390,70 @@ class ConversationManager:
 class DualAIDialogueEngine:
     """双AI对话引擎"""
     
-    def __init__(self, knowledge_base: RAGKnowledgeBase, conversation_manager: ConversationManager, llm_generator):
+    def __init__(self, knowledge_base: RAGKnowledgeBase, conversation_manager: ConversationManager, llm_generator, 
+                 enable_scoring: bool = True, base_model_path: str = None):
         self.kb = knowledge_base
         self.conv_mgr = conversation_manager
         self.llm_generator = llm_generator
+        self.enable_scoring = enable_scoring
+        self.scorer = None
+        
+        # 初始化评分器
+        if enable_scoring and base_model_path:
+            try:
+                from dialogue_scorer import DialogueAIScorer
+                print("正在初始化对话评分系统...")
+                self.scorer = DialogueAIScorer(
+                    base_model_path=base_model_path,
+                    tokenizer=getattr(llm_generator, 'tokenizer', None),
+                    model=getattr(llm_generator, 'model', None)
+                )
+                print("✓ 对话评分系统初始化成功")
+            except Exception as e:
+                print(f"⚠ 对话评分系统初始化失败: {e}")
+                self.enable_scoring = False
+    
+    def score_dialogue_turn(self, dialogue_content: str, speaker: str, dialogue_history: List[DialogueTurn]) -> Tuple[float, str, str]:
+        """对单条对话进行评分
+        
+        Args:
+            dialogue_content: 对话内容
+            speaker: 说话者
+            dialogue_history: 对话历史
+            
+        Returns:
+            tuple: (总分, 详细分数JSON, 反馈意见)
+        """
+        if not self.enable_scoring or not self.scorer:
+            return 0.0, "{}", "评分系统未启用"
+        
+        try:
+            # 获取角色数据
+            character_data = self.kb.character_data.get(speaker, {})
+            
+            # 转换对话历史格式
+            history_for_scoring = []
+            for turn in dialogue_history[-5:]:  # 最近5轮对话
+                history_for_scoring.append({
+                    'speaker': turn.speaker,
+                    'content': turn.content
+                })
+            
+            # 进行AI评分
+            score_result = self.scorer.score_dialogue(
+                dialogue_content=dialogue_content,
+                speaker=speaker,
+                character_data=character_data,
+                dialogue_history=history_for_scoring,
+                context_info=[]
+            )
+            
+            # 返回评分结果
+            return score_result.overall_score, json.dumps(score_result.scores), score_result.feedback
+            
+        except Exception as e:
+            print(f"⚠ 对话评分失败: {e}")
+            return 0.0, "{}", f"评分失败: {str(e)}"
     
     def generate_character_prompt(self, character_name: str, context_info: List[Dict], dialogue_history: List[DialogueTurn], 
                                 history_context_count: int = 3, context_info_count: int = 2) -> str:
@@ -494,9 +566,17 @@ class DualAIDialogueEngine:
             context_used = [f"{info['section']}.{info['subsection']}" for info in context_info[:context_info_count]]
             avg_relevance = sum(info['relevance_score'] for info in context_info[:context_info_count]) / len(context_info[:context_info_count]) if context_info else 0.0
             
-            # 保存对话轮次
+            # 对对话进行评分
+            if self.enable_scoring:
+                dialogue_score, score_details, score_feedback = self.score_dialogue_turn(response, current_speaker, dialogue_history)
+                print(f"  [评分: {dialogue_score:.2f}] {score_feedback}")
+            else:
+                dialogue_score, score_details, score_feedback = 0.0, "{}", ""
+            
+            # 保存对话轮次（包含评分信息）
             self.conv_mgr.add_dialogue_turn(
-                session_id, current_speaker, response, context_used, avg_relevance
+                session_id, current_speaker, response, context_used, avg_relevance,
+                dialogue_score, score_details, score_feedback
             )
             
             return response, context_used
@@ -605,14 +685,29 @@ class DualAIDialogueEngine:
                 max_new_tokens=150
             )
     
-            # 保存对话到数据库
+            # 保存对话到数据库并进行评分
             for result in conversation_results:
+                # 获取当前对话历史进行评分
+                current_dialogue_history = self.conv_mgr.get_conversation_history(session_id)
+                
+                # 对对话进行评分
+                if self.enable_scoring:
+                    dialogue_score, score_details, score_feedback = self.score_dialogue_turn(
+                        result['dialogue'], result['speaker'], current_dialogue_history
+                    )
+                    print(f"  [评分: {dialogue_score:.2f}] {score_feedback[:100]}...")
+                else:
+                    dialogue_score, score_details, score_feedback = 0.0, "{}", ""
+                
                 self.conv_mgr.add_dialogue_turn(
                     session_id, 
                     result['speaker'], 
                     result['dialogue'], 
                     [result.get('context_used', '')], 
-                    0.8  # 默认相关性分数
+                    0.8,  # 默认相关性分数
+                    dialogue_score,
+                    score_details, 
+                    score_feedback
                 )
         
 
diff --git a/AITrain/main_controller.py b/AITrain/main_controller.py
index e6c0f51..904febd 100644
--- a/AITrain/main_controller.py
+++ b/AITrain/main_controller.py
@@ -140,7 +140,7 @@ def run_dialogue_system():
         conv_mgr = ConversationManager("./conversation_data/conversations.db")
         
