+tFzJ^xph^u1!Fs{R&YjB6DROynJli$rXD07Gx_*UE5
zrbfsHqNyS0VU{4AP@*XR!ecC(Z|{su_z2z*$k8L};eD?xu85jFD=E5s%pn(U3PB+7
zk{vm%5ATMnW&;Q5m9h)DW?!479Laa-Yw^be{6;sgEKTyBlZtWZukItkxWxNL!<-S*
zeZj(Ec;)oemvSkJs#?-dY!5~~hWw~^NkiSGUO{`bO*00rX7nhV$6kJAT{Tlh4!1wGU+AbZ_!Il%!3Xh3d
z)geWXe@y6E-p>#@qaEN5TAalO>M(G$l(0&;-4fVZn#%ck%vCl_tA4$6??=a!pXw#A
zYZL3o%iZ(CZ*YcQW$zNUqx=!F;X!qlxWI1crDt+Si0j*PwK55fShzFt*~-s?ht(iU
z02dMD#zdvPlrI;~+nqt>MmK1QubLP0U^!f+K)OD}^9J@zPwjPfpEo{u&04cnj8%MD
z;FL%0DRK&N+NXR^Ra41McPG>`WVJIr6fWS5)x6g?mK-lcB}uS;-1=i=pUT?OlzSqewv>at60QJ%JjRR=sIF_}|7gHNv1lW#3}
zgN25)wP(biN2=|g2nfwt9aZ-o`~2%ijO5)}%T6{fk7!nO6lEu$W?v*QwH1J6XS5jj
z5ZWGHnhVc7aXCEm9Jeh~yxP40LS%5_1m_`a_~ipQPE7&K&s<#Ne$n1
zsWebPceBi97B2S4SZB13GfKYIV((Ed2r#9o;`0&T57}GB-Dpw`d?JSK81E6|E5|kN
zq$kVni0~(sM`Q=Xw_v=Jea~jY3Yb#ye5C7l!sosq`q@}k=y6HBiYQ;0_I@U`5taMv
zt>t;n)u+=DR-Ak)}wym0d(e
zG_ia13IY>*imbyczGkQ%W%xWu3&mTvs-jFIq)Qs_ZA+1*hs%ULHoM?AOZIao
z0v8XaUVF=kdvswm?P;ET$dQ^YVm87Fskk)MBEpeqo6`zhjDJhaipXwV;(rjctM{_6
zZPDj`05;d*1(AHMXDY8uNjK^n7l(#vgwqY)Z={`xEMn>}8hcxthBYp4#k)-~Z)4MeH-5VT6CKDRDzCr)(e){mUvF!xCcJ`mSP
zx{#d8ec_~lkByHUNyQ-ZDj(1Oi#oyOST#O6blNyF-5h*g=i&OL-D`xP5C`}UJz%Ng
z+0!v6enRkL3Gk3#8u(vB9A_-l78MN;SXi!1l$3IQe!=yo0K4P_jZe~oEYME@o}{-l
zP7LBWZ9^E*3yOn@o|Cs_Px&0@m2IdmG3-&zMs#KHW20glO+p7Z)%DMkOU4zjw;`N+
ze!7Yfl@{`hgJ#mT^=EqO!A$q5uN{R}4cGULQT5l40YO?HM@Nr5HFm9#w`U8Ka}mnw
zCi-unHnt=
zG$QSZ@Uf$j1iBflu~53B25im6x%qf*#sIwB+6N>OiEy7XUnIJqIU#tcWwafoL{h(H
zIjBi4=&V6_(5?0y-I~~UsI6mwD5A+iB$rj_5Y9)j!%ktn(#N1nC-*I5f@~!m5WJ#K
zfqrGpKuvw^n>fe<5hB2m*N277_Qlb4XX%FQWC}c&ZP+T8P5=1#>6Xt7*rSnIrjN!N
z?`OtAt!E!}Xu|(z#mD7Qt@f86<>qgvE}*?mei!|E=PfoogR=8okW{I4;$3$0z851u
zSCsD2-DDin>(c-QN2C#9R>I*@x=3p3vuzY6u;keDU
z&vO(5)edQhi4xv8AwNriCH|tf?cXx8NascoKz!B!L@}Q%UL9+yg|?=m;IdOq72q&X
zj_2&By7DOjGoFHOoAtsTY_BZ;8EsU5Y<)c0Y;L^Vh8dmR`V+)c0bEL)Vj3HV5=t8Q
z)tfZKjpC#g<~SQ?!!u?2oE?FYc4?@9kexo)7Gd#A6+5ZO>FrhqhC5d;LFM=pKO>iv
zWVfM`c<}<11_KYxME85-nO1p}%k(u}YkZtd%BK60##5U&N=uI36u~M9-jp=MWl&Nm
zu12&9;Y^TIPd+d9^O<&@ZCl3TXNjrZ7D^*XhayIIZfP<^y1AVrvAT7y#2;t}(X^x^3XbrA`EztZYbOys)9^kQ;gL`$P-I9kJ^K4yn2Mwrcak&zDJ-pkh8Z?o{l@)Ha^oK
zFGAM*3Vxwuu5Hy;^6fZw7z8K;-mAhRckjnw-zbb{e@hcW8)5nqgtMX|M`qp*Ja_s;
z>u8{BGXH{b@(^+|#699`FL&-kPqn-Pf1c5gADAOzW6U_vF>;Vfp?yQFJ%>B8|r+Vc-w%vOT-PX>_
z6#aG%ky$)X2$b~oVI7Al+m_wF7>S_c`r!o(rgLX_j{K5kYkH_QN=V6RUCl|?
z;5LkrL+iuD;$O
z6Ba)X`}WAZk%Hkm4l19E2_l@myhXguU!6yEq!oH_MKX)p&PxmC;Tu`Z46H*z{p
zy2iao^FxXrQ?{KqC!IhZMlx@p|CF(RwyC)#-KHl#N;85#z}yTS3hA6YFJEem>v_u`bpgn<+M-GqyQu
z=MyMzsIYui!^4M4tT=%n;#I6#4GB%Vwb_Q=Qdebk-}EuPWa(n}D_CPWV>oe41k0IV
zQeWVN@+C*kN=KzL-(&(wpzSQEULA9>zM};0E9?CP6cG-c!BdnO>T
zEY714Q@aaR5%hk_!IBw@@_UBpPadyjYSV;x)WU>oDpZYetRUf%o({y^XDIu>A`ceG9qytUW0KaPB=w43gE
zeyELJN@!0(?BoQbX1Qke9-$wV1pOTrf(AQEoy>9}MDb`6cqLr?rXCmHnfVU2LRiC3
z>9-UL1Lrt`)T&gsQX{412enQAq=y#rv4Sqe)6O3G8Zy6b&JuwP*a^Xc@MK#P^TEBI
z@)^V6^Dl&0g)ruFW-iii{zD~c3ucrF@aUA0?iJZ=g3HdI$oYmyuw>ufD&%cQTPaZ#MRia1-J%SMMPi;v%x$Ww%o&AY%}-m3cb21A$*P
zjwc&SuSiREgghasabF>rljk8w3Y0#v1=O-+7O``AX@XrUqG_X8dGH={gm6nvd4)3wBt9{0$-TPx?Mr#JrcS&Wwbl=l|
zjnQ_ObWe6~290b^FLti7d4yrcRlgwQ1KcbFN{a*hrni2JPYAomWkWO31$MFeeGtg8
ze^Q>$rl$l8a>(5Ig*XVpP=(@-|MEG~!zC2;AKImXGWoJ^OPj>Z|xyoCafV@%frLM^W?@v-?GYa<2r3Dp4r1XozL7vi4
zG3Vqe&;VPFH7rjG4tTJ8hp_6maU8~aMrP#)LDbW_aL`zerBGNVdcMu}4n)ZNdn%pD
zdG&o9C&$XsvZ|2Oh6vxCa17imA3?-QXX!sBtM%Al>)Owsl$#wX?L&BcaLx6f`Bbf7F)S
zLOknmZ8)PKM)+2cT51ZCjtUJ6eofLW_6`6aju+z97W9sL{u1ngm_Gqs7Y|1y9W}`w-rgXIN`j4{+5S!DCw@HBhYCFNSG!qVw_EnU^$^R3gT8OwbW`8ySrU|mSJR!8f>eAHIkald!H-&+Z
zi2dZGc^?NABi*KrK~Bi`%oZ;n<`x$bv)GJ&SWl=Oyfji)1^yD7s>#21c{{f{oa?MW
zw7Urg8ppea`8I~o`PutlJ`-rNxiBN6d1p(49WiwS%QQE@;(o1#b@Tcs2r+#NF>IfX
z`+NcH8zo|KI6;-mE5b}i#}a{K9iw*BK$vb*Sx93bgDIo2gu;RvT`wX@?#MT>M^aYJ
z)vkep77HjzzQp4s_R8a?kT)J+&qk-iDLo?-G#nG7hBE!_)!wR*((F%8otK>ahL>NZ
z&GtF*K9tmSSGlA^_S_^ruY(o;4}j>2K%?O-wntNXX|_ZYZk9kcDpk8k{7tJOy%4AU
zY0+XY6iHq?KZ_d^8FX%_VKs*OL2;ax=6EDz?i)bZE4Y(ve;SV1Xe=k&5-)SG^ar!W
zWV40iHMzusXU(;Waq7>e4rd$@A{|b$oE{-|waTL_XsUV~K#T(1;f&kW=%AfU=rzry
zB+E$E)M*04+5(+rmZM5>`=hZYORYDeEa-~XLL!=5r8ITH<*=rHt{e`1M%O)0pnXb{
zB?+bka7YG_8hEl5c6(k67uDkGpwF1J+nXxVIZc>z=j5BNr6tQ--)OZew9QaYGM(I~
zP@O!LrJ1T^lmr-OX(EDg_Eos^yAE+mKK}fs7=R#m33_s_G5_=hY6>pH=;*s3$d5jv
ztR~;2jSWczXnI=-xq|JuFBtSWBTZX&?L=yh4-#f@?&OM9uE<0D8TqE*H0`VXE)Bft
zZKK8QP=C4EIvi0={f-;3J>#Xoq3N7jr>%d2SGJ97S@W6}i?u}SDkq&*!*SmSr*YBc
zP)GF|#0Uwx{kqm!$Y#t@M0UqVYe$OvNWZ-WVcJz>^{AH}jVV~dTmr>3PCsRYDebV
zs*i{52^7fbkF>4Y68LFsWNinPAKYbEOZs&(%N>7ABL7RtOFJSIO~;G_NI-3PJ5V+4K{sQkt>H8idV1_n-+6+!rV2S%P+l&>FAlsY{8
z0y27jz>Cl{U-JykEnOz&>hqT|bQRzH#5@iK$)68MT{Z`sWB%=0!0qxt&+gWHyN-Nm
zh~l5HPNcK#00`7w@!&QKoEa@pzoWa<$%*5Ivp-;nTuz?W@0e?0m_Ym~764NJ3Ggho
zYp(~X`jOC&{D4~8JxI$C+r3qWNU!+0_)B^&?zWh{0`8OGg&c(~e%LRo;;^~Oeyu7-
z`thxT>R@|>CIWF%TcB3$f)_~msNYKr2J_8u^AXN(3mcXG@_KWj0RH8#**cnACK8;>
zPo;4{`<$uEZD4ck0Q7OYt`kweD`WLoerc-TDTIN)1jw*17&_7`h~@Mowe$$-0|ucN
zIc5?77`yM`A}~JNCXU}>h0H^T;(&^!Hcx{wHytpv#@!nLHMi++GVUfRjm;^tu0z*i
z1or$qxb+Nu*J}~Dx%jJV5vE;KjO!cU4qO;*tAkrH?CAMi)TsmH
zkMUc+^-~&)KaZkaAe(aZvps?URiz^Fz(4&NTg3V@ZRO@{z8Y)V6Jh}h`+_(sOJ8p|
z2xR1;mg?BrJqGK=Yt1Kn%g;kp;Ws@4!~+8Ws3Cx&eG#|toDQ!6SS>?I0lo-#GYRI^
z9GZgX0f$w9=9&ZWva`C=ho2_*rSZ(DI}df>1&7c2bC1~
z_GXQ&vU@nRKNG0Tvn=pM;Z+im=fn@_(0>h}qVZOZ2m~2p>=IP5Pn&VJYRa2`pvF>n
zWmgxHBR45@w%t3Z5^NPEUMnujoljMk?8Q}+s8P<|kZKKiqpj!5paDSw!TF+Ih}cIX
zrOW7!^2`-F%WTT{p*8_<$-vI?Tm(infs3T2b!I?j&KT7q?}*nyeY*r^6vR9ska3Z9
z&!-m`&M~?+etBgqEV4iECpF=-R8r}pw)EjP~Y-@nxfkQZk-hl>$v;|qaD_PZhz!F|tKX}ky4
zgl8v}(UR3WQvkxgScd*3qD_{ST-z{+lN^5G+KXRXXenI4a~HSZAMw$@voZZ`+9E`s
zr?&2F76@a;h4RXw#v<}EA7ZB=HTF9
zf6vId9zop?r$iQ(TXS@Ch1zbec5{b@=DPx~dkdQF^tCeIDiNpHXqNqc$e9d21Q&5d
z1I%+OYmg(ZvKizu&CIP@S!pNivp`MfIbQ8Fb`3+oIFm6y&@4Covw&L$p9tl7*hs2;-$+In2NZ(q4QdbsKh!e(RX592i87ITY_w0lwQU(YVTh**r0
zZ0(=Nr{xPVN4DH6K7e0%K|nd@4}WW-WKGwQFOZc76Y334v7iULG=Wv-Q_H@*Uvd&&
z>1*BJPt$g*6Y;VewA#EzGA%I53xx9FO|q|T=WSt+at1T<9;m`xGUW@d
z%IKrx6g#=o*jL`9VX${wXJFLHwREaXyRjL0nHc=Rz=}hP@@7Wr10kk7xXD>z_4z7{
z)i`P68BBt0E=uC-BQ2J86ix?;pzE1tel_s@C}a;G!gA}r-*tAhB(F+(
z(CDP4YhXI{=w4=%w*f_P&XfbU2U$%c%)lhlr=i=9dy$iaHV^eb)NjXjG3%&e2JWI(
z`8R)RfBU*D-3R{RI(?^eikzKJeR?kHKAy5$odSo4cC3KsFz&&$PUlas?(K7otH{3{
zM1^P1y>OVJUAPM-Ai-9|!Km#mG+L-j8(#43J%xGq5EY)zk9)aZZXVZepE0RL)rDG2
z58fS>BZwT&iPs~`rQ*!jN%qC|kB&;Wb0-N^AHjO|Sq;!kqjHz@3gSe!LP|70acuOB
zF)xQ!P={#?I!rq^&Y3ZFoQk@Y{D_F6Nt!s+d>K8e=nusQ)L=!;oXBxFTP^X-65hr)
zE48B41Evy^Npifm4&%~btsya~v*BX}IgVq{o&^ZfXF^Wft)!r9do6_04C!0Gro@BV(dL5blSk&aa=ji=X$soSxe
zMjN-B{jrWFq?#|Tx`gT(@x9WV)Y`8i%3bGFkM8e!$1l;fh7jh`eLw9r*(Io**w-sZ
z#|&%l39w%_3C0NX889$!G?-}zdeUnCv}Oq%FD%hi-jA)oci&$*Hv_m=>O{tM8Ego_
z-M8;-1A6*cj8i(N+9!vk^Ai8
zNAmvUCH9P=7*lL>JZ7H5rRF{
zo&8Kg+!czxoXj8nIzc1fjATOb{mbW8Ia`SLe8onzt|r??;%)>7;3v*d2jmbsl@;|X
zOW2S9M)IGjPj8gyE3-NL7%_+w@^;wRXgPT-GK@u{qw06@zP`-qVQ`P~%y}S>{0DyM
zx^YN#W`c}OoPeXd-KpIo9Ipo!#V6Frgv$I*PWlAZr2?PxqpWvWgulA=I4B0ha0ZIS
zZ-SxjrMlO5+L|wivnLyL#Uq-cSZI-J6T$k7L1*xqj`iL4S0!0Ly~B@7_9nHObLYN9
zjkdNv4If!H_MN}k10KGg9_N4=ry`2W^54OHugsoAjtX?_sTx-6%QJ3^UDg#ZA@K;;
zX=;x+>$Z(_Eyd*r-`#H!V%La^7DVsgw$a_rwP|3Sf*ORBQ}WuU3h-3VbhLRpCut#8
zOAs&jNP`{E{fFy4Cni}&+7qECw=8Dw3JT3t9OOjP_tJOi%6#$Ke}i-s(ifpThxs4Y
z$Z$PJ62|4{mHAmbFmR<)?M#VvZ8tKm-zGM;1{4Bh{StGYkLKm(c!y(o6pDGjHmDF3gYSAY`L0ot9K*myx?
zdZU;}rOx`QjmNPfkn$j-UBZq_if<`Xq$i%unTKB5JJnTWU4D}FKqIBKT}LoIPo7hM
zwaD?UNweRq*g`LWN@iocxUJVJW6yqOOXcXZT<1brB+hSJI+NW#C^2e(sji}b;e6Sd
zIvdHWL4s#UIX`>aNrDZe5$S*JyNy}K)5B20f062A@E-%J>3CLnYO}$
z-nv*jMaz89{mizAEVYmVL4f*xWaq11V{f}-c;!;U1F&7Qm1W9d;$f($?ELM+Sf}|5
zs`0{Hsh7((1$C6==WJQ@@{QwO`lw>Q<9#+&Is~7LcyK)D?}Bl8aF
zqg^(DC%gRs=W&G3*rouoa-Lr?Nrx$|@4h8AS{6PIvc!MQHl=BCGXpVAd8grcTa~A~
z1gJ$p6f`RG8Wb13t9Iqg#Z7lw7T{YEFg;Fs4|4huPp-X<^5zS7Vt^EydkF3>*0raT@{PbbI;Ag6>D
zN2)P^M#=^eXq9T)LcbJucVkYs`Gnp36ao1^!r13Q{@*)?APR}bQP&ucRyotQ@Of}y
zq%Bh&&bApmmi^A~
z+|XNz$(NW*>oxJJ@s#R{0SqV
z_svNU^A84Sqzq1sx((IM9`^y&!}ipXg|@}R9poLxxltn?*0Ua*XEc>L1J7^HcnPFN
z3kHqMi1_BLp_3AguH7*sZAK@e{=mb0QjC`>Z^f|ys@}M4Tml^svLlQip><414e;*A
zUTw>&&=IHFLwA*U2K_0=E*A4GjeAc2L5BowT;*|0gzAn^!qNVCcFny*={=5kv%Yj9
zIrnptTSBX3^`yQ&k69i2#Syx$>gUHuuiZyWw$N-dF9FZ*6;t>I<0n~mw(1O+$b&|hIE_iCdo-Amav;k-Yp{!xi(_G2rLvB{?&
zkqh?7@J-gEqLSY*UH7otJu2nJWmEhb=}KMYK#=7ys@)H%caArTF%MH$tp==)botR0
zkEp2NRnC8c38f7Qlph%8bx&UV#6xvcC#ys}i|4Un;Qgj!>2)n#VT)7!Q;z!2N&@i1
z_e*tkW0;~wk?*{dGmuIqZ(XdUSJtV`ate*s8u&h)od@~QOzsjbt3eR7LORhjZ2+|7
zthHOSQ{Gba>CQkWNo=Q`Y3N`=6Uo^<^v)z$U0IM>M^SNnh7!GXmzQd45m#X5lWC>|
zIdlKx>ef253s>mqI8h(wZ(Yf@@ut(svI-}V^O|9;kP*h6pE3&9<5Je=avxhg(6y{O
zcUMx5*PJfdreaJe;gv-lzUXw|cuVDctFq$-!~DxEfkHkT!kjNw2n^K+&uVgQRNj
zG!18^mOxrFp;MRi&txJ_{|rPIWG#{y1+qu1<|-`d9_*=C!ui_5lz$Kt6#DE5eI&gC
znY7nli!sjj^IgTZ+P0F+`>D)z=JiB0Nf$!e?f+#seGDvGr^dkD)Jle7i@{yTFm;)e6dt;rvVt1W^OTcbb8S7q|(NpGzRT
z2j)|{s6NiYfCGe&Y3?mR)yB7zV5g6I+$JnV1_P=QFABEtBKO`j22q3Mt=1-I=LG}-+0&t>q^aZ95bK8w0pD%WAdVV|IC{9{pYmNGG|~CJJ74AWW&PwU
zyp}|*40f*Taez5f8zBRwLwS{JG$89|0%Famwy=H;&jcXO77?(P1ZM?}ZpA;r<>|9R
z8a%60b{mm-?V-zeHwp^}kk$r>YGS^?O{feewQ3%kj(~um;|>`nS`79`=kPcckTDtS
zrvpz~{)R+)`E~z-X^F0NeXy%AVS+hVhsckT3;tJHu*beE{8p0}F7h(4&!mq*~me
zcVR3%7DU1Mfl9_ZK0rm*OPx`toDSP_2`DeqB)CXd38$E=hwYu2{&!l=j~-8SGXF*7
z(S1QGW0;2%5}BE=L?!JNvZZLz@K(FP{Nrpa^K^pz88IenKAi%E$ftgVhGXj
z;;EWok}|eu3Oq?AuOoI<-e+SX%lXhZmF1t|M@9K;Q@)M1Wma^Y1NiFC!&Ex*P6YZ#
z=a}0_)|35K22rJ%=AXo`pI)GKmpgh^*AIiKwkyXR5k?Bl@*XM;0%Vk(@;eLgHGD-J(2hAs6
zpQ=lBiKE9X0kXisD_k-2SHhd6iPTSWsnBft%#y8THjC#kMG>75qjrI*B*NTov8`l-
z5`V(>R8kefZox`h!%I5Hh^g(2p+G@+KuZyr;|~z>S8Nd-J$?Dm74Rr-KNfyo>0Z)X
zX-D=u`VsU4CIYK(>%Wd8Svt}eWTb*&g!rYGu$bU)UOg)6SPZdWT0|1h?RPQxhwo3a
z`duU+aGi&s7Bc);QV+6G-S`P$uG~gavVNdC&II!39C{jV@^lbJd6)S6`~O2rc7I#h
zc0%Da|JwZx$C~^*_FIb}obG=ut5IeS6FnLE=kpG*sFi$v&+sQ?@V1RJe&Kg%^<@qgVuEe5;P
zOT;Gw(}s?2he3&)%zPV5&!cw@+0$XHB+z|y$LGP0(-F9Ld%G`3haZ5HIL+vO*2t<3
zyzjR+_a?Qx-q7_`gO}#wZ$q$u{XtAVeN&saEyPgs(=TU-&Sk}
zvSo~5E9yhF%hag9F#iA>{H>3XY*r6osWbu$tJVXJk8-;xV_`$aG%kXqv9xFM)@kJF
zmp%`D4cD#yuT*McC(ELYk)4k76|lZwtaK1qWW+gM0~dfqnIoh`3Y07!g9^o}`1
zI%~W^R1uxC6z$3<&46~@j+uPN1Ayc}vP{R{dynNnBXAQSZcPI0b=-8wV{i8MHcL{YA7IG5CIWw33L^vmr`#h6%o?kA;U-X}x)c)pOSDgG<`sQNhAqy)Cc!
zj^yJ0o%$_1>!)&YZsdVQdL7P}8qNc6A30jVQP-10ib#Cy~I}
zZkWk@6V_*UlWTO-HcWV|z*V|qq0A!nuz(t7y)usvhZ)aIH=OxnhKiiuA?Gt?5(zltVH
zYBK@&ZeeDnvZm8nc2uFdDeMSH6JKlu_zNm7BQjt##em&=A=uvBzZI-i0&QI>j8c@F
zp?neht9%_Aqt5a|*;lIV>Z=B#J1ncrxsB*`O28!AB3ck*Vq?TEa<$
zV}auIl5>$msmh@g`ThEw#|q>MFq=y3uc&%F6y`BiSg9m9({(_RSh
z$;3`BoY?FRwQKAJy69PJZLM*Zs!fm}7A)404(fz95)2!2#>B_itJ&vHmyiVB-6?J+
z`@U?roXIn~-JA!f2mDxx+yw20u=?F>am5!Y+p@EBPlp;*d7$IoD4L|tzYtlTykw^u
z8I;p_GzzXnQ=<2xK`E9VKZ2g_3m5Xn+WB{kBZTf+3xEOvueL4*PluzD`4#s`REkOl
zoHR;+WfGJ2O?JnlojX_$+@~-rmFMN>PA#0%4ZKW^+pYC2WjIqnmRCY2iLEz`lU6sb
zv#mrwBt|7B9j&K-!oe4(qAAH}*2Dg^llzclb%*Rp-BUncx%0XRn+R1a>z4IwDdKL|
z0~GsOpI?F$k7q9H{t^bv4gWKC8@e`56&mlSe1yYLp9Ch^?|%
zIYOXKuF{S`HA?Ku%kOV_zidVVyP5cI^?bj(U)lGYh2@-|a`YH`s9gI{9JotU8Fg?<(yvj1s{<#}#
zOBCA2jA;1V;5NNxy}v72ZfW1Rz7m)Kc*GP}MM$E6x>1293vqb0h
z+Z{l2A+ssy>FJ|-bTJic@@1Uv=yIG|$FH(gM27e8oXChe&oOm8@&u{=^us&+Sgcxn
z?dx6VxUq7pVu7Kc%PiUnjcDz>=R%^+o;0qN9BTo!nGu08|BW)+epO-
z@z7z6BJWRb1w@~C4dIHsFi=FR=jDcM49Nun!aUuuEmJcYF5>P6)`1U6h7J@QqsaIZ
zAwApa>)6;p@+ymai)^Pi`Yo-_?kD~3pA_l#OqqsV9fPzL*pYPyw^mUI)IB-XO}R_d
zps1NS7M61nxd
zDSq*bEiYg1Eo{Ewc>Zb8TctD(&%LiBy3rqYL7N__RXaW(xvB^UH;i|xTCzx0dV
zk~$KPIOH4Y4xCUJOS;pHw$O^nz)PEVfwSfoXxSG5#emADArHV1$7Xa`n>RE=L(GP?
z$#h#>ZL%C+NhlZodWgv6x{Pz~Y{{jrE~&3c5*%N$Z5m>Mzv_Y8Ihk*E`#pU}raCJ$
z@(C3a;0W$Ia`xJFWR5i`@+ZZQ3S9TANL1a9R85Z7*sIMV6PoqK$zdz-`d`?nRY}Ae
zl17Y7RNPT&o1A3RF<)9c@BuPc70V;0)ysN2fq!Y3Y%%myOW%Ur630Xx
z;UxdFA~J1W%z!010rpe`)y|jHa4AD^sf#v=A$<4`sxiR_*1;Eh%X5kS5;mSD08lEX
z&zOd8vhVsn;*dk}%jN`xBwKyxT_4n6VWwDwhijD88e(Nae3Xc~RT2{{V}4+P=F&cfH7^sa!5r@lm}5
zL6APP-q-@sViY}_4{<-DeC_rLA0W`%l$}v|{!TYL)s6W(n^oU4pNpgI!tQ6mBP7jd
z@7!AR&2VV(vy&A?he*+n&nmq;Gh`Zq10}jLE@kV+>eq?>ogqH=fD=b(wwm3&RK4SN
zA_t?!sxAL*;5yPDiF%*UCVP1xwmniNMpOJl$9WC1gB`##Lf%avPyVwZYNYf
zD2@Y=Wl7q{ra9+E&BtodkXygwZ@Tpw0GnqUWc0g}N9ma1V9wvm9X}21LZ1t5sY$_~
zG;#eTQUc)_?%BIq*Cs2P!2EpCb9vz5gp=>ubBTlPh~V)}`@a%Al5Yzh5!?72m$2BK
zm$#U6kihbD4i4H+iG)+V^{135TA{^oY2s3$SoAos`C?j=N_SpdXYfLf?j!%FHq*4}
z(dP#og<dnZZ`xnmc@MaW!)<_DJifk
z<@A@x-SR!a)t?72`Q){epohF$yZ_^smNloAW&9>%YCxIU#+d%%J6iQ?KpOX^8y*Ht
zE6ZZHeLDMQd)6T
z*@{vE**5lxD6bEgBYN3{?0Tu{l2XHp<>`jk=RNjgQj~=+Y71mJ8ry{pnx)$C#&Y%*
zEEyE*Y8A6Rwt3yMi2(0|&plLWJH>3f=gi#ZZSHw5FVRlCS%DnB8$ksN@ONz!;KGHfN=A}^e9z5*E4)MwuQipKpu*^K*p+N
zt5vR0cmbL6y$rltw#cS+cHg$k~5cJZ;}GzVYqr;N#=jCcb0i|4?kc
zU>vkV7e3eO&8cv95(0{*2#MbZDaa@vx}rogi=I~@PXLC=Y>PHSI?lGB$%RLEhQz1w
zX@S)Mg^kY9&OkeJuPz(r8+K>?PP-f}$6F@M=N0Tg&RuP)SxBbZ+J0B@$d@yAZ0Q>c
zYotri)N_Nd_p(Mso;zcMwKjCyRr~(R`+qWV?hv)P$^)I;Q-*gs
za)Urv%2|WB(Dj~H)+@PP&kt}6P3x^Yf!9^Lk3$<52<~Rs`*fw_3JaZ{q4b5;gTkh^Bk%=
z-rX^J{0ghv=*~2o4;U*u-|2$%%;AEcI_@C|D0ZULyRx8q!YF#x@(6jDRLLr@5cQ46
z$Iljr>m4C4wWEm2);LX(&@5NyPoKUuFW1T_;eZ9oIrDpoUBclE->>a#r!>ibUfP@q
zpPcVVL_A_r8k#4P1N43E?E7&Tb}EohcKdMWKa;msJn^H0%{mAAI<~lM=kS
z0rk{q@z9nX-;k}|V-LHphaG)e`;LB*PGbw0<(`;3{=Lt&NfIz}VbQ8R=)UIcS523G
zuU`)?w8J|)nf$Q7KcaG#yEiAmktJ_URQHtla)(S6NJw`UDhbn%-G58QGH4&K6X<6j
zkJo1VhlCUp{u;?75ni2OK2FF?*V8n+1eTNncXIF40)zSD4l09r)7iF_rF(|rE4K4r
zs7x>oTFNX{zmYSRyt%x9bZoPq63ZR@?W5f9Sj(w?1@d`8W(N#v#nv
zqzCslK~iYDR>u|t%}kUr)_`05c>`K)o~H7^UZ;wQlC&D-XP1wu{)X!J3y0K9(DhqE
zI!iyfSw2{M)*(H&pK-FqmvaSi%)Rw2Ve0+%^21L19d#RP!c=b!Cl#9UvyDw~JkS}r
zro%zs0!q{+`E4a?j%+zaXThc7#l9QdvMKe5M(bBmlbd;ne>5_E#qsrq+dn4$
zoP0!?YNe7a#A`7Iy$|wf(m}=!Otc-#^O7A|ZBp)r;~t-A$SstQZIrz^;N)ca&`r36
z=64r?e#d8BdIzQ-EK*)ykop^jWR!xIwV;?Uk6Hc&%(Css*5=R;hXkkIyOY&c`^;7R
z1Rfv={I}muQ(ps$^Cl(ylDx~$GB7;4Tq#`{aR2@u%{(;G=80IsZdD6qaR*1X1hL2B
zZECF@IgF)lc_I_VZSOutz)TbmTzajMMA!~(;#mB@`O>?*HKCOVd;|
znSW|`Z&AaVkJLGuCfCn%)oXP#3*mDKu09mWE@$owR>?Ecrl>eHCZ6&OI>Ej(F2%7i
z>xu0reW`8%SIwSht3lr}ng>fX1CUPpv2lR|>l5+B3)~Ziq9-SOZW9k3dUyZFV_U8y
zX5|urn~?N53c_C=q??samGG{AcZ%OexQh@wG&DqPY0)f!fA>>x|JukUZ`y2A7%0fg
z%WDhcyQ~nMxsB!Ur;BXf-|{}f4>{S>UyUY=W_g)6!X0Y^1)H{~r)L+k%?M+bRG+^%
z@;NDrowdz)kUWeXm?w*H$@4fIj?xl-Az9{qaIC%W&lji`&sv5HGN!7-`r$lq
z4n{+`O`^Dj1UX`>)foOxOxtUQQBXMz)7Y?BY_?wRep3>X(R?*WfPMTtk}~M0z(FEF
zK3paJB;pB)Z#i+A#1=^_kVq-N&@Cc~iP8f8f=uWz>5;QMv-O9`MMdY(DAX2}=VIgN
zs0Mu9)~^h1|H^;+v;W(FcpYqewxBcEI{p{`(eeK$z9SB`X_>HfVM0pkT6ZmZUD>)?
zZ-4Z!IC`)paBf`+B<}0?g^7Jz7wYCQSnc4f){vWCC>4_xK!~A?5cjH`lmmauN@6`o$l!9(8+C|eJE(3^yNi<
z^cjQ=!