         # 检查模型路径
-        base_model_path = '/mnt/e/AI/Project02/AITrain/Qwen/Qwen3-4B'
+        base_model_path = '/mnt/g/Project02/AITrain/Qwen/Qwen3-4B'
         lora_model_path = './output/NPC_Dialogue_LoRA/final_model'
         
         if not os.path.exists(base_model_path):
@@ -186,8 +186,14 @@ def run_dialogue_system():
             character2_config
         )
         
-        # 创建对话引擎
-        dialogue_engine = DualAIDialogueEngine(kb, conv_mgr, dual_generator)
+        # 创建对话引擎（启用评分功能）
+        dialogue_engine = DualAIDialogueEngine(
+            kb, 
+            conv_mgr, 
+            dual_generator, 
+            enable_scoring=True, 
+            base_model_path=base_model_path
+        )
         
         # 创建对话会话
         characters = [char1_name, char2_name]
@@ -317,6 +323,954 @@ def create_demo_scenario():
         import traceback
         traceback.print_exc()
 
+def analyze_model_performance():
+    """分析模型性能"""
+    print("\n" + "="*60)
+    print("模型性能分析")
+    print("="*60)
+    
+    try:
+        from dual_ai_dialogue_system import ConversationManager
+        import sqlite3
+        import json
+        from datetime import datetime, timedelta
+        
+        conv_mgr = ConversationManager("./conversation_data/conversations.db")
+        
+        with sqlite3.connect(conv_mgr.db_path) as conn:
+            print("\n1. 总体性能趋势分析:")
+            
+            # 按时间段分析性能趋势
+            cursor = conn.execute("""
+                SELECT 
+                    DATE(timestamp) as date,
+                    COUNT(*) as dialogue_count,
+                    AVG(dialogue_score) as avg_score,
+                    AVG(CASE WHEN dialogue_score >= 8.0 THEN 1.0 ELSE 0.0 END) as high_quality_rate
+                FROM dialogue_turns 
+                WHERE dialogue_score > 0 
+                    AND timestamp >= datetime('now', '-7 days')
+                GROUP BY DATE(timestamp)
+                ORDER BY date DESC
+            """)
+            
+            trend_data = cursor.fetchall()
+            if trend_data:
+                print(f"  最近7天性能趋势:")
+                for date, count, avg_score, hq_rate in trend_data:
+                    print(f"    {date}: 平均{avg_score:.2f}分 ({count}轮对话, {hq_rate*100:.1f}%高质量)")
+            else:
+                print("  暂无足够数据进行趋势分析")
+            
+            print("\n2. 维度问题分析:")
+            
+            # 分析各维度的问题
+            cursor = conn.execute("""
+                SELECT score_details 
+                FROM dialogue_turns 
+                WHERE dialogue_score > 0 AND score_details != '{}'
+                ORDER BY timestamp DESC 
+                LIMIT 100
+            """)
+            
+            dimension_scores = {
+                'coherence': [],
+                'character_consistency': [],
+                'naturalness': [],
+                'information_density': [],
+                'creativity': []
+            }
+            
+            for (score_details,) in cursor.fetchall():
+                try:
+                    scores = json.loads(score_details)
+                    for dim, score in scores.items():
+                        if dim in dimension_scores:
+                            dimension_scores[dim].append(float(score))
+                except:
+                    continue
+            
+            dimension_names = {
+                'coherence': '连贯性',
+                'character_consistency': '角色一致性',
+                'naturalness': '自然度',
+                'information_density': '信息密度',
+                'creativity': '创意性'
+            }
+            
+            weak_dimensions = []
+            for dim, scores in dimension_scores.items():
+                if scores:
+                    avg_score = sum(scores) / len(scores)
+                    print(f"    {dimension_names[dim]}: 平均{avg_score:.2f}分 ({len(scores)}个样本)")
+                    if avg_score < 7.0:
+                        weak_dimensions.append(dim)
+            
+            if weak_dimensions:
+                print(f"\n  ⚠ 发现薄弱维度: {[dimension_names[d] for d in weak_dimensions]}")
+                print("  建议进行针对性优化训练")
+            
+            print("\n3. 角色表现分析:")
+            
+            # 分析不同角色的表现
+            cursor = conn.execute("""
+                SELECT 
+                    speaker,
+                    COUNT(*) as dialogue_count,
+                    AVG(dialogue_score) as avg_score,
+                    MIN(dialogue_score) as min_score,
+                    MAX(dialogue_score) as max_score,
+                    AVG(CASE WHEN dialogue_score >= 8.0 THEN 1.0 ELSE 0.0 END) as high_quality_rate
+                FROM dialogue_turns 
+                WHERE dialogue_score > 0
+                GROUP BY speaker
+                ORDER BY avg_score DESC
+            """)
+            
+            character_performance = cursor.fetchall()
+            if character_performance:
+                print("  角色表现排名:")
+                for i, (speaker, count, avg, min_s, max_s, hq_rate) in enumerate(character_performance, 1):
+                    status = "✓" if avg >= 7.5 else "⚠" if avg >= 6.5 else "✗"
+                    print(f"    {i}. {speaker} {status}")
+                    print(f"       平均{avg:.2f}分 (范围{min_s:.1f}-{max_s:.1f}, {hq_rate*100:.1f}%高质量, {count}轮)")
+            
+            print("\n4. 问题模式识别:")
+            
+            # 识别低分对话的常见问题
+            cursor = conn.execute("""
+                SELECT content, dialogue_score, score_feedback
+                FROM dialogue_turns 
+                WHERE dialogue_score > 0 AND dialogue_score < 6.0
+                ORDER BY dialogue_score ASC
+                LIMIT 5
+            """)
+            
+            low_score_examples = cursor.fetchall()
+            if low_score_examples:
+                print("  低分对话示例:")
+                for i, (content, score, feedback) in enumerate(low_score_examples, 1):
+                    print(f"    {i}. 分数{score:.1f}: {content[:50]}...")
+                    if feedback:
+                        print(f"       问题: {feedback[:80]}...")
+            else:
+                print("  暂无低分对话样本")
+            
+            print("\n5. 优化建议:")
+            
+            # 生成优化建议
+            suggestions = []
+            
+            if weak_dimensions:
+                if 'character_consistency' in weak_dimensions:
+                    suggestions.append("• 加强角色设定训练，增加角色特征描述的权重")
+                if 'creativity' in weak_dimensions:
+                    suggestions.append("• 增加创意性训练数据，提高对话的趣味性")
+                if 'coherence' in weak_dimensions:
+                    suggestions.append("• 优化上下文理解，加强对话逻辑连贯性")
+                if 'naturalness' in weak_dimensions:
+                    suggestions.append("• 增加自然语言训练，改善表达流畅度")
+                if 'information_density' in weak_dimensions:
+                    suggestions.append("• 优化信息组织，避免冗余表达")
+            
+            # 检查是否需要数据收集
+            cursor = conn.execute("SELECT COUNT(*) FROM dialogue_turns WHERE dialogue_score > 0")
+            total_scored = cursor.fetchone()[0]
+            
+            if total_scored < 50:
+                suggestions.append("• 需要收集更多评分数据以进行准确分析")
+            
+            if total_scored >= 100:
+                suggestions.append("• 数据量充足，建议开始模型迭代优化")
+            
+            if suggestions:
+                for suggestion in suggestions:
+                    print(f"    {suggestion}")
+            else:
+                print("    当前性能表现良好，继续保持！")
+                
+    except Exception as e:
+        print(f"✗ 性能分析失败: {e}")
+        import traceback
+        traceback.print_exc()
+
+def generate_training_dataset():
+    """生成训练数据集"""
+    print("\n" + "="*60)
+    print("生成训练数据集")
+    print("="*60)
+    
+    try:
+        from dual_ai_dialogue_system import ConversationManager
+        import sqlite3
+        import json
+        import os
+        from datetime import datetime
+        
+        conv_mgr = ConversationManager("./conversation_data/conversations.db")
+        
+        # 创建输出目录
+        output_dir = "./training_data"
+        os.makedirs(output_dir, exist_ok=True)
+        
+        print("请选择训练数据生成类型:")
+        print("1. 高质量对话数据集 (分数≥8.0)")
+        print("2. 问题对话改进数据集 (分数<6.0)")
+        print("3. 角色一致性训练集")
+        print("4. 创意性增强训练集")
+        print("5. 完整对话质量数据集")
+        
+        choice = input("请输入选择 (1-5): ").strip()
+        
+        with sqlite3.connect(conv_mgr.db_path) as conn:
+            training_data = []
+            
+            if choice == '1':
+                # 高质量对话数据集
+                print("\n生成高质量对话数据集...")
+                cursor = conn.execute("""
+                    SELECT speaker, content, score_details, score_feedback
+                    FROM dialogue_turns 
+                    WHERE dialogue_score >= 8.0
+                    ORDER BY dialogue_score DESC
+                    LIMIT 200
+                """)
+                
+                for speaker, content, score_details, feedback in cursor.fetchall():
+                    training_data.append({
+                        "type": "high_quality",
+                        "speaker": speaker,
+                        "content": content,
+                        "scores": json.loads(score_details) if score_details else {},
+                        "feedback": feedback,
+                        "label": "positive"
+                    })
+                
+                output_file = f"{output_dir}/high_quality_dialogues_{datetime.now().strftime('%Y%m%d_%H%M')}.json"
+                
+            elif choice == '2':
+                # 问题对话改进数据集
+                print("\n生成问题对话改进数据集...")
+                cursor = conn.execute("""
+                    SELECT speaker, content, score_details, score_feedback
+                    FROM dialogue_turns 
+                    WHERE dialogue_score < 6.0 AND dialogue_score > 0
+                    ORDER BY dialogue_score ASC
+                    LIMIT 100
+                """)
+                
+                for speaker, content, score_details, feedback in cursor.fetchall():
+                    # 为每个低分对话生成改进指导
+                    improvement_prompt = generate_improvement_prompt(content, feedback)
+                    
+                    training_data.append({
+                        "type": "improvement",
+                        "speaker": speaker,
+                        "original_content": content,
+                        "scores": json.loads(score_details) if score_details else {},
+                        "problems": feedback,
+                        "improvement_prompt": improvement_prompt,
+                        "label": "needs_improvement"