z-RHh69`gfd^y>F59;>9XDC`IgLa7}rQV*38Jkn=@ASVxbhGP6iE{D
zAgrQ)Z0GxH?_%E}l*XHs;JfKqMT#R9!gNF5U>)H+mS!_?^!=Zo3Kd+uBWs=q3yEgU
zh2Vu&{ZPeFL`iUrd#*k}l&+TBh>_N_i1{ZmElEB!NYWem(SiS~u`7*gB1@z0OSek1
zGmU@;`ve09h=_pcBs3ZrwqOJat3jehb{#|z2+?Nnu(UWd1Vl7p3xOa=lx7_^B`%E-
zNP?^;3PMX*B1;lNAek!YK6B>Z{H*g{y?W=oTi?A^-~HbEvngUL0l57C-qP$L>&yI}
zQT{Z3?PmDNcpN=q-j?8rIKLjI0L>~NNzyD5{Q<2*67N0udmxWMP=9*b_yoQ7c1)o7
z%2kqkRYnUIMAc`ph7LNrsR3PI`K3t8Mr76j1hB>a$8Rlf#;mSbam@p_TRG8WlXiOY
z#B=!3yzK9h$3CQR8DZ|q>S?9niTpO8duPee%GsoA9rN1yI`Kl|IK0i&FS(Ekn_hd=
zy2QqcirnnePM5{N9dHxo}hySISQGW#Oq&*hSi!xCX^%>@tDzy
zL6j|NKI!7AeiONhz8`lFqB@@pie>SKns^l{5d?@mA!ZGZoQ85z$HH(amRp*p&&Y>!Ic;kB4CX@`{2p5VkAS6>isgB42o(uMwuVKl
z_O$v|s9WW+VLhpX%x4w!K5f?83MUFvQ@Em$ETOBMTV`V^_#JSP7yTg!Z&;wzM}e1B
zsZ_f9y+@LOi4~{7)GB`18*(+jd@qqNr6*H{l~bmX4u2qX`DOtHdfysH`!CL2?|Ph@
zxQ!a2p1o~4G>v3t!_Leh5d7!p7KCS^I&B;=8Xu=8Uok8QQVuNUO3P{ih>a-w4OP(K
z8=sbstJkJ4#kd8IB6$1hi4>f7#bTcZ&C-cl3HO%1ibpBBZ1{w=2?`^h&7EGGH+f4*
zmcKzUZ7}xQqs5k}na;37z*GgPMeR?IBL6LGHuW18-C%sVH|G&go}ust)#RvtJlSg7
z1HY}_I}r=b_G!@erCQ=Eeeyyr?%=|Ow-V6;mj-h-Oa3BTF4UkuQeEzM`Mpuu{Y`;3
zq#>_{oR6w+-xC!|^@y>-((-7dg8T-_tjC~A+>8m)?($4O19@KPPEjqRy1>oZ1+55m
zl&5&Cjnsj_yeqmH$`(&bIutLSlITMUcA$ZL0XX=om>!wdYR$0qXP^{N9_5h>%iu*LFgVZ#Nhz^2E`mq=84La?ADBgCvWA^JE7Pi{9W
z-*1<0;ID~lN7l?6MhAcWa5xmDUmarXuMA@E45w6wflIHs_P$2+_ZhmvHQ?%-BvIY=
zOZ*;0W?5YsDngy6!POrdBT3x_w7=OkD2l~>g0egNh)$6%ET{ZNt)!9XNQ)CXbUau&
zam)F}=%BIXyyoj=z0n-gpyn$w=%wDGtdnqDQUwdP72uNsBKl1y4tqH6y;z_3F)1Y>6HwA
z)~SPcja)9G6>An8^P;BlW;uHpy}Dq*L5BAUs_*P4nqKr*HsbYrVfrH8v+L&AvtpAb
zpXv)6W8WfoxktK+Vc1idWmRR)GlM7!#tZtdJtWo5;TF@#4Al_Lr>1rLGr>g
zlcOV=veP(Gb1W9tPB6*}_Il9LLhapd{&Tj$Y2j-ys9v=XJ^!Z5@UcTeL_`F@Uw2!N
zfng}bdQ)ip#anfNm#rT7ve=|t?4NoC2(>Qv41*uqVF_>u`~)~$C8Yj7-uQ8ewJ#8$
z9c&;>-3S^i&?nR*G!kh3cxHXf2MD>g8`1b#$o$*-&+P-3vfTa()hvbne{WA)A~82O
zDE^6gQaU>wp1}s7bnBFKrr;e`Qgnh5$wtGS;l8ObK=Xk_RjXc1-BuV4-^KB}eh>75
z9?Zra_FkQ+w!$N(VBJIA1{}Zr4FGpe#_IhK0b*XKk!TMgo(6&KH8HrycH!OK@5@7}
zeQ^a$e}1-w2;isA62&wv??jYm*gC(6Z!&hxYgJ-Kvl4w5it*S1x4|MUjnvtB&Ihar
z7aOkhmkI=eJoGPBrw?5leEN#Vk{a^!W%wT9svZ-9VGAoYi5o|l#xSE=&$14{+h@84
zrd2M+?&!7tmx6{i_|(`s$L;=+_sWtbQpL_C8MliFd%XhZOcWa|Z>XJwq^RMXv#FP@
zcIzOTvQezaeZTG@hW{2I+&K53vooXaLa4hN06`-iq9*a!&`%yQzM!6TG!hRz;>{zW
zb5VW%lo}{Q^M&{KwxzQM+gMci=CW?pTj&A*3>ImD{SVB|&0!XVo0Vb)A%J1)D1B4s
zZI1?amYvMqm`;czptl)7{z|8peQr}9S>MXHBAoxHbvXFgE&J+#JfiGzX4NLHb~)~V
z%S9B*@JJ&DD#*7C4wwxFMZm|en0Hz?-FLtTWaZ}RdtX#wA0;jUX^RIyrV_wz3lGR~
zddJ4b(lx-W^4BR-FTgt&2pC(LJfKo4^U!6$HpyuXh!Rzli_h<{u^YR33V(b&>slI+
zBh{0$f~!W>H!2Fm{=Z#esK=?S4`z|m(qSRqOLW&xIHdh@xgCGoW0Jegpait;oatqK
z$7ocM+fwhV$;hI%(-
W!up5Z<<~*}-QtSDp00HY`T0K|gRg!7
diff --git a/EdgeCraftRAG/docker_compose/intel/gpu/arc/README.md b/EdgeCraftRAG/docker_compose/intel/gpu/arc/README.md
index 9b9ab4eb3d..ae187648bc 100755
--- a/EdgeCraftRAG/docker_compose/intel/gpu/arc/README.md
+++ b/EdgeCraftRAG/docker_compose/intel/gpu/arc/README.md
@@ -14,11 +14,9 @@ This section describes how to quickly deploy and test the EdgeCraftRAG service m
1. [Prerequisites](#1-prerequisites)
2. [Access the Code](#2-access-the-code)
-3. [Prepare models](#3-prepare-models)
-4. [Prepare env variables and configurations](#4-prepare-env-variables-and-configurations)
-5. [Deploy the Service on Arc GPU Using Docker Compose](#5-deploy-the-service-on-intel-gpu-using-docker-compose)
-6. [Access UI](#6-access-ui)
-7. [Cleanup the Deployment](#7-cleanup-the-deployment)
+3. [Run quick_start.sh](#3-run-quick_startsh)
+4. [Access UI](#4-access-ui)
+5. [Cleanup the Deployment](#5-cleanup-the-deployment)
### 1. Prerequisites
@@ -54,130 +52,46 @@ cd GenAIExamples/EdgeCraftRAG
>git checkout v1.5
>```
-### 3. Prepare models
+### 3. Run quick_start.sh
-There are 3 models need to be prepared: **Embedding**, **Reranking**, **LLM**
-You'll need to decide the inferencing framework for these models.
-
-#### Embedding and Reranking
-
-Embedding and reranking are usually servered by local OpenVINO inferencing, to prepare these 2 models:
-
-```bash
-# Prepare models for embedding, reranking:
-export MODEL_PATH="${PWD}/models" # Your model path for embedding, reranking and LLM models
-mkdir -p $MODEL_PATH
-pip install --upgrade --upgrade-strategy eager "optimum[openvino]"
-optimum-cli export openvino -m BAAI/bge-small-en-v1.5 ${MODEL_PATH}/BAAI/bge-small-en-v1.5 --task sentence-similarity
-optimum-cli export openvino -m BAAI/bge-reranker-large ${MODEL_PATH}/BAAI/bge-reranker-large --task text-classification
-```
-
-#### LLM
-
-##### openVINO
-If you have Core Ultra platform only, please prepare openVINO models:
-You can also run openVINO models on discrete GPU.
-
-```bash
-# Prepare LLM model for openVINO
-optimum-cli export openvino --model Qwen/Qwen3-8B ${MODEL_PATH}/Qwen/Qwen3-8B/INT4_compressed_weights --task text-generation-with-past --weight-format int4 --group-size 128 --ratio 0.8
-```
-
-##### vLLM
-Alternatively, if you have discrete GPU and want to use vLLM, please prepare models for vLLM:
+Run quick start script from the `EdgeCraftRAG` root directory:
```bash
-# Prepare LLM model for vLLM
-export LLM_MODEL="Qwen/Qwen3-8B" # Your model id
-pip install modelscope
-modelscope download --model $LLM_MODEL --local_dir "${MODEL_PATH}/${LLM_MODEL}"
-# Optionally, you can also download models with huggingface:
-# pip install -U huggingface_hub
-# huggingface-cli download $LLM_MODEL --local-dir "${MODEL_PATH}/${LLM_MODEL}"
+./tools/quick_start.sh
```
-### 4. Prepare env variables and configurations
-
-#### Prepare env variables for deployment
+By default, this script starts local OpenVINO deployment when no environment variables are configured.
-```bash
-ip_address=$(hostname -I | awk '{print $1}')
-# Use `ip a` to check your active ip
-export HOST_IP=$ip_address # Your host ip
-
-# Check group id of video and render
-export VIDEOGROUPID=$(getent group video | cut -d: -f3)
-export RENDERGROUPID=$(getent group render | cut -d: -f3)
-
-# If you have a proxy configured, execute below line
-export no_proxy=${no_proxy},${HOST_IP},edgecraftrag,edgecraftrag-server
-export NO_PROXY=${NO_PROXY},${HOST_IP},edgecraftrag,edgecraftrag-server
-# If you have a HF mirror configured, it will be imported to the container
-# export HF_ENDPOINT=https://hf-mirror.com # your HF mirror endpoint"
-
-# Make sure all 3 folders have 1000:1000 permission, otherwise
-export DOC_PATH=${PWD}/tests
-export TMPFILE_PATH=${PWD}/tests
-chown 1000:1000 ${MODEL_PATH} ${DOC_PATH} ${TMPFILE_PATH}
-# In addition, also make sure the .cache folder has 1000:1000 permission, otherwise
-chown 1000:1000 -R $HOME/.cache
-```
+If you prefer manual model preparation, env setup, and docker compose options, please refer to [Manual deployment details in Advanced Setup](../../../../docs/Advanced_Setup.md#manual-deployment-details-for-arc-platform).
-For more advanced env variables and configurations, please refer to [Prepare env variables for vLLM deployment](../../../../docs/Advanced_Setup.md#prepare-env-variables-for-vllm-deployment)
+### 4. Access UI
-Set Milvus DB and chat history round for inference:
-
-```bash
-# EC-RAG support Milvus as persistent database, by default milvus is disabled, you can choose to set MILVUS_ENABLED=1 to enable it
-export MILVUS_ENABLED=0
-# If you enable Milvus, the default storage path is PWD, uncomment if you want to change:
-# export DOCKER_VOLUME_DIRECTORY= # change to your preference
-
-# EC-RAG support chat history round setting, by default chat history is disabled, you can set CHAT_HISTORY_ROUND to control it
-# export CHAT_HISTORY_ROUND= # change to your preference
-```
-
-### 5. Deploy the Service on Intel GPU Using Docker Compose
-
-#### Option a. Deploy openVINO LLM based EC-RAG for Core Ultra, Arc B60, Arc A770
-
-Make sure you have prepared [openVINO models](#openvino)
-```bash
-docker compose -f docker_compose/intel/gpu/arc/compose.yaml up -d
-```
-
-#### Option b.1. Deploy vLLM based EC-RAG for Arc B60
-Make sure you have prepared [vLLM models](#vllm)
-
-```bash
-docker compose --profile b60 -f docker_compose/intel/gpu/arc/compose.yaml up -d
-```
+Open your browser, access http://${HOST_IP}:8082
-#### Option b.2. Deploy vLLM based EC-RAG for Arc A770
-Make sure you have prepared [vLLM models](#vllm)
+After startup completes, `quick_start.sh` will print:
-```bash
-docker compose --profile a770 -f docker_compose/intel/gpu/arc/compose.yaml up -d
+```text
+Service launched successfully.
+UI access URL: http://${HOST_IP}:8082
+If you are accessing from another machine, replace ${HOST_IP} with your server's reachable IP or hostname.
```
-### 6. Access UI
-
-Open your browser, access http://${HOST_IP}:8082
-
> Your browser should be running on the same host of your console, otherwise you will need to access UI with your host domain name instead of ${HOST_IP}.
Below is the UI front page, for detailed operations on UI and EC-RAG settings, please refer to [Explore_Edge_Craft_RAG](../../../../docs/Explore_Edge_Craft_RAG.md)

-### 7. Cleanup the Deployment
+### 5. Cleanup the Deployment
-To stop the containers associated with the deployment, execute the following command:
+To stop the containers associated with the deployment, execute the helper script command:
+```bash
+./tools/quick_start.sh cleanup
```
-docker compose -f docker_compose/intel/gpu/arc/compose.yaml down
-```
-All the EdgeCraftRAG containers will be stopped and then removed on completion of the "down" command.
+All the EdgeCraftRAG containers will be stopped and then removed on completion.
+
+If you prefer the manual docker compose cleanup command, please refer to [Manual cleanup details in Advanced Setup](../../../../docs/Advanced_Setup.md#6-cleanup-the-deployment-manual).
## EdgeCraftRAG Docker Compose Files
diff --git a/EdgeCraftRAG/docker_compose/intel/gpu/arc/README_zh.md b/EdgeCraftRAG/docker_compose/intel/gpu/arc/README_zh.md
index ba73024aab..8162ae7ee1 100644
--- a/EdgeCraftRAG/docker_compose/intel/gpu/arc/README_zh.md
+++ b/EdgeCraftRAG/docker_compose/intel/gpu/arc/README_zh.md
@@ -14,11 +14,9 @@
1. [前置条件](#1-前置条件)
2. [获取代码](#2-获取代码)
-3. [准备模型](#3-准备模型)
-4. [准备环境变量和配置](#4-准备环境变量和配置)
-5. [使用 Docker Compose 在 Arc GPU 上部署服务](#5-使用-docker-compose-在-intel-gpu-上部署服务)
-6. [访问 UI](#6-访问-ui)
-7. [清理部署](#7-清理部署)
+3. [运行 quick_start.sh](#3-运行-quick_startsh)
+4. [访问 UI](#4-访问-ui)
+5. [清理部署](#5-清理部署)
### 1. 前置条件
@@ -54,133 +52,46 @@ cd GenAIExamples/EdgeCraftRAG
>git checkout v1.5
>```
-### 3. 准备模型
+### 3. 运行 quick_start.sh
-需要准备 3 个模型:**Embedding**、**Reranking**、**LLM**。
-你需要为这些模型选择推理框架。
-
-#### Embedding 和 Reranking
-
-Embedding 和 reranking 通常由本地 OpenVINO 推理提供服务。准备这 2 个模型:
-
-```bash
-# 准备 embedding、reranking 模型:
-export MODEL_PATH="${PWD}/models" # embedding、reranking、LLM 模型路径
-mkdir -p $MODEL_PATH
-pip install --upgrade --upgrade-strategy eager "optimum[openvino]"
-optimum-cli export openvino -m BAAI/bge-small-en-v1.5 ${MODEL_PATH}/BAAI/bge-small-en-v1.5 --task sentence-similarity
-optimum-cli export openvino -m BAAI/bge-reranker-large ${MODEL_PATH}/BAAI/bge-reranker-large --task text-classification
-```
-
-#### LLM
-
-##### openVINO
-如果你只有 Core Ultra 平台,请准备 openVINO 模型:
-你也可以在独立 GPU 上运行 openVINO 模型。
-
-```bash
-# 准备 openVINO 的 LLM 模型
-optimum-cli export openvino --model Qwen/Qwen3-8B ${MODEL_PATH}/Qwen/Qwen3-8B/INT4_compressed_weights --task text-generation-with-past --weight-format int4 --group-size 128 --ratio 0.8
-```
-
-##### vLLM
-另外,如果你有独立显卡,可以为 vLLM 准备模型:
-
-```bash
-# 准备 vLLM 的 LLM 模型
-export LLM_MODEL="Qwen/Qwen3-8B" # 你的模型 id
-pip install modelscope
-modelscope download --model $LLM_MODEL --local_dir "${MODEL_PATH}/${LLM_MODEL}"
-# 可选:你也可以用 huggingface 下载模型:
-# pip install -U huggingface_hub
-# huggingface-cli download $LLM_MODEL --local-dir "${MODEL_PATH}/${LLM_MODEL}"
-```
-
-### 4. 准备环境变量和配置
-
-#### 为部署准备环境变量
-
-```bash
-ip_address=$(hostname -I | awk '{print $1}')
-# 使用 `ip a` 查看当前活动 ip
-export HOST_IP=$ip_address # 主机 ip
-
-# 查看 video 与 render 的 group id
-export VIDEOGROUPID=$(getent group video | cut -d: -f3)
-export RENDERGROUPID=$(getent group render | cut -d: -f3)
-
-# 若已配置代理,执行以下命令
-export no_proxy=${no_proxy},${HOST_IP},edgecraftrag,edgecraftrag-server
-export NO_PROXY=${NO_PROXY},${HOST_IP},edgecraftrag,edgecraftrag-server
-# 如果配置了 HF 镜像,会传入容器
-# export HF_ENDPOINT=https://hf-mirror.com # 你的 HF 镜像地址"
-
-# 确保以下 3 个文件夹权限为 1000:1000,否则
-export DOC_PATH=${PWD}/tests
-export TMPFILE_PATH=${PWD}/tests
-chown 1000:1000 ${MODEL_PATH} ${DOC_PATH} ${TMPFILE_PATH}
-# 此外还要确保 .cache 文件夹权限为 1000:1000,否则
-chown 1000:1000 -R $HOME/.cache
-```
-
-如需更高级的环境变量和配置,请参考 [Prepare env variables for vLLM deployment](../../../../docs/Advanced_Setup.md#prepare-env-variables-for-vllm-deployment)
-
-为推理设置 Milvus 数据库与聊天历史轮数:
+在 `EdgeCraftRAG` 根目录下运行快速启动脚本:
```bash
-# EC-RAG 支持 Milvus 持久化数据库,默认关闭;可设置 MILVUS_ENABLED=1 开启
-export MILVUS_ENABLED=0
-# 如果启用 Milvus,默认存储路径为 PWD,如需修改请取消注释:
-# export DOCKER_VOLUME_DIRECTORY= # 按需修改
-
-# EC-RAG 支持聊天历史轮数设置,默认关闭;可通过 CHAT_HISTORY_ROUND 控制
-# export CHAT_HISTORY_ROUND= # 按需修改
+./tools/quick_start.sh
```
-### 5. 使用 Docker Compose 在 Intel GPU 上部署服务
-
-#### Option a. 为 Core Ultra、Arc B60、Arc A770 部署基于 openVINO LLM 的 EC-RAG
-
-请确保已准备好 [openVINO 模型](#openvino)
+在不配置任何环境变量时,脚本默认启动本地 OpenVINO 部署。
-```bash
-docker compose -f docker_compose/intel/gpu/arc/compose.yaml up -d
-```
-
-#### Option b.1. 为 Arc B60 部署基于 vLLM 的 EC-RAG
+如果你希望使用手动方式(模型准备、环境变量配置、Docker Compose 启动),请参考 [Advanced Setup 中的手动部署说明](../../../../docs/Advanced_Setup_zh.md#arc-平台手动部署详细说明)。
-请确保已准备好 [vLLM 模型](#vllm)
+### 4. 访问 UI
-```bash
-docker compose --profile b60 -f docker_compose/intel/gpu/arc/compose.yaml up -d
-```
-
-#### Option b.2. 为 Arc A770 部署基于 vLLM 的 EC-RAG
+打开浏览器访问 http://${HOST_IP}:8082
-请确保已准备好 [vLLM 模型](#vllm)
+启动完成后,`quick_start.sh` 会输出:
-```bash
-docker compose --profile a770 -f docker_compose/intel/gpu/arc/compose.yaml up -d
+```text
+Service launched successfully.
+UI access URL: http://${HOST_IP}:8082
+If you are accessing from another machine, replace ${HOST_IP} with your server's reachable IP or hostname.