+                    })
+                
+                output_file = f"{output_dir}/improvement_dialogues_{datetime.now().strftime('%Y%m%d_%H%M')}.json"
+                
+            elif choice == '3':
+                # 角色一致性训练集
+                print("\n生成角色一致性训练集...")
+                cursor = conn.execute("""
+                    SELECT speaker, content, score_details
+                    FROM dialogue_turns 
+                    WHERE dialogue_score > 0
+                    ORDER BY json_extract(score_details, '$.character_consistency') DESC
+                    LIMIT 150
+                """)
+                
+                for speaker, content, score_details in cursor.fetchall():
+                    scores = json.loads(score_details) if score_details else {}
+                    char_consistency = scores.get('character_consistency', 0)
+                    
+                    training_data.append({
+                        "type": "character_consistency",
+                        "speaker": speaker,
+                        "content": content,
+                        "character_consistency_score": char_consistency,
+                        "label": "high_consistency" if char_consistency >= 8.0 else "medium_consistency"
+                    })
+                
+                output_file = f"{output_dir}/character_consistency_{datetime.now().strftime('%Y%m%d_%H%M')}.json"
+                
+            elif choice == '4':
+                # 创意性增强训练集
+                print("\n生成创意性增强训练集...")
+                cursor = conn.execute("""
+                    SELECT speaker, content, score_details
+                    FROM dialogue_turns 
+                    WHERE dialogue_score > 0
+                    ORDER BY json_extract(score_details, '$.creativity') DESC
+                    LIMIT 150
+                """)
+                
+                for speaker, content, score_details in cursor.fetchall():
+                    scores = json.loads(score_details) if score_details else {}
+                    creativity = scores.get('creativity', 0)
+                    
+                    training_data.append({
+                        "type": "creativity",
+                        "speaker": speaker,
+                        "content": content,
+                        "creativity_score": creativity,
+                        "label": "high_creativity" if creativity >= 8.0 else "medium_creativity"
+                    })
+                
+                output_file = f"{output_dir}/creativity_enhancement_{datetime.now().strftime('%Y%m%d_%H%M')}.json"
+                
+            elif choice == '5':
+                # 完整对话质量数据集
+                print("\n生成完整对话质量数据集...")
+                cursor = conn.execute("""
+                    SELECT speaker, content, dialogue_score, score_details, score_feedback
+                    FROM dialogue_turns 
+                    WHERE dialogue_score > 0
+                    ORDER BY timestamp DESC
+                    LIMIT 300
+                """)
+                
+                for speaker, content, score, score_details, feedback in cursor.fetchall():
+                    training_data.append({
+                        "type": "complete_dataset",
+                        "speaker": speaker,
+                        "content": content,
+                        "overall_score": score,
+                        "dimension_scores": json.loads(score_details) if score_details else {},
+                        "feedback": feedback,
+                        "quality_label": get_quality_label(score)
+                    })
+                
+                output_file = f"{output_dir}/complete_quality_dataset_{datetime.now().strftime('%Y%m%d_%H%M')}.json"
+            
+            else:
+                print("❌ 无效选择")
+                return
+            
+            if training_data:
+                # 保存训练数据
+                with open(output_file, 'w', encoding='utf-8') as f:
+                    json.dump({
+                        "metadata": {
+                            "created_at": datetime.now().isoformat(),
+                            "total_samples": len(training_data),
+                            "data_type": choice,
+                            "source": "dialogue_scoring_system"
+                        },
+                        "data": training_data
+                    }, f, ensure_ascii=False, indent=2)
+                
+                print(f"\n✓ 训练数据集生成成功！")
+                print(f"  - 文件路径: {output_file}")
+                print(f"  - 数据条数: {len(training_data)}")
+                print(f"  - 数据类型: {get_dataset_description(choice)}")
+                
+                # 生成数据集统计信息
+                generate_dataset_statistics(training_data, choice)
+                
+            else:
+                print("❌ 未找到符合条件的数据")
+                
+    except Exception as e:
+        print(f"✗ 训练数据集生成失败: {e}")
+        import traceback
+        traceback.print_exc()
+
+def generate_improvement_prompt(content, feedback):
+    """生成改进提示"""
+    return f"""原对话: {content}
+
+问题分析: {feedback}
+
+改进要求:
+1. 保持角色特征和设定
+2. 增强对话的逻辑性和连贯性
+3. 提升表达的自然度
+4. 增加信息密度，避免冗余
+5. 适当增加创意元素
+
+请生成一个改进版本的对话。"""
+
+def get_quality_label(score):
+    """根据分数获取质量标签"""
+    if score >= 9.0:
+        return "excellent"
+    elif score >= 8.0:
+        return "good"
+    elif score >= 7.0:
+        return "average"
+    elif score >= 6.0:
+        return "below_average"
+    else:
+        return "poor"
+
+def get_dataset_description(choice):
+    """获取数据集描述"""
+    descriptions = {
+        '1': "高质量对话数据集",
+        '2': "问题对话改进数据集",
+        '3': "角色一致性训练集",
+        '4': "创意性增强训练集",
+        '5': "完整对话质量数据集"
+    }
+    return descriptions.get(choice, "未知类型")
+
+def generate_dataset_statistics(training_data, data_type):
+    """生成数据集统计信息"""
+    print(f"\n数据集统计信息:")
+    
+    if data_type == '1':  # 高质量数据集
+        speakers = {}
+        for item in training_data:
+            speaker = item['speaker']
+            speakers[speaker] = speakers.get(speaker, 0) + 1
+        
+        print(f"  - 角色分布:")
+        for speaker, count in speakers.items():
+            print(f"    • {speaker}: {count}条对话")
+    
+    elif data_type == '5':  # 完整数据集
+        quality_dist = {}
+        for item in training_data:
+            label = item['quality_label']
+            quality_dist[label] = quality_dist.get(label, 0) + 1
+        
+        print(f"  - 质量分布:")
+        for label, count in quality_dist.items():
+            print(f"    • {label}: {count}条对话")
+
+def run_model_optimization():
+    """运行模型迭代优化"""
+    print("\n" + "="*60)
+    print("模型迭代优化")
+    print("="*60)
+    
+    try:
+        from dual_ai_dialogue_system import ConversationManager
+        import sqlite3
+        import json
+        import os
+        from datetime import datetime
+        
+        conv_mgr = ConversationManager("./conversation_data/conversations.db")
+        
+        print("模型优化选项:")
+        print("1. 分析优化需求")
+        print("2. 生成LoRA训练脚本")
+        print("3. 创建提示优化配置")
+        print("4. 执行增量训练")
+        print("5. 性能对比验证")
+        
+        choice = input("请输入选择 (1-5): ").strip()
+        
+        if choice == '1':
+            # 分析优化需求
+            print("\n=== 优化需求分析 ===")
+            
+            with sqlite3.connect(conv_mgr.db_path) as conn:
+                # 获取性能数据
+                cursor = conn.execute("""
+                    SELECT 
+                        COUNT(*) as total,
+                        AVG(dialogue_score) as avg_score,
+                        AVG(CASE WHEN dialogue_score >= 8.0 THEN 1.0 ELSE 0.0 END) as high_quality_rate
+                    FROM dialogue_turns WHERE dialogue_score > 0
+                """)
+                
+                total, avg_score, hq_rate = cursor.fetchone()
+                
+                print(f"当前性能指标:")
+                print(f"  - 总评分样本: {total}")
+                print(f"  - 平均分数: {avg_score:.2f}")
+                print(f"  - 高质量率: {hq_rate*100:.1f}%")
+                
+                # 分析优化潜力
+                optimization_needs = []
+                
+                if avg_score < 7.0:
+                    optimization_needs.append("整体质量需要提升")
+                
+                if hq_rate < 0.6:
+                    optimization_needs.append("高质量对话比例偏低")
+                
+                # 分析各维度表现
+                cursor = conn.execute("""
+                    SELECT score_details FROM dialogue_turns 
+                    WHERE dialogue_score > 0 AND score_details != '{}'
+                    ORDER BY timestamp DESC LIMIT 100
+                """)
+                
+                dim_scores = {'coherence': [], 'character_consistency': [], 
+                             'naturalness': [], 'information_density': [], 'creativity': []}
+                
+                for (score_details,) in cursor.fetchall():
+                    try:
+                        scores = json.loads(score_details)
+                        for dim, score in scores.items():
+                            if dim in dim_scores:
+                                dim_scores[dim].append(float(score))
+                    except:
+                        continue
+                
+                weak_dimensions = []
+                print(f"\n维度分析:")
+                for dim, scores in dim_scores.items():
+                    if scores:
+                        avg = sum(scores) / len(scores)
+                        print(f"  - {dim}: {avg:.2f}分")
+                        if avg < 7.0:
+                            weak_dimensions.append(dim)
+                
+                if weak_dimensions:
+                    optimization_needs.append(f"薄弱维度: {weak_dimensions}")
+                
+                print(f"\n优化建议:")
+                if optimization_needs:
+                    for i, need in enumerate(optimization_needs, 1):
+                        print(f"  {i}. {need}")
+                else:
+                    print("  当前模型表现良好，可考虑细微调优")
+                
+        elif choice == '2':
+            # 生成LoRA训练脚本
+            print("\n=== 生成LoRA训练脚本 ===")
+            
+            script_content = generate_lora_training_script()
+            script_path = "./scripts/iterative_lora_training.py"
+            
+            os.makedirs("./scripts", exist_ok=True)
+            with open(script_path, 'w', encoding='utf-8') as f:
+                f.write(script_content)
+            
+            print(f"✓ LoRA训练脚本已生成: {script_path}")
+            print("使用方法:")
+            print("  1. 先运行训练数据生成 (选项8)")
+            print("  2. 修改脚本中的路径配置")