```
-### 6. 访问 UI
-
-打开浏览器访问 http://${HOST_IP}:8082
-
> 浏览器应运行在与控制台相同的主机上;否则你需要使用主机域名而不是 ${HOST_IP} 来访问 UI。
-下图为 UI 首页。有关 UI 操作和 EC-RAG 设置的详细说明,请参考 [Explore_Edge_Craft_RAG](../../../../docs/Explore_Edge_Craft_RAG.md)
+下图为 UI 首页。有关 UI 操作和 EC-RAG 设置的详细说明,请参考 [Explore_Edge_Craft_RAG](../../../../docs/Explore_Edge_Craft_RAG_zh.md)

-### 7. 清理部署
+### 5. 清理部署
-若要停止与本次部署关联的容器,请执行以下命令:
+若要停止与本次部署关联的容器,请执行脚本命令:
+```bash
+./tools/quick_start.sh cleanup
```
-docker compose -f docker_compose/intel/gpu/arc/compose.yaml down
-```
-执行完 `down` 命令后,所有 EdgeCraftRAG 容器都会停止并被移除。
+执行完成后,所有 EdgeCraftRAG 容器都会停止并被移除。
+
+如果你希望使用手动 docker compose 清理命令,请参考 [Advanced Setup 中的手动清理说明](../../../../docs/Advanced_Setup_zh.md#6-清理部署手动)。
## EdgeCraftRAG Docker Compose 文件
diff --git a/EdgeCraftRAG/docs/API_Guide.md b/EdgeCraftRAG/docs/API_Guide.md
index 1845d0ba36..2b0c318bc6 100644
--- a/EdgeCraftRAG/docs/API_Guide.md
+++ b/EdgeCraftRAG/docs/API_Guide.md
@@ -1,50 +1,77 @@
-# Edge Craft Retrieval-Augmented Generation API guide
+# Edge Craft Retrieval-Augmented Generation API Guide
+
+> **Base URLs**
+> - EC-RAG Server: `http://${HOST_IP}:16010`
+> - EC-RAG Mega Service: `http://${HOST_IP}:16011`
+
+---
## Pipeline Management
### Create a pipeline
```bash
-curl -X POST http://${HOST_IP}:16010/v1/settings/pipelines -H "Content-Type: application/json" -d @tests/test_pipeline_local_llm.json | jq '.'
+curl -X POST http://${HOST_IP}:16010/v1/settings/pipelines \
+ -H "Content-Type: application/json" \
+ -d @tests/test_pipeline_local_llm.json | jq '.'
```
### Update a pipeline
```bash
-curl -X PATCH http://${HOST_IP}:16010/v1/settings/pipelines/rag_test_local_llm -H "Content-Type: application/json" -d @tests/test_pipeline_local_llm.json | jq '.'
+curl -X PATCH http://${HOST_IP}:16010/v1/settings/pipelines/{name} \
+ -H "Content-Type: application/json" \
+ -d @tests/test_pipeline_local_llm.json | jq '.'
```
### Check all pipelines
```bash
-curl -X GET http://${HOST_IP}:16010/v1/settings/pipelines -H "Content-Type: application/json" | jq '.'
+curl -X GET http://${HOST_IP}:16010/v1/settings/pipelines \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+### Check a specific pipeline
+
+```bash
+curl -X GET http://${HOST_IP}:16010/v1/settings/pipelines/{name} \
+ -H "Content-Type: application/json" | jq '.'
```
### Activate a pipeline
```bash
-curl -X PATCH http://${HOST_IP}:16010/v1/settings/pipelines/rag_test_local_llm -H "Content-Type: application/json" -d '{"active": "true"}' | jq '.'
+curl -X PATCH http://${HOST_IP}:16010/v1/settings/pipelines/{name} \
+ -H "Content-Type: application/json" \
+ -d '{"active": "true"}' | jq '.'
```
### Remove a pipeline
```bash
# Firstly, deactivate the pipeline if the pipeline status is active
-curl -X PATCH http://${HOST_IP}:16010/v1/settings/pipelines/rag_test_local_llm -H "Content-Type: application/json" -d '{"active": "false"}' | jq '.'
+curl -X PATCH http://${HOST_IP}:16010/v1/settings/pipelines/{name} \
+ -H "Content-Type: application/json" \
+ -d '{"active": "false"}' | jq '.'
+
# Then delete the pipeline
-curl -X DELETE http://${HOST_IP}:16010/v1/settings/pipelines/rag_test_local_llm -H "Content-Type: application/json" | jq '.'
+curl -X DELETE http://${HOST_IP}:16010/v1/settings/pipelines/{name} \
+ -H "Content-Type: application/json" | jq '.'
```
-### Get pipeline json
+### Get pipeline JSON
```bash
-curl -X GET http://${HOST_IP}:16010/v1/settings/pipelines/{name}/json -H "Content-Type: application/json" | jq '.'
+curl -X GET http://${HOST_IP}:16010/v1/settings/pipelines/{name}/json \
+ -H "Content-Type: application/json" | jq '.'
```
-### Import pipeline from a json file
+### Import pipeline from a JSON file
```bash
-curl -X POST http://${HOST_IP}:16010/v1/settings/pipelines/import -H "Content-Type: multipart/form-data" -F "file=@your_test_pipeline_json_file.txt"| jq '.'
+curl -X POST http://${HOST_IP}:16010/v1/settings/pipelines/import \
+ -H "Content-Type: multipart/form-data" \
+ -F "file=@your_pipeline.json" | jq '.'
```
### Enable and check benchmark for pipelines
@@ -56,19 +83,28 @@ Benchmarking activities may significantly reduce system performance.
**DO NOT** perform benchmarking in a production environment.
```bash
-# Set ENABLE_BENCHMARK as true before launch services
+# Set ENABLE_BENCHMARK as true before launching services
export ENABLE_BENCHMARK="true"
-# check the benchmark data for pipeline {pipeline_name}
-curl -X GET http://${HOST_IP}:16010/v1/settings/pipelines/{pipeline_name}/benchmarks -H "Content-Type: application/json" | jq '.'
+# Check the benchmark data for the active pipeline
+curl -X GET http://${HOST_IP}:16010/v1/settings/pipeline/benchmark \
+ -H "Content-Type: application/json" | jq '.'
+
+# Check the benchmark data for a specific pipeline
+curl -X GET http://${HOST_IP}:16010/v1/settings/pipelines/{pipeline_name}/benchmarks \
+ -H "Content-Type: application/json" | jq '.'
```
+---
+
## Model Management
### Load a model
```bash
-curl -X POST http://${HOST_IP}:16010/v1/settings/models -H "Content-Type: application/json" -d '{"model_type": "reranker", "model_id": "BAAI/bge-reranker-large", "model_path": "./models/bge_ov_reranker", "device": "cpu", "weight": "INT4"}' | jq '.'
+curl -X POST http://${HOST_IP}:16010/v1/settings/models \
+ -H "Content-Type: application/json" \
+ -d '{"model_type": "reranker", "model_id": "BAAI/bge-reranker-large", "model_path": "./models/bge_ov_reranker", "device": "cpu", "weight": "INT4"}' | jq '.'
```
It will take some time to load the model.
@@ -76,108 +112,432 @@ It will take some time to load the model.
### Check all models
```bash
-curl -X GET http://${HOST_IP}:16010/v1/settings/models -H "Content-Type: application/json" | jq '.'
+curl -X GET http://${HOST_IP}:16010/v1/settings/models \
+ -H "Content-Type: application/json" | jq '.'
```
-### Update a model
+### Check a specific model
```bash
-curl -X PATCH http://${HOST_IP}:16010/v1/settings/models/BAAI/bge-reranker-large -H "Content-Type: application/json" -d '{"model_type": "reranker", "model_id": "BAAI/bge-reranker-large", "model_path": "./models/bge_ov_reranker", "device": "gpu", "weight": "INT4"}' | jq '.'
+curl -X GET http://${HOST_IP}:16010/v1/settings/models/BAAI/bge-reranker-large \
+ -H "Content-Type: application/json" | jq '.'
```
-### Check a certain model
+### Update a model
```bash
-curl -X GET http://${HOST_IP}:16010/v1/settings/models/BAAI/bge-reranker-large -H "Content-Type: application/json" | jq '.'
+curl -X PATCH http://${HOST_IP}:16010/v1/settings/models/BAAI/bge-reranker-large \
+ -H "Content-Type: application/json" \
+ -d '{"model_type": "reranker", "model_id": "BAAI/bge-reranker-large", "model_path": "./models/bge_ov_reranker", "device": "gpu", "weight": "INT4"}' | jq '.'
```
### Delete a model
```bash
-curl -X DELETE http://${HOST_IP}:16010/v1/settings/models/BAAI/bge-reranker-large -H "Content-Type: application/json" | jq '.'
+curl -X DELETE http://${HOST_IP}:16010/v1/settings/models/BAAI/bge-reranker-large \
+ -H "Content-Type: application/json" | jq '.'
```
+### Get available model weights
+
+Query the available compression weights (INT4 / INT8 / FP16) for a given model path:
+
+```bash
+curl -X GET "http://${HOST_IP}:16010/v1/settings/weight/BAAI/bge-reranker-large" \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+### Get available model IDs by type
+
+Supported `model_type` values: `LLM`, `vLLM`, `reranker`, `embedding`, `vLLM_embedding`, `kbadmin_embedding_model`
+
+```bash
+# List available local LLM models
+curl -X GET "http://${HOST_IP}:16010/v1/settings/avail-models/LLM" \
+ -H "Content-Type: application/json" | jq '.'
+
+# List models served by a vLLM server (optional server_address parameter)
+curl -X GET "http://${HOST_IP}:16010/v1/settings/avail-models/vLLM?server_address=http://localhost:8086" \
+ -H "Content-Type: application/json" | jq '.'
+
+# List available embedding models
+curl -X GET "http://${HOST_IP}:16010/v1/settings/avail-models/embedding" \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+### Get available models from a vLLM server
+
+```bash
+curl -X GET "http://${HOST_IP}:16010/v1/available_models?server_address=http://localhost:8086" \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+---
+
## Knowledge Base Management
### Create a knowledge base
```bash
-curl -X POST http://${HOST_IP}:16010/v1/knowledge -H "Content-Type: application/json" -d @tests/configs/test_kb.json | jq '.'
+curl -X POST http://${HOST_IP}:16010/v1/knowledge \
+ -H "Content-Type: application/json" \
+ -d @tests/configs/test_kb.json | jq '.'
```
-### Update a knowledge base
+### Check all knowledge bases
+
+```bash
+curl -X GET http://${HOST_IP}:16010/v1/knowledge \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+### Check a specific knowledge base
+
+```bash
+curl -X GET http://${HOST_IP}:16010/v1/knowledge/default_kb \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+### Get knowledge base JSON
+
+```bash
+curl -X GET http://${HOST_IP}:16010/v1/knowledge/default_kb/json \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+### Get knowledge base file map (paginated)
```bash
-curl -X PATCH http://${HOST_IP}:16010/v1/knowledge/patch -H "Content-Type: application/json" -d '{"name": "default_kb","active":"True","description": "Your knowledge base Description"}' | jq '.'
+curl -X GET "http://${HOST_IP}:16010/v1/knowledge/default_kb/filemap?page_num=1&page_size=20" \
+ -H "Content-Type: application/json" | jq '.'
```
-### Check all knowledge base
+### Update a knowledge base
```bash
-curl -X GET http://${HOST_IP}:16010/v1/knowledge -H "Content-Type: application/json" | jq '.'
+curl -X PATCH http://${HOST_IP}:16010/v1/knowledge/patch \
+ -H "Content-Type: application/json" \
+ -d '{"name": "default_kb", "active": "True", "description": "Your knowledge base description"}' | jq '.'
```
-### Activate knowledge base
+### Activate a knowledge base
```bash
-curl -X PATCH http://${HOST_IP}:16010/v1/knowledge/patch -H "Content-Type: application/json" -d '{"name": "default_kb","active":true}' | jq '.'
+curl -X PATCH http://${HOST_IP}:16010/v1/knowledge/patch \
+ -H "Content-Type: application/json" \
+ -d '{"name": "default_kb", "active": true}' | jq '.'
```
### Remove a knowledge base
```bash
-curl -X DELETE http://${HOST_IP}:16010/v1/knowledge/default_kb -H "Content-Type: application/json" | jq '.'
+curl -X DELETE http://${HOST_IP}:16010/v1/knowledge/default_kb \
+ -H "Content-Type: application/json" | jq '.'
```
### Add file to knowledge base
```bash
-curl -X POST http://${HOST_IP}:16010/v1/knowledge/default_kb/files -H "Content-Type: application/json" -d '{"local_path": "/home/user/ui_cache/#REPLACE WITH YOUR DIR WITHIN MOUNTED DOC PATH#"}' | jq '.'
+curl -X POST http://${HOST_IP}:16010/v1/knowledge/default_kb/files \
+ -H "Content-Type: application/json" \
+ -d '{"local_path": "/home/user/ui_cache/#REPLACE WITH YOUR FILE OR DIR PATH#"}' | jq '.'
```
-### Delete file to knowledge base
+### Delete file from knowledge base
```bash
-curl -X DELETE http://${HOST_IP}:16010/v1/knowledge/default_kb/files -H "Content-Type: application/json" -d '{"local_path": "/home/user/ui_cache/#REPLACE WITH YOUR DIR WITHIN MOUNTED DOC PATH#"}' | jq '.'
+curl -X DELETE http://${HOST_IP}:16010/v1/knowledge/default_kb/files \
+ -H "Content-Type: application/json" \
+ -d '{"local_path": "/home/user/ui_cache/#REPLACE WITH YOUR FILE PATH#"}' | jq '.'
```
+---
+
+## Experience Knowledge Base Management
+
+Experience knowledge bases store curated Q&A pairs that can be retrieved to augment pipeline responses.
+
+### Get all experiences
+
+```bash
+curl -X GET http://${HOST_IP}:16010/v1/experiences \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+### Get experience by ID or question
+
+```bash
+curl -X POST http://${HOST_IP}:16010/v1/experience \
+ -H "Content-Type: application/json" \
+ -d '{"idx": "your-experience-id"}' | jq '.'
+```
+
+### Update an experience
+
+```bash
+curl -X PATCH http://${HOST_IP}:16010/v1/experiences \
+ -H "Content-Type: application/json" \
+ -d '{"idx": "your-experience-id", "question": "Updated question?", "content": "Updated answer"}' | jq '.'
+```
+
+### Delete an experience
+
+```bash
+curl -X DELETE http://${HOST_IP}:16010/v1/experiences \
+ -H "Content-Type: application/json" \
+ -d '{"idx": "your-experience-id"}' | jq '.'
+```
+
+### Add experiences from a file
+
+```bash
+curl -X POST http://${HOST_IP}:16010/v1/experiences/files \
+ -H "Content-Type: application/json" \
+ -d '{"local_path": "/home/user/ui_cache/experiences.json"}' | jq '.'
+```
+
+### Check and add multiple experiences (duplicate check first)
+
+```bash
+# Step 1: Check for duplicates; if none, experiences are added automatically
+curl -X POST http://${HOST_IP}:16010/v1/multiple_experiences/check \
+ -H "Content-Type: application/json" \
+ -d '[{"question": "What is EC-RAG?", "content": "EdgeCraft RAG is ..."}]' | jq '.'
+
+# Step 2 (only if duplicates detected): Confirm with overwrite flag
+# flag=true → overwrite duplicates; flag=false → append
+curl -X POST "http://${HOST_IP}:16010/v1/multiple_experiences/confirm?flag=true" \
+ -H "Content-Type: application/json" \
+ -d '[{"question": "What is EC-RAG?", "content": "EdgeCraft RAG is ..."}]' | jq '.'
+```
+
+---
+
+## Agent Management
+
+### Check all agents
+
+```bash
+curl -X GET http://${HOST_IP}:16010/v1/agents \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+### Check a specific agent
+
+```bash
+curl -X GET http://${HOST_IP}:16010/v1/agents/{name} \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+### Get default configs for an agent type
+
+```bash
+curl -X GET http://${HOST_IP}:16010/v1/agents/configs/{agent_type} \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+### Create an agent
+
+```bash
+curl -X POST http://${HOST_IP}:16010/v1/agents \
+ -H "Content-Type: application/json" \
+ -d '{"name": "my_agent", "type": "react_llm", "pipeline_idx": "your-pipeline-idx"}' | jq '.'
+```
+
+### Update an agent
+
+```bash
+curl -X PATCH http://${HOST_IP}:16010/v1/agents/{name} \
+ -H "Content-Type: application/json" \
+ -d '{"name": "my_agent", "active": true}' | jq '.'
+```
+
+### Delete an agent
+
+```bash
+curl -X DELETE http://${HOST_IP}:16010/v1/agents/{name} \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+---
+
## System Prompt Management
### Get system prompt
```bash
-curl -X GET http://${HOST_IP}:16010/v1/chatqna/prompt -H "Content-Type: application/json" | jq '.'
+curl -X GET http://${HOST_IP}:16010/v1/chatqna/prompt \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+### Get tagged system prompt
+
+```bash
+curl -X GET http://${HOST_IP}:16010/v1/chatqna/prompt/tagged \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+### Get default system prompt
+
+```bash
+curl -X GET http://${HOST_IP}:16010/v1/chatqna/prompt/default \
+ -H "Content-Type: application/json" | jq '.'
```
### Update system prompt
```bash
-curl -X POST http://${HOST_IP}:16010/v1/chatqna/prompt -H "Content-Type: application/json" -d '{"prompt":"This is a template prompt"}' | jq '.'
+curl -X POST http://${HOST_IP}:16010/v1/chatqna/prompt \
+ -H "Content-Type: application/json" \
+ -d '{"prompt": "This is a custom prompt template"}' | jq '.'
+```
+
+### Upload system prompt from file
+
+```bash
+curl -X POST http://${HOST_IP}:16010/v1/chatqna/prompt-file \
+ -H "Content-Type: multipart/form-data" \
+ -F "file=@your_prompt_file.txt" | jq '.'
+```
+
+### Reset system prompt to default
+
+```bash
+curl -X POST http://${HOST_IP}:16010/v1/chatqna/prompt/reset \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+---
+
+## Data Management
+
+### Get all nodes (chunks) in the active knowledge base
+
+```bash
+curl -X GET http://${HOST_IP}:16010/v1/data/nodes \
+ -H "Content-Type: application/json" | jq '.'
```
-### Reset system prompt
+### Get nodes by document name
```bash
-curl -X POST http://${HOST_IP}:16010/v1/chatqna/prompt/reset -H "Content-Type: application/json" | jq '.'
+curl -X GET "http://${HOST_IP}:16010/v1/data/{document_name}/nodes" \
+ -H "Content-Type: application/json" | jq '.'
```
-### Use custom system prompt file
+### Get all document names in the active knowledge base
```bash
-curl -X POST http://${HOST_IP}:16010/v1/chatqna/prompt-file -H "Content-Type: multipart/form-data" -F "file=@your_prompt_file.txt"
+curl -X GET http://${HOST_IP}:16010/v1/data/documents \
+ -H "Content-Type: application/json" | jq '.'
```
+### Get all files
+
+```bash
+curl -X GET http://${HOST_IP}:16010/v1/data/files \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+### Get a specific file
+
+```bash
+curl -X GET http://${HOST_IP}:16010/v1/data/files/{name} \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+### Upload a file (from UI)
+
+```bash
+curl -X POST "http://${HOST_IP}:16010/v1/data/file/{file_name}" \
+ -H "Content-Type: multipart/form-data" \
+ -F "file=@/path/to/your/document.pdf" | jq '.'