+            print(f"  3. 运行: python {script_path}")
+            
+        elif choice == '3':
+            # 创建提示优化配置
+            print("\n=== 创建提示优化配置 ===")
+            
+            config = generate_prompt_optimization_config()
+            config_path = "./config/prompt_optimization.json"
+            
+            os.makedirs("./config", exist_ok=True)
+            with open(config_path, 'w', encoding='utf-8') as f:
+                json.dump(config, f, ensure_ascii=False, indent=2)
+            
+            print(f"✓ 提示优化配置已生成: {config_path}")
+            print("配置包含:")
+            print("  - 动态提示调整规则")
+            print("  - 质量阈值设置")
+            print("  - 生成参数优化")
+            
+        elif choice == '4':
+            # 执行增量训练
+            print("\n=== 执行增量训练 ===")
+            
+            # 检查训练数据
+            training_dir = "./training_data"
+            if not os.path.exists(training_dir):
+                print("❌ 训练数据目录不存在，请先生成训练数据 (选项8)")
+                return
+            
+            training_files = [f for f in os.listdir(training_dir) if f.endswith('.json')]
+            if not training_files:
+                print("❌ 未找到训练数据文件，请先生成训练数据 (选项8)")
+                return
+            
+            print(f"找到训练数据文件:")
+            for i, file in enumerate(training_files, 1):
+                print(f"  {i}. {file}")
+            
+            file_idx = input(f"选择训练数据文件 (1-{len(training_files)}): ").strip()
+            try:
+                selected_file = training_files[int(file_idx) - 1]
+                training_file_path = os.path.join(training_dir, selected_file)
+                
+                print(f"将使用训练文件: {selected_file}")
+                print("⚠ 注意：实际训练需要配置正确的模型路径和计算资源")
+                
+                # 生成训练命令
+                training_command = generate_training_command(training_file_path)
+                print(f"建议训练命令:")
+                print(f"  {training_command}")
+                
+                # 可选：执行训练（需要用户确认）
+                confirm = input("是否现在执行训练？(y/N): ").strip().lower()
+                if confirm == 'y':
+                    print("开始增量训练...")
+                    # 这里可以添加实际的训练执行逻辑
+                    print("⚠ 训练功能需要根据实际环境配置")
+                
+            except (ValueError, IndexError):
+                print("❌ 无效的文件选择")
+                
+        elif choice == '5':
+            # 性能对比验证
+            print("\n=== 性能对比验证 ===")
+            
+            print("验证选项:")
+            print("1. 历史性能趋势对比")
+            print("2. A/B测试配置生成")
+            print("3. 模型版本性能对比")
+            
+            verify_choice = input("请输入选择 (1-3): ").strip()
+            
+            if verify_choice == '1':
+                # 历史性能趋势
+                with sqlite3.connect(conv_mgr.db_path) as conn:
+                    cursor = conn.execute("""
+                        SELECT 
+                            DATE(timestamp) as date,
+                            AVG(dialogue_score) as avg_score,
+                            COUNT(*) as count
+                        FROM dialogue_turns 
+                        WHERE dialogue_score > 0 
+                            AND timestamp >= datetime('now', '-30 days')
+                        GROUP BY DATE(timestamp)
+                        ORDER BY date ASC
+                    """)
+                    
+                    trend_data = cursor.fetchall()
+                    if trend_data:
+                        print(f"30天性能趋势:")
+                        for date, avg_score, count in trend_data:
+                            trend = "📈" if avg_score > 7.5 else "📉" if avg_score < 6.5 else "📊"
+                            print(f"  {date}: {avg_score:.2f}分 {trend} ({count}条对话)")
+                    else:
+                        print("暂无足够的历史数据")
+                        
+            elif verify_choice == '2':
+                # A/B测试配置
+                ab_config = generate_ab_test_config()
+                ab_config_path = "./config/ab_test_config.json"
+                
+                os.makedirs("./config", exist_ok=True)
+                with open(ab_config_path, 'w', encoding='utf-8') as f:
+                    json.dump(ab_config, f, ensure_ascii=False, indent=2)
+                
+                print(f"✓ A/B测试配置已生成: {ab_config_path}")
+                print("配置包含:")
+                print("  - 对照组和实验组设置")
+                print("  - 评估指标定义")
+                print("  - 测试持续时间配置")
+                
+            elif verify_choice == '3':
+                print("模型版本对比功能开发中...")
+                print("建议手动记录不同版本的性能指标进行对比")
+        
+        else:
+            print("❌ 无效选择")
+            
+    except Exception as e:
+        print(f"✗ 模型优化失败: {e}")
+        import traceback
+        traceback.print_exc()
+
+def generate_lora_training_script():
+    """生成LoRA训练脚本"""
+    return '''#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+基于评分数据的LoRA增量训练脚本
+自动生成 - 请根据实际环境调整配置
+"""
+
+import json
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from peft import LoraConfig, get_peft_model, TaskType
+import os
+
+class IterativeLoRATrainer:
+    def __init__(self, base_model_path, training_data_path, output_path):
+        self.base_model_path = base_model_path
+        self.training_data_path = training_data_path
+        self.output_path = output_path
+        
+        # LoRA配置
+        self.lora_config = LoraConfig(
+            task_type=TaskType.CAUSAL_LM,
+            inference_mode=False,
+            r=16,  # LoRA rank
+            lora_alpha=32,
+            lora_dropout=0.1,
+            target_modules=["q_proj", "v_proj", "k_proj", "o_proj"]
+        )
+    
+    def load_training_data(self):
+        """加载训练数据"""
+        with open(self.training_data_path, 'r', encoding='utf-8') as f:
+            data = json.load(f)
+        return data['data']
+    
+    def prepare_training_samples(self, data):
+        """准备训练样本"""
+        samples = []
+        
+        for item in data:
+            if item.get('label') == 'positive' or item.get('overall_score', 0) >= 8.0:
+                # 高质量样本
+                sample = {
+                    'input': f"角色: {item['speaker']}\\n请生成高质量对话:",
+                    'output': item['content'],
+                    'quality_score': item.get('overall_score', 8.0)
+                }
+                samples.append(sample)
+                
+        return samples
+    
+    def train(self):
+        """执行训练"""
+        print("开始LoRA增量训练...")
+        
+        # 加载模型和分词器
+        tokenizer = AutoTokenizer.from_pretrained(self.base_model_path)
+        model = AutoModelForCausalLM.from_pretrained(
+            self.base_model_path,
+            torch_dtype=torch.bfloat16,
+            device_map="auto"
+        )
+        
+        # 应用LoRA
+        model = get_peft_model(model, self.lora_config)
+        
+        # 加载训练数据
+        training_data = self.load_training_data()
+        training_samples = self.prepare_training_samples(training_data)
+        
+        print(f"训练样本数量: {len(training_samples)}")
+        
+        # 这里添加实际的训练循环
+        # 建议使用transformers的Trainer或自定义训练循环
+        
+        # 保存模型
+        model.save_pretrained(self.output_path)
+        tokenizer.save_pretrained(self.output_path)
+        
+        print(f"训练完成，模型保存到: {self.output_path}")
+
+if __name__ == '__main__':
+    # 配置参数 - 请根据实际情况修改
+    BASE_MODEL_PATH = '/mnt/e/AI/Project02/AITrain/Qwen/Qwen3-4B'
+    TRAINING_DATA_PATH = './training_data/high_quality_dialogues_latest.json'
+    OUTPUT_PATH = './output/iterative_lora_v2'
+    
+    trainer = IterativeLoRATrainer(BASE_MODEL_PATH, TRAINING_DATA_PATH, OUTPUT_PATH)
+    trainer.train()
+'''
+
+def generate_prompt_optimization_config():
+    """生成提示优化配置"""
+    return {
+        "optimization_rules": {
+            "quality_thresholds": {
+                "excellent": 9.0,
+                "good": 8.0,
+                "acceptable": 7.0,
+                "needs_improvement": 6.0
+            },
+            "adaptive_adjustments": {
+                "low_coherence": {
+                    "add_context_emphasis": True,
+                    "increase_history_weight": 1.2,
+                    "add_logical_constraints": True
+                },
+                "low_character_consistency": {
+                    "enhance_character_description": True,
+                    "add_personality_reminders": True,
+                    "increase_character_weight": 1.5
+                },
+                "low_creativity": {
+                    "add_creativity_prompts": True,
+                    "increase_temperature": 0.1,
+                    "diversify_examples": True
+                }
+            }
+        },
+        "dynamic_prompts": {
+            "quality_boost_templates": [
+                "请生成一个富有创意且符合角色特征的高质量对话",
+                "确保对话逻辑连贯，信息丰富，表达自然",
+                "体现角色的独特性格和说话风格"
+            ],
+            "problem_specific_guidance": {
+                "repetitive": "避免重复之前的内容，提供新的信息和观点",
+                "inconsistent": "严格遵循角色设定，保持人格一致性",
+                "dull": "增加对话的趣味性和深度，使用生动的表达"
+            }
+        },
+        "generation_parameters": {
+            "adaptive_temperature": {
+                "high_creativity_needed": 0.9,
+                "normal": 0.8,
+                "high_consistency_needed": 0.7
+            },
+            "adaptive_top_p": {
+                "creative_mode": 0.9,
+                "balanced_mode": 0.8,
+                "conservative_mode": 0.7
+            }
+        }
+    }
+
+def generate_ab_test_config():
+    """生成A/B测试配置"""
+    return {
+        "test_name": "model_optimization_validation",
+        "created_at": datetime.now().isoformat(),
+        "groups": {
+            "control": {
+                "name": "原始模型",
+                "description": "未优化的基础模型",
+                "model_path": "/path/to/base/model",
+                "sample_ratio": 0.5
+            },
+            "experimental": {
+                "name": "优化模型",
+                "description": "经过迭代优化的模型",
+                "model_path": "/path/to/optimized/model",
+                "sample_ratio": 0.5
+            }
+        },
+        "evaluation_metrics": {
+            "primary": [
+                "overall_dialogue_score",
+                "character_consistency_score",
+                "creativity_score"
+            ],
+            "secondary": [
+                "user_satisfaction",
+                "response_time",
+                "coherence_score"
+            ]
+        },
+        "test_duration": {
+            "target_samples": 200,
+            "max_duration_days": 7,
+            "min_samples_per_group": 50
+        },