+```
+
+---
+
+## Session Management
+
+### Get all sessions
+
+```bash
+curl -X GET http://${HOST_IP}:16010/v1/sessions \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+### Get a session by ID
+
+```bash
+curl -X GET http://${HOST_IP}:16010/v1/session/{session_id} \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+---
+
+## System Information
+
+### Get system status
+
+Returns CPU usage, memory usage, disk usage, OS info, and current time.
+
+```bash
+curl -X GET http://${HOST_IP}:16010/v1/system/info \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+### Get available inference devices
+
+Returns OpenVINO-available devices (e.g., CPU, GPU, AUTO).
+
+```bash
+curl -X GET http://${HOST_IP}:16010/v1/system/device \
+ -H "Content-Type: application/json" | jq '.'
+```
+
+---
+
## ChatQnA
### Retrieval API
+Retrieve relevant context chunks from the active knowledge base without running LLM generation.
+
+```bash
+curl -X POST http://${HOST_IP}:16010/v1/retrieval \
+ -H "Content-Type: application/json" \
+ -d '{"messages": "Your question here", "top_n": 5, "max_tokens": 512}' | jq '.'
+```
+
+### ChatQnA API (Mega Service)
+
+Send a question through the full RAG pipeline (retrieval + LLM generation).
+
```bash
-curl -X POST http://${HOST_IP}:16010/v1/retrieval -H "Content-Type: application/json" -d '{"messages":"#Please enter the question you need to retrieve here#", "top_n":5, "max_tokens":512}' | jq '.'
+curl -X POST http://${HOST_IP}:16011/v1/chatqna \
+ -H "Content-Type: application/json" \
+ -d '{"messages": "Your question here", "top_n": 5, "max_tokens": 512}' | jq '.'
+```
+
+### RAGQnA API (with contexts in response)
+Returns the LLM answer together with the retrieved context chunks.
+
+```bash
+# Non-streaming
+curl -X POST http://${HOST_IP}:16010/v1/ragqna \
+ -H "Content-Type: application/json" \
+ -d '{"messages": "Your question here", "top_n": 5, "max_tokens": 512, "stream": false}' | jq '.'
+
+# Streaming
+curl -X POST http://${HOST_IP}:16010/v1/ragqna \
+ -H "Content-Type: application/json" \
+ -d '{"messages": "Your question here", "top_n": 5, "max_tokens": 512, "stream": true}'
```
-### ChatQnA API
+### Check vLLM server connection
```bash
-curl -X POST http://${HOST_IP}:16011/v1/chatqna -H "Content-Type: application/json" -d '{"messages":"#REPLACE WITH YOUR QUESTION HERE#", "top_n":5, "max_tokens":512}' | jq '.'
+curl -X POST http://${HOST_IP}:16010/v1/check/vllm \
+ -H "Content-Type: application/json" \
+ -d '{"server_address": "http://localhost:8086", "model_name": "Qwen/Qwen3-8B"}' | jq '.'
```
diff --git a/EdgeCraftRAG/docs/Advanced_Setup.md b/EdgeCraftRAG/docs/Advanced_Setup.md
index 4d17e28983..7811a5d82f 100644
--- a/EdgeCraftRAG/docs/Advanced_Setup.md
+++ b/EdgeCraftRAG/docs/Advanced_Setup.md
@@ -1,34 +1,69 @@
# Edge Craft Retrieval-Augmented Generation Advanced Setup
-## OpenVINO Local Inference
+[中文版](Advanced_Setup_zh.md)
-EC-RAG support using local OpenVINO models to do inference, please follow below steps to run local inference:
+## Build Docker Images for Mega Service, Server and UI by your own
-### 1. (Optional) Build Docker Images for Mega Service, Server and UI by your own
-
-**All the docker images can be automatically pulled**, If you want to build the images by your own, please follow the steps:
+**All the docker images can be automatically pulled**, if you want to build the images by your own, please use helper script:
```bash
-git clone https://github.com/opea-project/GenAIExamples.git
-cd GenAIExamples/EdgeCraftRAG
-docker build --no-cache --pull --build-arg http_proxy=$http_proxy --build-arg https_proxy=$https_proxy --build-arg no_proxy="$no_proxy" -t opea/edgecraftrag:latest -f Dockerfile .
-docker build --no-cache --pull --build-arg http_proxy=$http_proxy --build-arg https_proxy=$https_proxy --build-arg no_proxy="$no_proxy" -t opea/edgecraftrag-server:latest -f Dockerfile.server .
-docker build --no-cache --pull --build-arg http_proxy=$http_proxy --build-arg https_proxy=$https_proxy --build-arg no_proxy="$no_proxy" -t opea/edgecraftrag-ui:latest -f ui/docker/Dockerfile.ui .
+# Build all images
+./tools/build_images.sh
+
+# Build a single image
+./tools/build_images.sh mega
+./tools/build_images.sh server
+./tools/build_images.sh ui
+
+# Build multiple selected images
+./tools/build_images.sh mega server
```
-### 2. Prepare models
+## Manual deployment details for Arc platform
+
+### Prepare models
+
+There are 3 models need to be prepared: **Embedding**, **Reranking**, **LLM**
+You'll need to decide the inferencing framework for these models.
+
+#### Embedding and Reranking
+
+Embedding and reranking are usually servered by local OpenVINO inferencing, to prepare these 2 models:
```bash
-# Prepare models for embedding, reranking and generation, you can also choose other OpenVINO optimized models
-export MODEL_PATH="${PWD}/ov_models" # Your model path for embedding, reranking and LLM models
+# Prepare models for embedding, reranking:
+export MODEL_PATH="${PWD}/models" # Your model path for embedding, reranking and LLM models
mkdir -p $MODEL_PATH
pip install --upgrade --upgrade-strategy eager "optimum[openvino]"
optimum-cli export openvino -m BAAI/bge-small-en-v1.5 ${MODEL_PATH}/BAAI/bge-small-en-v1.5 --task sentence-similarity
optimum-cli export openvino -m BAAI/bge-reranker-large ${MODEL_PATH}/BAAI/bge-reranker-large --task text-classification
+```
+
+#### LLM
+
+##### openVINO
+If you have Core Ultra platform only, please prepare openVINO models:
+You can also run openVINO models on discrete GPU.
+
+```bash
+# Prepare LLM model for openVINO
optimum-cli export openvino --model Qwen/Qwen3-8B ${MODEL_PATH}/Qwen/Qwen3-8B/INT4_compressed_weights --task text-generation-with-past --weight-format int4 --group-size 128 --ratio 0.8
```
-### 3. Prepare env variables and configurations
+##### vLLM
+Alternatively, if you have discrete GPU and want to use vLLM, please prepare models for vLLM:
+
+```bash
+# Prepare LLM model for vLLM
+export LLM_MODEL="Qwen/Qwen3-8B" # Your model id
+pip install modelscope
+modelscope download --model $LLM_MODEL --local_dir "${MODEL_PATH}/${LLM_MODEL}"
+# Optionally, you can also download models with huggingface:
+# pip install -U huggingface_hub
+# huggingface-cli download $LLM_MODEL --local-dir "${MODEL_PATH}/${LLM_MODEL}"
+```
+
+### Prepare env variables and configurations
```bash
ip_address=$(hostname -I | awk '{print $1}')
@@ -48,12 +83,12 @@ export NO_PROXY=${NO_PROXY},${HOST_IP},edgecraftrag,edgecraftrag-server
# Make sure all 3 folders have 1000:1000 permission, otherwise
export DOC_PATH=${PWD}/tests
export TMPFILE_PATH=${PWD}/tests
-chown 1000:1000 -R ${MODEL_PATH} ${DOC_PATH} ${TMPFILE_PATH}
+chown 1000:1000 ${MODEL_PATH} ${DOC_PATH} ${TMPFILE_PATH}
# In addition, also make sure the .cache folder has 1000:1000 permission, otherwise
chown 1000:1000 -R $HOME/.cache
```
-### 4. Start Edge Craft RAG Services with Docker Compose
+Set Milvus DB and chat history round for inference:
```bash
# EC-RAG support Milvus as persistent database, by default milvus is disabled, you can choose to set MILVUS_ENABLED=1 to enable it
@@ -63,25 +98,48 @@ export MILVUS_ENABLED=0
# EC-RAG support chat history round setting, by default chat history is disabled, you can set CHAT_HISTORY_ROUND to control it
# export CHAT_HISTORY_ROUND= # change to your preference
+```
+
+### Deploy the Service on Intel GPU Using Docker Compose
-# EC-RAG support pipeline performance benchmark, use ENABLE_BENCHMARK=true/false to turn on/off benchmark
-# export ENABLE_BENCHMARK= # change to your preference
+#### Option a. Deploy openVINO LLM based EC-RAG for Core Ultra, Arc B60, Arc A770
-export MAX_MODEL_LEN=5000
-# Launch EC-RAG service with compose
+```bash
docker compose -f docker_compose/intel/gpu/arc/compose.yaml up -d
```
+#### Option b.1. Deploy vLLM based EC-RAG for Arc B60
+
+```bash
+docker compose --profile b60 -f docker_compose/intel/gpu/arc/compose.yaml up -d
+```
+
+#### Option b.2. Deploy vLLM based EC-RAG for Arc A770
+
+```bash
+docker compose --profile a770 -f docker_compose/intel/gpu/arc/compose.yaml up -d
+```
+
+### 6. Cleanup the Deployment (Manual)
+
+To stop the containers associated with the deployment, execute the following command:
+
+```bash
+docker compose -f docker_compose/intel/gpu/arc/compose.yaml down
+```
+
+All the EdgeCraftRAG containers will be stopped and then removed on completion of the `down` command.
+
## EC-RAG with Kbadmin
EC-RAG support kbadmin as a knowledge base manager
Please make sure all the kbadmin services have been launched
-EC-RAG Docker Images preparation is the same as local inference section, please refer to [Build Docker Images](#1-optional-build-docker-images-for-mega-service-server-and-ui-by-your-own)
-Model preparation is the same as vLLM inference section, please refer to [Prepare models](../docker_compose/intel/gpu/arc/README.md#2-prepare-models)
+EC-RAG Docker Images preparation is the same as local inference section, please refer to [Build Docker Images](#build-docker-images-for-mega-service-server-and-ui-by-your-own)
+Model preparation is the same as Arc manual deployment section, please refer to [Prepare models](#prepare-models)
### 1. Start Edge Craft RAG Services with Docker Compose
-This section is the same as default vLLM inference section, please refer to [Prepare env variables and configurations](../docker_compose/intel/gpu/arc/README.md#prepare-env-variables-and-configurations) and [Start Edge Craft RAG Services with Docker Compose](../docker_compose/intel/gpu/arc/README.md#deploy-the-service-on-arc-a770-using-docker-compose)
+This section is the same as Arc manual deployment section, please refer to [Prepare env variables and configurations](#prepare-env-variables-and-configurations) and [Deploy the Service on Intel GPU Using Docker Compose](#deploy-the-service-on-intel-gpu-using-docker-compose)
### 2. Access Kbadmin UI
@@ -92,38 +150,16 @@ please refer to [ChatQnA with Kbadmin in UI](./Explore_Edge_Craft_RAG.md#chatqna
In this sample, we will use Qwen3-30B-A3B deployment on 4 Arc B60 GPUs as an example.
Before started, please prepare models into MODEL_PATH and prepare docker images
-### Prepare env variables and configurations
-
```bash
export MODEL_PATH="${PWD}/models" # Your model path
export LLM_MODEL="Qwen/Qwen3-30B-A3B"
-ip_address=$(hostname -I | awk '{print $1}')
-# Use `ip a` to check your active ip
-export HOST_IP=$ip_address # Your host ip
-
-# Check group id of video and render
-export VIDEOGROUPID=$(getent group video | cut -d: -f3)
-export RENDERGROUPID=$(getent group render | cut -d: -f3)
+pip install modelscope
+modelscope download --model $LLM_MODEL --local_dir "${MODEL_PATH}/${LLM_MODEL}"
-# If you have a proxy configured, execute below line
-export no_proxy=${no_proxy},${HOST_IP},edgecraftrag,edgecraftrag-server
-export NO_PROXY=${NO_PROXY},${HOST_IP},edgecraftrag,edgecraftrag-server
-# If you have a HF mirror configured, it will be imported to the container
-# export HF_ENDPOINT=https://hf-mirror.com # your HF mirror endpoint"
-
-# Make sure all 3 folders have 1000:1000 permission, otherwise
-export DOC_PATH=${PWD}/tests
-export TMPFILE_PATH=${PWD}/tests
-chown 1000:1000 ${MODEL_PATH} ${DOC_PATH} ${TMPFILE_PATH}
-# In addition, also make sure the .cache folder has 1000:1000 permission, otherwise
-chown 1000:1000 -R $HOME/.cache
-```
-
-### Deploy the Service on Arc B60 Using Docker Compose
-
-```bash
# vLLM envs
export TP=4 # for multi GPU, you can change TP value
export ZE_AFFINITY_MASK=0,1,2,3 # for multi GPU, you can export ZE_AFFINITY_MASK=0,1,2...
docker compose --profile b60 -f docker_compose/intel/gpu/arc/compose.yaml up -d
```
+
+
diff --git a/EdgeCraftRAG/docs/Advanced_Setup_zh.md b/EdgeCraftRAG/docs/Advanced_Setup_zh.md
new file mode 100644
index 0000000000..b167633abb
--- /dev/null
+++ b/EdgeCraftRAG/docs/Advanced_Setup_zh.md
@@ -0,0 +1,163 @@
+# Edge Craft 检索增强生成(EC-RAG)高级部署说明
+
+[English](Advanced_Setup.md)
+
+## 自行构建 Mega Service、Server 和 UI 镜像
+
+**所有 Docker 镜像都可以自动拉取**。如果你希望自行构建镜像,请使用辅助脚本:
+
+```bash
+# 构建全部镜像
+./tools/build_images.sh
+
+# 单独构建某个镜像
+./tools/build_images.sh mega
+./tools/build_images.sh server
+./tools/build_images.sh ui
+
+# 组合构建多个镜像
+./tools/build_images.sh mega server
+```
+
+## Arc 平台手动部署详细说明
+
+### 准备模型
+
+需要准备 3 类模型:**Embedding**、**Reranking**、**LLM**。
+你需要根据场景选择对应推理框架。
+
+#### Embedding 与 Reranking
+
+Embedding 与 Reranking 通常由本地 OpenVINO 推理提供,可按如下方式准备:
+
+```bash
+# 准备 embedding、reranking 模型:
+export MODEL_PATH="${PWD}/models" # embedding、reranking、LLM 模型目录
+mkdir -p $MODEL_PATH
+pip install --upgrade --upgrade-strategy eager "optimum[openvino]"
+optimum-cli export openvino -m BAAI/bge-small-en-v1.5 ${MODEL_PATH}/BAAI/bge-small-en-v1.5 --task sentence-similarity
+optimum-cli export openvino -m BAAI/bge-reranker-large ${MODEL_PATH}/BAAI/bge-reranker-large --task text-classification
+```
+
+#### LLM
+
+##### openVINO
+如果仅使用 Core Ultra 平台,请准备 openVINO 模型:
+你也可以在独立 GPU 上运行 openVINO 模型。
+
+```bash
+# 准备 openVINO 的 LLM 模型
+optimum-cli export openvino --model Qwen/Qwen3-8B ${MODEL_PATH}/Qwen/Qwen3-8B/INT4_compressed_weights --task text-generation-with-past --weight-format int4 --group-size 128 --ratio 0.8
+```
+
+##### vLLM
+如果你有独立 GPU 并希望使用 vLLM,可按如下方式准备模型:
+
+```bash
+# 准备 vLLM 的 LLM 模型
+export LLM_MODEL="Qwen/Qwen3-8B" # 模型 ID
+pip install modelscope
+modelscope download --model $LLM_MODEL --local_dir "${MODEL_PATH}/${LLM_MODEL}"
+# 可选:也可以使用 huggingface 下载模型:
+# pip install -U huggingface_hub
+# huggingface-cli download $LLM_MODEL --local-dir "${MODEL_PATH}/${LLM_MODEL}"
+```
+
+### 准备环境变量与配置
+
+```bash
+ip_address=$(hostname -I | awk '{print $1}')
+# 可使用 `ip a` 查看当前活动 ip
+export HOST_IP=$ip_address # 主机 IP
+
+# 获取 video / render 组 ID
+export VIDEOGROUPID=$(getent group video | cut -d: -f3)
+export RENDERGROUPID=$(getent group render | cut -d: -f3)
+
+# 如果你配置了代理,请执行以下命令
+export no_proxy=${no_proxy},${HOST_IP},edgecraftrag,edgecraftrag-server
+export NO_PROXY=${NO_PROXY},${HOST_IP},edgecraftrag,edgecraftrag-server
+# 如果你配置了 HF 镜像,会被注入容器
+# export HF_ENDPOINT=https://hf-mirror.com # 你的 HF 镜像地址
+
+# 确保以下 3 个目录权限为 1000:1000
+export DOC_PATH=${PWD}/tests
+export TMPFILE_PATH=${PWD}/tests
+chown 1000:1000 ${MODEL_PATH} ${DOC_PATH} ${TMPFILE_PATH}
+# 同时确保 .cache 目录权限为 1000:1000
+chown 1000:1000 -R $HOME/.cache
+```
+
+设置 Milvus 和聊天历史轮数:
+
+```bash
+# EC-RAG 支持 Milvus 持久化数据库,默认关闭;设置 MILVUS_ENABLED=1 可开启
+export MILVUS_ENABLED=0
+# 启用 Milvus 时,默认存储路径为 PWD,如需修改请取消注释:
+# export DOCKER_VOLUME_DIRECTORY= # 按需修改
+
+# EC-RAG 支持聊天历史轮数,默认关闭;可通过 CHAT_HISTORY_ROUND 控制
+# export CHAT_HISTORY_ROUND= # 按需修改
+```
+
+### 使用 Docker Compose 在 Intel GPU 上部署服务
+
+#### 选项 a:为 Core Ultra / Arc B60 / Arc A770 部署基于 openVINO LLM 的 EC-RAG
+
+```bash
+docker compose -f docker_compose/intel/gpu/arc/compose.yaml up -d
+```
+
+#### 选项 b.1:为 Arc B60 部署基于 vLLM 的 EC-RAG
+
+```bash
+docker compose --profile b60 -f docker_compose/intel/gpu/arc/compose.yaml up -d
+```
+
+#### 选项 b.2:为 Arc A770 部署基于 vLLM 的 EC-RAG
+
+```bash
+docker compose --profile a770 -f docker_compose/intel/gpu/arc/compose.yaml up -d
+```
+
+### 6. 清理部署(手动)
+
+若要停止本次部署相关容器,请执行:
+
+```bash
+docker compose -f docker_compose/intel/gpu/arc/compose.yaml down
+```
+
+执行 `down` 后,EdgeCraftRAG 相关容器将停止并移除。
+
+## EC-RAG 与 Kbadmin
+
+EC-RAG 支持 kbadmin 作为知识库管理器。
+请先确保所有 kbadmin 服务已启动。
+EC-RAG 镜像准备与本地推理章节一致,请参考 [自行构建镜像](#自行构建-mega-serviceserver-和-ui-镜像)。
+模型准备与 Arc 手动部署章节一致,请参考 [准备模型](#准备模型)。
+
+### 1. 使用 Docker Compose 启动 Edge Craft RAG 服务
+
+此部分与 Arc 手动部署一致,请参考 [准备环境变量与配置](#准备环境变量与配置) 和 [使用 Docker Compose 在 Intel GPU 上部署服务](#使用-docker-compose-在-intel-gpu-上部署服务)。
+
+### 2. 访问 Kbadmin UI
+
+请参考 [ChatQnA with Kbadmin in UI](./Explore_Edge_Craft_RAG.md#chatqna-with-kbadmin-in-ui)。
+
+## 在多 GPU 上部署 EC-RAG
+
+本示例以 4 张 Arc B60 部署 Qwen3-30B-A3B 为例。
+开始前请先准备模型到 `MODEL_PATH` 并准备好 Docker 镜像。
+
+```bash
+export MODEL_PATH="${PWD}/models" # 模型路径
+export LLM_MODEL="Qwen/Qwen3-30B-A3B"
+pip install modelscope
+modelscope download --model $LLM_MODEL --local_dir "${MODEL_PATH}/${LLM_MODEL}"
+
+# vLLM 环境变量
+export TP=4 # 多卡时可按需调整 TP
+export ZE_AFFINITY_MASK=0,1,2,3 # 多卡时可按需调整
+docker compose --profile b60 -f docker_compose/intel/gpu/arc/compose.yaml up -d
+```
diff --git a/EdgeCraftRAG/docs/Agentic_RAG_Guide.md b/EdgeCraftRAG/docs/Agentic_RAG_Guide.md
index c499c8ede1..b353b7e990 100644
--- a/EdgeCraftRAG/docs/Agentic_RAG_Guide.md
+++ b/EdgeCraftRAG/docs/Agentic_RAG_Guide.md
@@ -1,5 +1,7 @@
# EC-RAG Agent Guide
+[中文版](Agentic_RAG_Guide_zh.md)
+
This guide explains how to enable, configure, and use Agent capabilities within EdgeCraftRAG (EC-RAG). It covers the creation and management of different agent types to enhance retrieval and generation workflows.(Currently, EC-RAG Agent is an experimental option.)
## Overview
diff --git a/EdgeCraftRAG/docs/Agentic_RAG_Guide_Zh.md b/EdgeCraftRAG/docs/Agentic_RAG_Guide_zh.md
similarity index 99%
rename from EdgeCraftRAG/docs/Agentic_RAG_Guide_Zh.md
rename to EdgeCraftRAG/docs/Agentic_RAG_Guide_zh.md
index 5862a3f448..3c9802288d 100644
--- a/EdgeCraftRAG/docs/Agentic_RAG_Guide_Zh.md
+++ b/EdgeCraftRAG/docs/Agentic_RAG_Guide_zh.md
@@ -1,5 +1,7 @@
# EC-RAG智能体指南
+[English](Agentic_RAG_Guide.md)
+
目前,EC-RAG智能体将作为试用版功能开放体验。
本指南将为您介绍如何在 EdgeCraftRAG (EC-RAG) 中启用、配置和使用智能体(Agent)功能,涵盖不同类型智能体的创建和管理。
diff --git a/EdgeCraftRAG/docs/Experience_Guide.md b/EdgeCraftRAG/docs/Experience_Guide.md
index 4be0dcfeb8..8d31242b1e 100644
--- a/EdgeCraftRAG/docs/Experience_Guide.md
+++ b/EdgeCraftRAG/docs/Experience_Guide.md
@@ -1,5 +1,7 @@
# EC-RAG Experience Guide
+[中文版](Experience_Guide_zh.md)
+
This guide explains how to enable, configure, and use Experience capabilities within EdgeCraftRAG (EC-RAG). It covers the creation and management of experience to enhance the answer quality
## Overview
diff --git a/EdgeCraftRAG/docs/Experience_Guide_Zh.md b/EdgeCraftRAG/docs/Experience_Guide_zh.md
similarity index 98%
rename from EdgeCraftRAG/docs/Experience_Guide_Zh.md
rename to EdgeCraftRAG/docs/Experience_Guide_zh.md
index 8c39ad6380..2a3602f61b 100644
--- a/EdgeCraftRAG/docs/Experience_Guide_Zh.md
+++ b/EdgeCraftRAG/docs/Experience_Guide_zh.md
@@ -1,5 +1,7 @@
# EC-RAG 经验注入指南
+[English](Experience_Guide.md)
+
本指南将为您介绍如何在 EdgeCraftRAG (EC-RAG) 中启用、配置和使用经验注入(Experience)功能,涵盖 Experience 的创建和管理,旨在提升您使用EC-RAG时的回答质量。
## 概述
diff --git a/EdgeCraftRAG/docs/Explore_Edge_Craft_RAG.md b/EdgeCraftRAG/docs/Explore_Edge_Craft_RAG.md
index 0a5a91ba5e..b8e89091ca 100644
--- a/EdgeCraftRAG/docs/Explore_Edge_Craft_RAG.md
+++ b/EdgeCraftRAG/docs/Explore_Edge_Craft_RAG.md
@@ -1,5 +1,7 @@
# Explore Edge Craft RAG
+[中文版](Explore_Edge_Craft_RAG_zh.md)
+
## ChatQnA with LLM Example in UI
### Create Pipeline
@@ -7,7 +9,7 @@
To create a default pipeline, you need to click the `Create Pipeline` button in the `Pipeline Setting` page.

-Then follow the pipeline create guide in UI to set your pipeline, please note that in `Indexer Type` you can set MilvusVector as indexer(Please make sure Milvus is enabled before set MilvusVector as indexer, you can refer to [Enable Milvus](../docker_compose/intel/gpu/arc/README.md#deploy-the-service-using-docker-compose)).
+Then follow the pipeline create guide in UI to set your pipeline, please note that in `Indexer Type` you can set MilvusVector as indexer(Please make sure Milvus is enabled before set MilvusVector as indexer, you can refer to [Enable Milvus](Advanced_Setup.md#deploy-the-service-on-intel-gpu-using-docker-compose)).
if choosing MilvusVector, you need to verify vector uri first, please input 'Your_IP:milvus_port' then click `Test` button. Note that milvus_port is 19530

diff --git a/EdgeCraftRAG/docs/Explore_Edge_Craft_RAG_zh.md b/EdgeCraftRAG/docs/Explore_Edge_Craft_RAG_zh.md
new file mode 100644
index 0000000000..8ad8c6d1d7
--- /dev/null
+++ b/EdgeCraftRAG/docs/Explore_Edge_Craft_RAG_zh.md
@@ -0,0 +1,57 @@
+# 探索 Edge Craft RAG
+
+[English](Explore_Edge_Craft_RAG.md)
+
+## 在 UI 中使用 LLM 进行 ChatQnA 示例
+
+### 创建流水线
+
+要创建默认流水线,请在 `Pipeline Setting`(流水线设置)页面点击 `Create Pipeline` 按钮。
+
+
+然后按照 UI 中的流水线创建向导进行配置。请注意,在 `Indexer Type`(索引器类型)中可以选择 MilvusVector 作为索引器(请确保在选择 MilvusVector 之前已启用 Milvus,可参考 [启用 Milvus](Advanced_Setup_zh.md#使用-docker-compose-在-intel-gpu-上部署服务))。
+如果选择 MilvusVector,需要先验证向量数据库 URI,请输入 `Your_IP:milvus_port`,然后点击 `Test` 按钮。注意 milvus_port 默认为 19530。
+
+
+然后,在 `Generator`(生成器)配置页面,将 LLM Inference Type 选择为 `Vllm`。
+在 `Large Language Model` 字段中输入您的 LLM 模型名称,例如 `Qwen/Qwen3-8B`。
+在 `Vllm Url` 字段中输入 `IP:vllm_port`,然后点击 `Test` 按钮。注意 vllm_port 默认为 `8086`。
+(注意:如果测试失败,可能是因为 vLLM 服务尚未就绪,可等待 30 秒后重试。)
+
+
+您也可以通过 `Operation` 字段创建多条流水线或更新/删除已有流水线,但请注意处于激活状态的流水线无法被更新。
+
+
+### 上传文件与 ChatQnA
+
+流水线创建完成后,前往 `Knowledge Base`(知识库)页面,点击 `Create Knowledge Base` 按钮创建知识库。
+
+
+创建知识库时,请选择 `Activated`(激活)选项,只有处于激活状态的知识库中的文件才能在 ChatQnA 中被检索。
+
+
+知识库创建完成后,即可上传用于检索的文档。
+
+
+然后,在 `Chat` 页面的聊天框中提交您的问题。
+
+
+## 在 UI 中使用 Kbadmin 进行 ChatQnA
+
+### Kbadmin 流水线
+
+在 `Node Parser` 页面,选择 `kbadmin` 作为节点解析器。
+
+
+在 `Indexer` 页面,填写 Embedding 服务和向量数据库信息,注意 Embedding 服务端口为 13020,向量数据库端口为 29530。
+
+
+### 上传文件与 ChatQnA
+
+流水线创建完成后,前往 `Knowledge Base` 页面,点击 `Create Knowledge Base` 按钮创建知识库。
+请在 `Type` 中选择 `kbadmin`,并从 kbadmin UI 页面中已创建的知识库列表中选择 kb 名称。
+
+
+
+然后,在 `Chat` 页面的聊天框中提交您的问题。
+
diff --git a/EdgeCraftRAG/edgecraftrag/api/v1/agent.py b/EdgeCraftRAG/edgecraftrag/api/v1/agent.py
index bcae2e4e95..a7bf33ed4a 100644
--- a/EdgeCraftRAG/edgecraftrag/api/v1/agent.py
+++ b/EdgeCraftRAG/edgecraftrag/api/v1/agent.py
@@ -16,7 +16,7 @@
# GET Agents
-@agent_app.get(path="/v1/settings/agents")
+@agent_app.get(path="/v1/agents")
async def get_all_agents():
out = []
agents = ctx.get_agent_mgr().get_agents()
@@ -38,7 +38,7 @@ async def get_all_agents():
# GET Agent
-@agent_app.get(path="/v1/settings/agents/{name}")
+@agent_app.get(path="/v1/agents/{name}")
async def get_agent(name):
agent = ctx.get_agent_mgr().get_agent_by_name(name)
if agent:
@@ -58,7 +58,7 @@ async def get_agent(name):
# POST Agent
-@agent_app.post(path="/v1/settings/agents")
+@agent_app.post(path="/v1/agents")
async def create_agent(request: AgentCreateIn, status_code=status.HTTP_201_CREATED):
try:
agent = ctx.get_agent_mgr().create_agent(request)
@@ -70,7 +70,7 @@ async def create_agent(request: AgentCreateIn, status_code=status.HTTP_201_CREAT
# PATCH Agent
-@agent_app.patch(path="/v1/settings/agents/{name}")
+@agent_app.patch(path="/v1/agents/{name}")
async def update_agent(name, request: AgentCreateIn):
try:
agentmgr = ctx.get_agent_mgr()
@@ -91,7 +91,7 @@ async def update_agent(name, request: AgentCreateIn):
# DELETE Agent
-@agent_app.delete(path="/v1/settings/agents/{name}", status_code=status.HTTP_204_NO_CONTENT)
+@agent_app.delete(path="/v1/agents/{name}", status_code=status.HTTP_204_NO_CONTENT)
async def delete_agent(name):
try:
agentmgr = ctx.get_agent_mgr()
@@ -108,7 +108,7 @@ async def delete_agent(name):
# GET Agent Type default configs
-@agent_app.get(path="/v1/settings/agents/configs/{agent_type}")
+@agent_app.get(path="/v1/agents/configs/{agent_type}")
async def get_agent_default_configs(agent_type):
try:
if agent_type in [e.value for e in AgentType]:
diff --git a/EdgeCraftRAG/edgecraftrag/api/v1/knowledge_base.py b/EdgeCraftRAG/edgecraftrag/api/v1/knowledge_base.py
index 7f044dd276..d3074af270 100644
--- a/EdgeCraftRAG/edgecraftrag/api/v1/knowledge_base.py
+++ b/EdgeCraftRAG/edgecraftrag/api/v1/knowledge_base.py
@@ -75,7 +75,7 @@ async def get_knowledge_base(knowledge_name: str):
# Get the specified knowledge base json.
@kb_app.get("/v1/knowledge/{knowledge_name}/json")
-async def get_knowledge_base(knowledge_name: str):
+async def get_knowledge_base_json(knowledge_name: str):
kb = ctx.knowledgemgr.get_knowledge_base_by_name_or_id(knowledge_name)
return kb.get_knowledge_json
diff --git a/EdgeCraftRAG/edgecraftrag/config_repository.py b/EdgeCraftRAG/edgecraftrag/config_repository.py
index f3d0a62c66..c2f42f4900 100644
--- a/EdgeCraftRAG/edgecraftrag/config_repository.py
+++ b/EdgeCraftRAG/edgecraftrag/config_repository.py
@@ -347,7 +347,6 @@ async def save_knowledge_configurations(operation: str = None, kb=None):
target_kb["experience_active"] = kb.experience_active
target_kb["all_document_maps"] = kb.all_document_maps
target_kb["file_paths"] = kb.file_paths
- "experience_active"
target_idx = target_kb.get("idx")
if not target_idx:
return {"message": "Missing 'idx' in knowledgebase data"}
diff --git a/EdgeCraftRAG/tests/configs/agent.json b/EdgeCraftRAG/tests/configs/agent.json
new file mode 100644
index 0000000000..78dbb9c383
--- /dev/null
+++ b/EdgeCraftRAG/tests/configs/agent.json
@@ -0,0 +1,12 @@
+[
+ {
+ "idx": "bb92fe12-2635-4667-abdd-5bf6ef9af8bd",
+ "name": "aaa",
+ "type": "simple",
+ "pipeline_idx": "0e591c6b-1e82-48d7-9458-98477f172a7e",
+ "configs": {
+ "max_retrievals": 7
+ },
+ "active": true
+ }
+]
\ No newline at end of file
diff --git a/EdgeCraftRAG/tests/configs/pipeline.json b/EdgeCraftRAG/tests/configs/pipeline.json
new file mode 100644
index 0000000000..777d1b79dc
--- /dev/null
+++ b/EdgeCraftRAG/tests/configs/pipeline.json
@@ -0,0 +1,42 @@
+[
+ {
+ "idx": "0e591c6b-1e82-48d7-9458-98477f172a7e",
+ "name": "default",
+ "retriever": {
+ "retriever_type": "vectorsimilarity",
+ "retrieve_topk": 30
+ },
+ "postprocessor": [
+ {
+ "processor_type": "reranker",
+ "reranker_model": {
+ "model_type": "LLM",
+ "model_id": "BAAI/bge-reranker-large",
+ "model_path": "./models/BAAI/bge-reranker-large",
+ "weight": "INT4",
+ "device": "AUTO",
+ "api_base": null
+ },
+ "top_n": 25
+ }
+ ],
+ "generator": [
+ {
+ "generator_type": "chatqna",
+ "prompt_path": "./default_prompt.txt",
+ "prompt_content": null,
+ "model": {
+ "model_type": "LLM",
+ "model_id": "Qwen/Qwen2-7B-Instruct",
+ "model_path": "./models/Qwen/Qwen2-7B-Instruct/INT4_compressed_weights/",
+ "weight": "INT4",
+ "device": "AUTO",
+ "api_base": null
+ },
+ "inference_type": "local",
+ "vllm_endpoint": null
+ }
+ ],
+ "active": true
+ }
+]
\ No newline at end of file
diff --git a/EdgeCraftRAG/tools/build_images.sh b/EdgeCraftRAG/tools/build_images.sh
new file mode 100755
index 0000000000..6a2bd0ed2f
--- /dev/null
+++ b/EdgeCraftRAG/tools/build_images.sh
@@ -0,0 +1,82 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+PROJECT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+
+cd "${PROJECT_DIR}"
+
+build_mega() {
+ echo "Building opea/edgecraftrag:latest"
+ docker build --no-cache --pull \
+ --build-arg http_proxy="${http_proxy-}" \
+ --build-arg https_proxy="${https_proxy-}" \
+ --build-arg no_proxy="${no_proxy-}" \
+ -t opea/edgecraftrag:latest \
+ -f Dockerfile .
+}
+
+build_server() {
+ echo "Building opea/edgecraftrag-server:latest"
+ docker build --no-cache --pull \
+ --build-arg http_proxy="${http_proxy-}" \
+ --build-arg https_proxy="${https_proxy-}" \
+ --build-arg no_proxy="${no_proxy-}" \
+ -t opea/edgecraftrag-server:latest \
+ -f Dockerfile.server .
+}
+
+build_ui() {
+ echo "Building opea/edgecraftrag-ui:latest"
+ docker build --no-cache --pull \
+ --build-arg http_proxy="${http_proxy-}" \
+ --build-arg https_proxy="${https_proxy-}" \
+ --build-arg no_proxy="${no_proxy-}" \
+ -t opea/edgecraftrag-ui:latest \
+ -f ui/docker/Dockerfile.ui .
+}
+
+print_usage() {
+ echo "Usage: ./tools/build_images.sh [mega|server|ui|all] ..."
+ echo "Examples:"
+ echo " ./tools/build_images.sh"
+ echo " ./tools/build_images.sh ui"
+ echo " ./tools/build_images.sh mega server"
+}
+
+if [ "$#" -eq 0 ]; then
+ build_mega
+ build_server
+ build_ui
+ echo "All images built successfully."
+ exit 0
+fi
+
+for target in "$@"; do
+ case "$target" in
+ mega)
+ build_mega
+ ;;
+ server)
+ build_server
+ ;;
+ ui)
+ build_ui
+ ;;
+ all)
+ build_mega
+ build_server
+ build_ui
+ ;;
+ -h|--help|help)
+ print_usage
+ exit 0
+ ;;
+ *)
+ echo "Unknown target: $target"
+ print_usage
+ exit 1
+ ;;
+ esac
+done
+
+echo "Requested image build(s) completed successfully."
diff --git a/EdgeCraftRAG/tools/quick_start.sh b/EdgeCraftRAG/tools/quick_start.sh
index 7634891ecd..fe4580bf66 100755
--- a/EdgeCraftRAG/tools/quick_start.sh
+++ b/EdgeCraftRAG/tools/quick_start.sh
@@ -4,27 +4,53 @@
set -e
-WORKPATH=$(dirname "$(pwd)")
+SCRIPT_DIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+WORKPATH=$(cd "${SCRIPT_DIR}/.." && pwd)
ip_address=$(hostname -I | awk '{print $1}')
HOST_IP=$ip_address
+# global defaults to avoid docker compose warnings on unset variables
+export HOST_IP=${HOST_IP:-"${ip_address}"}
+export MODEL_PATH=${MODEL_PATH:-"${WORKPATH}/models"}
+export LLM_MODEL=${LLM_MODEL:-"Qwen/Qwen3-8B"}
+export DOC_PATH=${DOC_PATH:-"${WORKPATH}/tests"}
+export TMPFILE_PATH=${TMPFILE_PATH:-"${WORKPATH}/tests"}
+export MILVUS_ENABLED=${MILVUS_ENABLED:-"0"}
+export CHAT_HISTORY_ROUND=${CHAT_HISTORY_ROUND:-"0"}
+
#use python venv
-ENV_NAME="ecrag_venv"
-python -m venv $ENV_NAME
+ENV_NAME="${WORKPATH}/ecrag_venv"
+
+# check if python3-venv (ensurepip) is fully available; install if missing
+if ! python3 -c "import ensurepip" &>/dev/null; then
+ echo "python3-venv (ensurepip) not found, installing..."
+ PY_VER=$(python3 -c "import sys; print(f'{sys.version_info.major}.{sys.version_info.minor}')")
+ if command -v apt-get &>/dev/null; then
+ sudo apt-get install -y "python${PY_VER}-venv"
+ elif command -v dnf &>/dev/null; then
+ sudo dnf install -y python3-virtualenv
+ elif command -v yum &>/dev/null; then
+ sudo yum install -y python3-virtualenv
+ else
+ echo "ERROR: Cannot install python3-venv: unsupported package manager. Please install it manually."
+ exit 1
+ fi
+fi
-# check venv
-if [ ! -d "$ENV_NAME" ]; then
- echo "Failed to create virtual environment"
- exit 1
+# create venv if missing or broken (activate script absent)
+if [ ! -f "${ENV_NAME}/bin/activate" ] && [ ! -f "${ENV_NAME}/Scripts/activate" ]; then
+ echo "Creating virtual environment at ${ENV_NAME}..."
+ rm -rf "${ENV_NAME}"
+ python3 -m venv "${ENV_NAME}"
fi
# activate venv
-if [ -f "$ENV_NAME/bin/activate" ]; then
- source $ENV_NAME/bin/activate
-elif [ -f "$ENV_NAME/Scripts/activate" ]; then
- source $ENV_NAME/Scripts/activate
+if [ -f "${ENV_NAME}/bin/activate" ]; then
+ source "${ENV_NAME}/bin/activate"
+elif [ -f "${ENV_NAME}/Scripts/activate" ]; then
+ source "${ENV_NAME}/Scripts/activate"
else
- echo "Failed to activate virtual environment"
+ echo "ERROR: Failed to activate virtual environment at ${ENV_NAME}"
exit 1
fi
@@ -42,6 +68,13 @@ get_enable_function() {
echo ${user_input:-$default_value}
}
+print_ui_access_info() {
+ echo ""
+ echo "Service launched successfully."
+ echo "UI access URL: http://${HOST_IP}:8082"
+ echo "If you are accessing from another machine, replace ${HOST_IP} with the server's reachable IP or hostname."
+}
+
function start_vllm_services() {
COMPOSE_FILE="compose.yaml"
echo "stop former service..."
@@ -54,7 +87,7 @@ function start_vllm_services() {
MILVUS_ENABLED=$(get_enable_function "MILVUS DB(Enter 1 for enable)" "0")
CHAT_HISTORY_ROUND=$(get_user_input "chat history round" "0")
LLM_MODEL=$(get_user_input "your LLM model" "Qwen/Qwen3-8B")
- MODEL_PATH=$(get_user_input "your model path" "${PWD}/models")
+ MODEL_PATH=$(get_user_input "your model path" "${WORKPATH}/models")
read -p "Have you prepare models in ${MODEL_PATH}:(yes/no) [yes]" user_input
user_input=${user_input:-"yes"}
@@ -64,6 +97,7 @@ function start_vllm_services() {
# Reranker: ${MODEL_PATH}/BAAI/bge-reranker-large
# llm :${MODEL_PATH}/${LLM_MODEL} (从huggingface或modelscope下载的原始模型,而不是经过OpenVINO转换的模型!)
echo "you skipped model downloading, please make sure you have prepared all models under ${MODEL_PATH}"
+ ensure_required_models_for_vllm
else
echo "you have not prepare models, starting to download models into ${MODEL_PATH}..."
mkdir -p $MODEL_PATH
@@ -116,7 +150,7 @@ function start_vllm_services() {
n=$((n+1))
done
rm -rf ipex-llm-serving-xpu-container.log
- echo "service launched, please visit UI at ${HOST_IP}:8082"
+ print_ui_access_info
}
@@ -132,7 +166,7 @@ function start_services() {
MILVUS_ENABLED=$(get_enable_function "MILVUS DB(Enter 1 for enable)" "0")
CHAT_HISTORY_ROUND=$(get_user_input "chat history round" "0")
LLM_MODEL=$(get_user_input "your LLM model" "Qwen/Qwen3-8B")
- MODEL_PATH=$(get_user_input "your model path" "${PWD}/models")
+ MODEL_PATH=$(get_user_input "your model path" "${WORKPATH}/models")
read -p "Have you prepare models in ${MODEL_PATH}:(yes/no) [yes]" user_input
user_input=${user_input:-"yes"}
@@ -142,6 +176,7 @@ function start_services() {
# Reranker: ${MODEL_PATH}/BAAI/bge-reranker-large
# llm :${MODEL_PATH}/${LLM_MODEL}/INT4_compressed_weights
echo "you skipped model downloading, please make sure you have prepared all models under ${MODEL_PATH}"
+ ensure_required_models_for_ov
else
read -p "you have not prepare models, do you need one-click model downloading into ${MODEL_PATH}:(yes/no) [yes]" your_input
your_input=${your_input:-"yes"}
@@ -185,6 +220,7 @@ function start_services() {
COMPOSE_FILE="compose.yaml"
echo "starting service..."
docker compose -f $WORKPATH/docker_compose/intel/gpu/arc/$COMPOSE_FILE up -d
+ print_ui_access_info
}
@@ -200,16 +236,90 @@ function check_baai_folder() {
fi
}
+ensure_openvino_tooling() {
+ if ! command -v optimum-cli >/dev/null 2>&1; then
+ echo "[Model Check] 'optimum-cli' not found, installing optimum-intel[openvino]..."
+ python -m pip install --upgrade-strategy eager "optimum-intel[openvino]"
+ fi
+}
+
+ensure_huggingface_tooling() {
+ if ! command -v huggingface-cli >/dev/null 2>&1; then
+ echo "[Model Check] 'huggingface-cli' not found, installing huggingface_hub..."
+ python -m pip install huggingface_hub
+ fi
+}
+
+ensure_embedding_and_reranker_models() {
+ local embedding_dir="${MODEL_PATH}/BAAI/bge-small-en-v1.5"
+ local reranker_dir="${MODEL_PATH}/BAAI/bge-reranker-large"
+
+ if [ ! -f "${embedding_dir}/openvino_model.xml" ]; then
+ echo "[Model Check] Embedding model missing: ${embedding_dir}"
+ echo "[Model Check] Downloading embedding model..."
+ ensure_openvino_tooling
+ mkdir -p "${embedding_dir}"
+ optimum-cli export openvino -m BAAI/bge-small-en-v1.5 "${embedding_dir}" --task sentence-similarity
+ else
+ echo "[Model Check] Embedding model exists: ${embedding_dir}"
+ fi
+
+ if [ ! -f "${reranker_dir}/openvino_model.xml" ]; then
+ echo "[Model Check] Reranker model missing: ${reranker_dir}"
+ echo "[Model Check] Downloading reranker model..."
+ ensure_openvino_tooling
+ mkdir -p "${reranker_dir}"
+ optimum-cli export openvino -m BAAI/bge-reranker-large "${reranker_dir}" --task text-classification
+ else
+ echo "[Model Check] Reranker model exists: ${reranker_dir}"
+ fi
+}
+
+ensure_llm_model_for_vllm() {
+ local llm_dir="${MODEL_PATH}/${LLM_MODEL}"
+ if [ ! -f "${llm_dir}/config.json" ]; then
+ echo "[Model Check] vLLM LLM model missing: ${llm_dir}"
+ echo "[Model Check] Downloading LLM model '${LLM_MODEL}'..."
+ ensure_huggingface_tooling
+ mkdir -p "${llm_dir}"
+ huggingface-cli download "${LLM_MODEL}" --local-dir "${llm_dir}"
+ else
+ echo "[Model Check] vLLM LLM model exists: ${llm_dir}"
+ fi
+}
+
+ensure_llm_model_for_ov() {
+ local ov_llm_dir="${MODEL_PATH}/${LLM_MODEL}/INT4_compressed_weights"
+ if [ ! -f "${ov_llm_dir}/openvino_model.xml" ]; then
+ echo "[Model Check] OpenVINO LLM model missing: ${ov_llm_dir}"
+ echo "[Model Check] Downloading and converting LLM model '${LLM_MODEL}' to INT4 OpenVINO..."
+ ensure_openvino_tooling
+ mkdir -p "${ov_llm_dir}"
+ optimum-cli export openvino --model "${LLM_MODEL}" "${ov_llm_dir}" --task text-generation-with-past --weight-format int4 --group-size 128 --ratio 0.8
+ else
+ echo "[Model Check] OpenVINO LLM model exists: ${ov_llm_dir}"
+ fi
+}
+
+ensure_required_models_for_vllm() {
+ ensure_embedding_and_reranker_models
+ ensure_llm_model_for_vllm
+}
+
+ensure_required_models_for_ov() {
+ ensure_embedding_and_reranker_models
+ ensure_llm_model_for_ov
+}
+
function quick_start_vllm_services() {
- WORKPATH=$(dirname "$PWD")
COMPOSE_FILE="compose.yaml"
EC_RAG_SERVICE_PORT=16010
docker compose -f $WORKPATH/docker_compose/intel/gpu/arc/$COMPOSE_FILE down
ip_address=$(hostname -I | awk '{print $1}')
export HOST_IP=${HOST_IP:-"${ip_address}"}
- export MODEL_PATH=${MODEL_PATH:-"${PWD}/models"}
+ export MODEL_PATH=${MODEL_PATH:-"${WORKPATH}/models"}
export DOC_PATH=${DOC_PATH:-"$WORKPATH/tests"}
export TMPFILE_PATH=${TMPFILE_PATH:-"$WORKPATH/tests"}
export DP_NUM=${DP_NUM:-1}
@@ -223,12 +333,12 @@ function quick_start_vllm_services() {
export QUANTIZATION=${QUANTIZATION:-fp8}
export CCL_DG2_USM=${CCL_DG2_USM:-0}
export LLM_MODEL=${LLM_MODEL:-Qwen/Qwen3-8B}
- export LLM_MODEL_PATH=${LLM_MODEL_PATH:-"${MODEL_PATH}/Qwen/Qwen3-8B"}
+ export LLM_MODEL_PATH=${LLM_MODEL_PATH:-"${MODEL_PATH}/${LLM_MODEL}"}
export VIDEOGROUPID=$(getent group video | cut -d: -f3)
export RENDERGROUPID=$(getent group render | cut -d: -f3)
export VLLM_SERVICE_PORT_A770=8086
- check_baai_folder
+ ensure_required_models_for_vllm
export HF_CACHE=${HF_CACHE:-"${HOME}/.cache"}
export no_proxy="localhost, 127.0.0.1, 192.168.1.1, ${HOST_IP}"
if [ ! -d "${HF_CACHE}" ]; then
@@ -251,7 +361,7 @@ function quick_start_vllm_services() {
n=$((n+1))
done
rm -rf ipex-llm-serving-xpu-container.log
- echo "service launched, please visit UI at ${HOST_IP}:8082"
+ print_ui_access_info
}
@@ -267,12 +377,12 @@ function quick_start_ov_services() {
export MILVUS_ENABLED=${MILVUS_ENABLED:-1}
export CHAT_HISTORY_ROUND=${CHAT_HISTORY_ROUND:-"0"}
export LLM_MODEL=${LLM_MODEL:-"Qwen/Qwen3-8B"}
- export MODEL_PATH=${MODEL_PATH:-"${PWD}/models"}
+ export MODEL_PATH=${MODEL_PATH:-"${WORKPATH}/models"}
export VIDEOGROUPID=$(getent group video | cut -d: -f3)
export RENDERGROUPID=$(getent group render | cut -d: -f3)
export MAX_MODEL_LEN=5000
- check_baai_folder
+ ensure_required_models_for_ov
export HF_CACHE=${HF_CACHE:-"${HOME}/.cache"}
if [ ! -d "${HF_CACHE}" ]; then
mkdir -p "${HF_CACHE}"
@@ -287,13 +397,14 @@ function quick_start_ov_services() {
echo "Starting service..."
docker compose -f "$WORKPATH/docker_compose/intel/gpu/arc/$COMPOSE_FILE" up -d
+ print_ui_access_info
}
function start_vLLM_B60_services() {
COMPOSE_FILE="compose.yaml"
echo "stop former service..."
- export MODEL_PATH=${MODEL_PATH:-"${PWD}/models"}
+ export MODEL_PATH=${MODEL_PATH:-"${WORKPATH}/models"}
docker compose -f $WORKPATH/docker_compose/intel/gpu/arc/$COMPOSE_FILE down
ip_address=$(hostname -I | awk '{print $1}')
@@ -303,7 +414,7 @@ function start_vLLM_B60_services() {
MILVUS_ENABLED=$(get_enable_function "MILVUS DB(Enter 1 for enable)" "0")
CHAT_HISTORY_ROUND=$(get_user_input "chat history round" "0")
LLM_MODEL=$(get_user_input "your LLM model" "Qwen/Qwen3-8B")
- MODEL_PATH=$(get_user_input "your model path" "${PWD}/models")
+ MODEL_PATH=$(get_user_input "your model path" "${WORKPATH}/models")
read -p "Have you prepare models in ${MODEL_PATH}:(yes/no) [yes]" user_input
user_input=${user_input:-"yes"}
@@ -313,6 +424,7 @@ function start_vLLM_B60_services() {
# Reranker: ${MODEL_PATH}/BAAI/bge-reranker-large
# llm :${MODEL_PATH}/${LLM_MODEL} (从huggingface或modelscope下载的原始模型,而不是经过OpenVINO转换的模型!)
echo "you skipped model downloading, please make sure you have prepared all models under ${MODEL_PATH}"
+ ensure_required_models_for_vllm
else
echo "you have not prepare models, starting to download models into ${MODEL_PATH}..."
mkdir -p $MODEL_PATH
@@ -382,19 +494,18 @@ function start_vLLM_B60_services() {
n=$((n+1))
done
rm -rf ipex-llm-serving-xpu-container.log
- echo "service launched, please visit UI at ${HOST_IP}:8082"
+ print_ui_access_info
}
function quick_start_vllm_B60_services() {
- WORKPATH=$(dirname "$PWD")
COMPOSE_FILE="compose.yaml"
EC_RAG_SERVICE_PORT=16010
docker compose -f $WORKPATH/docker_compose/intel/gpu/arc/$COMPOSE_FILE down
ip_address=$(hostname -I | awk '{print $1}')
export HOST_IP=${HOST_IP:-"${ip_address}"}
- export MODEL_PATH=${MODEL_PATH:-"${PWD}/models"}
+ export MODEL_PATH=${MODEL_PATH:-"${WORKPATH}/models"}
export DOC_PATH=${DOC_PATH:-"$WORKPATH/tests"}
export TMPFILE_PATH=${TMPFILE_PATH:-"$WORKPATH/tests"}
export MILVUS_ENABLED=${MILVUS_ENABLED:-1}
@@ -419,7 +530,7 @@ function quick_start_vllm_B60_services() {
export QUANTIZATION=${QUANTIZATION:-fp8}
- check_baai_folder
+ ensure_required_models_for_vllm
export no_proxy="localhost, 127.0.0.1, 192.168.1.1, ${HOST_IP}"
sudo chown -R 1000:1000 ${MODEL_PATH} ${DOC_PATH} ${TMPFILE_PATH}
docker compose --profile b60 -f $WORKPATH/docker_compose/intel/gpu/arc/$COMPOSE_FILE up -d
@@ -434,14 +545,27 @@ function quick_start_vllm_B60_services() {
n=$((n+1))
done
rm -rf ipex-llm-serving-xpu-container.log
- echo "service launched, please visit UI at ${HOST_IP}:8082"
+ print_ui_access_info
+}
+
+
+function quick_cleanup_services() {
+ COMPOSE_FILE="compose.yaml"
+ echo "Stopping EdgeCraftRAG services..."
+ docker compose -f "$WORKPATH/docker_compose/intel/gpu/arc/$COMPOSE_FILE" down
+ echo "Cleanup completed."
}
function main {
+ if [[ "${1:-}" == "cleanup" ]]; then
+ quick_cleanup_services
+ exit 0
+ fi
+
if [[ $- == *i* ]]; then
- read -p "Do you want to start vLLM or local OpenVINO services? (vLLM_A770/vLLM_B60/ov) [vLLM_A770]: " user_input
- user_input=${user_input:-"vLLM_A770"}
+ read -p "Do you want to start vLLM or local OpenVINO services? (vLLM_A770/vLLM_B60/ov) [ov]: " user_input
+ user_input=${user_input:-"ov"}
if [[ "$user_input" == "vLLM_A770" ]]; then
start_vllm_services
elif [[ "$user_input" == "vLLM_B60" ]]; then
diff --git a/EdgeCraftRAG/ui/docker/Dockerfile.ui b/EdgeCraftRAG/ui/docker/Dockerfile.ui
index 80cda697b3..1296f51c98 100644
--- a/EdgeCraftRAG/ui/docker/Dockerfile.ui
+++ b/EdgeCraftRAG/ui/docker/Dockerfile.ui
@@ -1,25 +1,23 @@
-FROM node:20.18.1 AS vue-app
-
+FROM node:22.14.0 AS vue-app
+
RUN apt-get update -y && apt-get install -y git
-
-RUN useradd -m -s /bin/bash user && \
- mkdir -p /home/user && \
- chown -R user /home/user/
-
-COPY ./ui/vue /home/user/ui
-
-RUN mkdir -p /home/user/ui
-ENV TMPFILE_PATH=/home/user/ui
-
+
WORKDIR /home/user/ui
-
+
+ENV TMPFILE_PATH=/home/user/ui
+ENV NODE_OPTIONS=--max-old-space-size=4096
+ENV NPM_CONFIG_AUDIT=false
+ENV NPM_CONFIG_FUND=false
+
+COPY ./ui/vue/package*.json ./
RUN npm install
-
+
+COPY ./ui/vue ./
RUN npm run build
FROM nginx:stable-alpine
-COPY --from=vue-app home/user/ui/dist /usr/share/nginx/html
+COPY --from=vue-app /home/user/ui/dist /usr/share/nginx/html
COPY ./ui/vue/nginx.conf /etc/nginx/nginx.conf
diff --git a/EdgeCraftRAG/ui/vue/src/api/agent/index.ts b/EdgeCraftRAG/ui/vue/src/api/agent/index.ts
index ee6fc29b3e..9bf55f33ac 100644
--- a/EdgeCraftRAG/ui/vue/src/api/agent/index.ts
+++ b/EdgeCraftRAG/ui/vue/src/api/agent/index.ts
@@ -5,14 +5,14 @@ import request from "../request";
export const getAgentList = () => {
return request({
- url: "/v1/settings/agents",
+ url: "/v1/agents",
method: "get",
});
};
export const getAgentDetailByName = (name: String) => {
return request({
- url: `/v1/settings/agents/${name}`,
+ url: `/v1/agents/${name}`,
method: "get",
});
};
@@ -27,7 +27,7 @@ export const getFreechatList = (params: Object) => {
export const requestAgentCreate = (data: Object) => {
return request({
- url: "/v1/settings/agents",
+ url: "/v1/agents",
method: "post",
data,
showLoading: true,
@@ -37,7 +37,7 @@ export const requestAgentCreate = (data: Object) => {
};
export const requestAgentUpdate = (name: String, data: Object) => {
return request({
- url: `/v1/settings/agents/${name}`,
+ url: `/v1/agents/${name}`,
method: "patch",
data,
showLoading: true,
@@ -48,7 +48,7 @@ export const requestAgentUpdate = (name: String, data: Object) => {
export const requestAgentDelete = (name: String) => {
return request({
- url: `/v1/settings/agents/${name}`,
+ url: `/v1/agents/${name}`,
method: "delete",
showLoading: true,
showSuccessMsg: true,
@@ -58,7 +58,7 @@ export const requestAgentDelete = (name: String) => {
export const getAgentConfigs = (type: String) => {
return request({
- url: `/v1/settings/agents/configs/${type}`,
+ url: `/v1/agents/configs/${type}`,
method: "get",
});
};
From 75b2ad3a2291acac4e5d5870ae5492212fa176cf Mon Sep 17 00:00:00 2001
From: Yongbozzz
Date: Thu, 12 Mar 2026 11:05:19 +0800
Subject: [PATCH 3/6] remove helm for upstream
Signed-off-by: Yongbozzz
---
.../assets/img/Explore_Edge_Craft_RAG_01.jpg | Bin 0 -> 192368 bytes
.../assets/img/Explore_Edge_Craft_RAG_02.jpg | Bin 0 -> 184820 bytes
.../assets/img/Explore_Edge_Craft_RAG_03.jpg | Bin 0 -> 152093 bytes
.../assets/img/Explore_Edge_Craft_RAG_04.jpg | Bin 0 -> 115565 bytes
.../assets/img/Explore_Edge_Craft_RAG_05.jpg | Bin 0 -> 154383 bytes
.../assets/img/Explore_Edge_Craft_RAG_06.jpg | Bin 0 -> 72962 bytes
.../assets/img/Explore_Edge_Craft_RAG_07.jpg | Bin 0 -> 102716 bytes
.../assets/img/Explore_Edge_Craft_RAG_08.jpg | Bin 0 -> 101805 bytes
.../assets/img/Explore_Edge_Craft_RAG_09.png | Bin 0 -> 122839 bytes
.../assets/img/Explore_Edge_Craft_RAG_10.png | Bin 0 -> 115197 bytes
.../assets/img/Explore_Edge_Craft_RAG_11.png | Bin 0 -> 137856 bytes
.../docker_compose/intel/gpu/arc/README.md | 2 +
.../docker_compose/intel/gpu/arc/README_zh.md | 2 +
EdgeCraftRAG/docs/Explore_Edge_Craft_RAG.md | 53 +++--
.../docs/Explore_Edge_Craft_RAG_zh.md | 46 ++---
EdgeCraftRAG/kubernetes/helm/Chart.yaml | 6 -
EdgeCraftRAG/kubernetes/helm/README.md | 94 ---------
EdgeCraftRAG/kubernetes/helm/README_zh.md | 94 ---------
.../helm/templates/configmap-env.yaml | 36 ----
.../daemonset-edgecraftrag-server.yaml | 58 ------
.../templates/daemonset-llm-serving-xpu.yaml | 58 ------
.../helm/templates/deployment-ecrag.yaml | 45 -----
.../templates/deployment-edgecraftrag-ui.yaml | 45 -----
.../helm/templates/service-ecrag.yaml | 11 --
.../service-edgecraftrag-server.yaml | 11 --
.../templates/service-edgecraftrag-ui.yaml | 13 --
.../templates/service-llm-serving-xpu.yaml | 11 --
EdgeCraftRAG/kubernetes/helm/values.yaml | 57 ------
EdgeCraftRAG/tests/configs/agent.json | 12 --
EdgeCraftRAG/tests/configs/pipeline.json | 42 ----
EdgeCraftRAG/tools/README.md | 179 +++++++++++++++--
EdgeCraftRAG/tools/README_zh.md | 181 ++++++++++++++++++
EdgeCraftRAG/tools/quick_start.sh | 42 ++--
33 files changed, 410 insertions(+), 688 deletions(-)
create mode 100644 EdgeCraftRAG/assets/img/Explore_Edge_Craft_RAG_01.jpg
create mode 100644 EdgeCraftRAG/assets/img/Explore_Edge_Craft_RAG_02.jpg
create mode 100644 EdgeCraftRAG/assets/img/Explore_Edge_Craft_RAG_03.jpg
create mode 100644 EdgeCraftRAG/assets/img/Explore_Edge_Craft_RAG_04.jpg
create mode 100644 EdgeCraftRAG/assets/img/Explore_Edge_Craft_RAG_05.jpg
create mode 100644 EdgeCraftRAG/assets/img/Explore_Edge_Craft_RAG_06.jpg
create mode 100644 EdgeCraftRAG/assets/img/Explore_Edge_Craft_RAG_07.jpg
create mode 100644 EdgeCraftRAG/assets/img/Explore_Edge_Craft_RAG_08.jpg
create mode 100644 EdgeCraftRAG/assets/img/Explore_Edge_Craft_RAG_09.png
create mode 100644 EdgeCraftRAG/assets/img/Explore_Edge_Craft_RAG_10.png
create mode 100644 EdgeCraftRAG/assets/img/Explore_Edge_Craft_RAG_11.png
delete mode 100644 EdgeCraftRAG/kubernetes/helm/Chart.yaml
delete mode 100644 EdgeCraftRAG/kubernetes/helm/README.md
delete mode 100644 EdgeCraftRAG/kubernetes/helm/README_zh.md
delete mode 100644 EdgeCraftRAG/kubernetes/helm/templates/configmap-env.yaml
delete mode 100644 EdgeCraftRAG/kubernetes/helm/templates/daemonset-edgecraftrag-server.yaml
delete mode 100644 EdgeCraftRAG/kubernetes/helm/templates/daemonset-llm-serving-xpu.yaml
delete mode 100644 EdgeCraftRAG/kubernetes/helm/templates/deployment-ecrag.yaml
delete mode 100644 EdgeCraftRAG/kubernetes/helm/templates/deployment-edgecraftrag-ui.yaml
delete mode 100644 EdgeCraftRAG/kubernetes/helm/templates/service-ecrag.yaml
delete mode 100644 EdgeCraftRAG/kubernetes/helm/templates/service-edgecraftrag-server.yaml
delete mode 100644 EdgeCraftRAG/kubernetes/helm/templates/service-edgecraftrag-ui.yaml
delete mode 100644 EdgeCraftRAG/kubernetes/helm/templates/service-llm-serving-xpu.yaml
delete mode 100644 EdgeCraftRAG/kubernetes/helm/values.yaml
delete mode 100644 EdgeCraftRAG/tests/configs/agent.json
delete mode 100644 EdgeCraftRAG/tests/configs/pipeline.json
create mode 100644 EdgeCraftRAG/tools/README_zh.md
diff --git a/EdgeCraftRAG/assets/img/Explore_Edge_Craft_RAG_01.jpg b/EdgeCraftRAG/assets/img/Explore_Edge_Craft_RAG_01.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..99c76a97472194ed7677f47b19d3beda74e27ee5
GIT binary patch
literal 192368
zcmeFYdss~U_cuOnhvYO8bw@R*gfOI(RQq;JMROBT7XhA%|*^gk(f1
zR9nY!9+_$=nx>&qNi&*i&$MUHv&MaYKHuN>d;Nab^|^k}U(fa2TwSqeX7BZWt+ih3
zwbuJR+7H@h%4m;e?#n0x1`MEVfxi@OBW1DffB##5bT|3mx`FP)+GiAlp#uYlj2Jjz
zGUYde0Rs&NXe%i+3T43HzwS->`_F*i1`ZlLL~p46u;K8Ag3*-U1`Hhd+n|Ai2M-zq
z?~aDoDT53KkC`-Q;SfXbt$LF~$J!oEzBbfiQC``&RW)+Ux!d+1(;sGJ{JY6`t0~q~
zr%kuBpXcB>-)ZrZrOVuwyLk{0Qg
zudT9gZS9Jm-O8TcK3t{Nt!n^f;J+dJYhnLmT?VkO-v$jDI7m;ot^vP=!_Pp2L4zmF
z88T*}x8BxJ!^yTshmKv8d@Zj`-(v16`M7QSYlazF+DWVwx~2VE*?+dMWBnK9&ZlV5~Gg|TmVxZm$Jvt|X~3%2t#
z33&Xt=yFw7ME+7Xo+eJnL?j{_*04`D?yuDI<_Cms!qQO5TKWrh1F8C?+Lr@lFc
z;A?3kk#89h6SNflnFF<(_|NDPN;$
z^1H6%(HdvD136}co|a6W_smRB>(|#(JtyDF(qtusRqb#L30Zpk)}6rCgv?^v)3}^_cf^-l%`M))
zB0OF=&yO4x;CJ(2>YCRUT%S<#b@}vln}p-O+hps}1FbGN?XEK}IhmPRhX}DoMvFsd
zcr$YX`%OLa(1rhOoin`!rMYJl+iAZS`({Rj`Jy8L+qkWZB@?3kH`d5%6@iEEWVJ->
zH}kJFXt&*!xA11M?+SkB0v;{Gh`$zlM=uVN*ti__5bmtEUuo1)8Rb%%d$VnxJo=7w
zH749y?D@)N-f;ikc;3B9NexpG*3Y}BJ-Psu`@NplT7LrJo)l=Q~#<SgFS+n2L>9?0gK7hKm}PRW>V{?>i#SB+}~+Jcpf
zjqt-hitrO`OI)wAbZpJ|1NE$xS0gGa8u)R_3y(JU4Zq#t#Lf)OUiZ6&=F9c=R{AXG
z1zU&O%fr7f)lvq}#>4vqBl@Mj6&i=~Gc2b%*D?5Q=Q&ws{`@9d{SVj4u}?*vUup?@
z_Qnl?g|-c4Lyn0aBr-egJP#LbzPV?ymEd;d{PR{L4@YxX>R
z+}&N~lkkveUdNb*-Q(A0T_vm(8_Ha-jr?4CBiQa0W*DSqpT9lxefy!yURj}my?k3w
zi*8iDRD48B(Ys5|BOAHK_gEt?^G+c8+KG@Jd!hSjK5a$
z{_aHK>&tf?-d~-vF5zvphf-l~w9ju%1?%AH+>0`ko0H>4>~k>Om*LPp_Wt>OAN=UK
zzc=6Z-*^A5(fEi2{DOS_vD-Zf>t>kc6}d6TG?_#?hK$37^IH4{m;L$b=N{fjW`bi^
zY((MU9(B=^b!iN}nlS#_XD!(&&MRgsXV+JMB?aw0p%-5@=VkCm?(rn=5KILsaj4KR
zR@OL2DiBH6v4iPbroBJFHNu4z8K$)b_TdpV-)9^=?eVs~(|q-=fo|t+JYU3p
zi{Uevo%RZ}*GH4YsNx;2NAhV`yyWWD*Y_pavFl^w$yXC7k4r8b$>^_VSSJaUS85$1Y6bNAWZ^}U1B
zuRkozaIqEbIDPu=oaiBInv;AgoHv)B;Rmc+Au~MI(q#G)kF$8aPvutIJ#}ZiNB++=
z7iW=KQk7)(IMu$3FC*`*Ka#$yHfj1fm(wl82qE9JF&_6!$UgESG;)^lgmTu>%lmpG
zFSN`JKmBU@4xcZXBbrK9WDh)AzD-d(RDRW(+o2Teah0=y`?RzDRS=O4iCs
z$1nEZj(*@uC&S~gGC#?x)?e85WAD}{)bT|J4=^hj
zANu|3C7M*U`36e8xsvHA7o1>CYPnJ}ynK8Wk=h-xsI_aE;t}^&+qOkZn7h{tUYsMqQF$>fbTT`tCpwiDv6XQIXC!`e_S<}^_62K^
zZ%3%&TU{+i!C
z<`v({=Af_?U$0kCJoR1X;9|0BG=F;{H?9L8q<@<
z4G!JALQ9$Px=oQW?1F8^#+Xy}belE%e&4e{>-7G#>#}k2tGP4$yMz}mywlJtm}aw^
zEFw;~gvXsLTI}-t4jy=En{E5!rXABmW(k^T_SIj#O4nt2j9D$4F`;r+VD0Yn>uj37
zD_>5%j?F?M|07xTpUSWQ6W%8!!lh92Shn&vTFNO_c*{tn!p~H=Vp`~}_A$0OpEfxkqcaJ~cQ;}5?o9;ids;9H!#=J*ApO*#Q=efS_0(IQZLymBtO)^ny-iQ$71F*~=OHy4-*OFl
zx{muzDv;l4_+(D!<>r!gEhQ7YPRwEZb~AUr9O>p>r+j)Vh?J^OX{*dKP{eHF=LO3nQ6?s=3{y?Q)Yp6C3d3W
zkAiw)l2n{u)`89Xk
zcd!mx8g=uBaOsU(dYB*StAu-{Kuqcj-&5B8*C@hzjM1it-U^L{F*INqp#Dp
z=H0QE{Ju)|tUmL_w3lOMhds!$&b&Rptc<_x+W1jxroEiCdpgzdcx$Hj%Ny4cPpomQ
zPP;~bwDW@F|Y&!9OO+KbK@>qz9yg=Iq{@
zFrC7<_~-1^Bd$K+cNa0zo9sUqURRV|jM%vLXIN3iXx0d><)?CM!F82YpZT@@`)#TV
z`V;%WlR|-A~d+Vn8w5Imm%x5HfvF@j4Uf8nZ
z^3HWLp4e|)D4jF*+u_}5pU!D+oQvdI?y#{@u5IXC_Fb8Cg=2h@^~T0ePUbm$v?%Gf
zD#-T?oq6}holP4D9JGBi=+xcG=PthAb~o>=G_ZMN3q4`Yr463;JGR};-9GO8xic@H
z6-_)gbmFOFbNsC2A9+eG#kfsB(C!Dq6P4jsEd@-Y{id(%
zs^seDd69nGS?OT&$DwWwv>FoQ!d;Wnlp5DuDhRiu<7}B2ioa{r9N9-EhElvUf)o?RL^R>
zaM#vv>n%sjZ66g>%d%8CruSwjzY%q0ZY3B>e?b5eLlYYPI>@@qciACz5IjZuf
zG|Hv8gacZNwWG~}DSSfSZ+GHhXYsPIj#K-CdIN=tiGfq|udLja-}O%A_~Nt8=glF!
zmL69QG_+?vOue2!`A>ej8HwDVmOxbgNJW=sovSg=iMJGgtffp|A?}?^xN0eNF_Nh6
zJ8jqUOMJfs_+$%|b~ZKS@8}CVJW@;9=5QQSmSv05d^3sJTFP0dU}R;8KH3(;$bk;`
z71obxDSPMsYcu@EuImX>r{5B$Uj0vhsIM9LR0qOPyk7H&i0+>rA@snrZ?c>cd-o|?
zv!x3@BY(;qW
zX^qwNurq3y0Vfm#6XcOAM{uZ@7MbjvU?GSIPBLrWXL!z*ZCUN}dTtv1QO5W+UuUeH
zer~#(!E}ycoM+0&wdZKl6c>3^+-JfSU%{ODdt{ed!agO&+iWvhKpX5VjlI7LIuqZ#
z*eP}vrnAPZ%9`OEcz50;EbwI3+N_({Y*%x%%(;gL2Lj4G$XTS=hqd~B&rDiNc+AG~
z`>Z*fa)(1tE^sU&0-l-Ej2}ki7B6>-n)k-Ci7p>b>hD#A
z^fEPh&LN7zW?856W%Rq2q6pwSVTtT%#K<~gWXR!L5uBA+>ZaPe#kMTB(D1{1w!F7L
zy+0(cz4wBXYu1!h+6P`n&y-8d-i(AcPCRSNt(N=3*R$*7ADT>FIl39fR_<88)