+        "statistical_settings": {
+            "confidence_level": 0.95,
+            "minimum_effect_size": 0.3,
+            "power": 0.8
+        }
+    }
+
+def generate_training_command(training_file_path):
+    """生成训练命令"""
+    return f"python ./scripts/iterative_lora_training.py --data {training_file_path} --output ./output/optimized_model_v{datetime.now().strftime('%Y%m%d')}"
+
+def show_scoring_statistics():
+    """显示对话评分统计"""
+    print("\n" + "="*60)
+    print("对话评分统计")
+    print("="*60)
+    
+    try:
+        from dual_ai_dialogue_system import ConversationManager
+        import json
+        
+        conv_mgr = ConversationManager("./conversation_data/conversations.db")
+        
+        # 查询评分数据
+        import sqlite3
+        with sqlite3.connect(conv_mgr.db_path) as conn:
+            # 总体统计
+            cursor = conn.execute("""
+                SELECT 
+                    COUNT(*) as total_turns,
+                    AVG(dialogue_score) as avg_score,
+                    MAX(dialogue_score) as max_score,
+                    MIN(dialogue_score) as min_score
+                FROM dialogue_turns 
+                WHERE dialogue_score > 0
+            """)
+            
+            stats = cursor.fetchone()
+            if stats and stats[0] > 0:
+                total_turns, avg_score, max_score, min_score = stats
+                print(f"总体统计:")
+                print(f"  - 已评分对话轮数: {total_turns}")
+                print(f"  - 平均分数: {avg_score:.2f}")
+                print(f"  - 最高分数: {max_score:.2f}")
+                print(f"  - 最低分数: {min_score:.2f}")
+            else:
+                print("暂无评分数据")
+                return
+            
+            # 按角色统计
+            print(f"\n按角色统计:")
+            cursor = conn.execute("""
+                SELECT 
+                    speaker,
+                    COUNT(*) as turns,
+                    AVG(dialogue_score) as avg_score,
+                    MAX(dialogue_score) as max_score
+                FROM dialogue_turns 
+                WHERE dialogue_score > 0
+                GROUP BY speaker
+                ORDER BY avg_score DESC
+            """)
+            
+            for row in cursor.fetchall():
+                speaker, turns, avg_score, max_score = row
+                print(f"  - {speaker}: 平均{avg_score:.2f}分 (最高{max_score:.2f}分, {turns}轮对话)")
+            
+            # 最近高分对话
+            print(f"\n最近高分对话 (分数≥8.0):")
+            cursor = conn.execute("""
+                SELECT speaker, content, dialogue_score, score_feedback, timestamp
+                FROM dialogue_turns 
+                WHERE dialogue_score >= 8.0
+                ORDER BY timestamp DESC
+                LIMIT 5
+            """)
+            
+            high_score_turns = cursor.fetchall()
+            if high_score_turns:
+                for speaker, content, score, feedback, timestamp in high_score_turns:
+                    print(f"  [{timestamp[:16]}] {speaker} ({score:.2f}分)")
+                    print(f"    内容: {content[:80]}...")
+                    if feedback:
+                        print(f"    评价: {feedback[:60]}...")
+                    print()
+            else:
+                print("  暂无高分对话")
+                
+            # 分数分布统计
+            print(f"\n分数分布:")
+            cursor = conn.execute("""
+                SELECT 
+                    CASE 
+                        WHEN dialogue_score >= 9.0 THEN '优秀 (9-10分)'
+                        WHEN dialogue_score >= 8.0 THEN '良好 (8-9分)'
+                        WHEN dialogue_score >= 7.0 THEN '中等 (7-8分)'
+                        WHEN dialogue_score >= 6.0 THEN '及格 (6-7分)'
+                        ELSE '待改进 (<6分)'
+                    END as score_range,
+                    COUNT(*) as count
+                FROM dialogue_turns 
+                WHERE dialogue_score > 0
+                GROUP BY score_range
+                ORDER BY MIN(dialogue_score) DESC
+            """)
+            
+            for score_range, count in cursor.fetchall():
+                percentage = (count / total_turns) * 100
+                print(f"  - {score_range}: {count}轮 ({percentage:.1f}%)")
+        
+    except Exception as e:
+        print(f"✗ 评分统计查询失败: {e}")
+
 def show_system_status():
     """显示系统状态"""
     print("\n" + "="*60)
@@ -382,11 +1336,15 @@ def main():
         print("3. 启动双AI对话系统 (支持双模型对话)")
         print("4. 创建演示对话场景")
         print("5. 系统状态检查")
-        print("6. 查看使用说明")
+        print("6. 查看对话评分统计")
+        print("7. 模型性能分析与优化")
+        print("8. 生成训练数据集")
+        print("9. 模型迭代优化")
+        print("10. 查看使用说明")
         print("0. 退出")
         print("="*50)
         
-        choice = input("请输入选择 (0-6): ").strip()
+        choice = input("请输入选择 (0-10): ").strip()
         
         if choice == '0':
             print("\n感谢使用双AI角色对话系统！")
@@ -408,6 +1366,18 @@ def main():
             show_system_status()
             
         elif choice == '6':
+            show_scoring_statistics()
+            
+        elif choice == '7':
+            analyze_model_performance()
+            
+        elif choice == '8':
+            generate_training_dataset()
+            
+        elif choice == '9':
+            run_model_optimization()
+            
+        elif choice == '10':
             show_usage_guide()
             
         else: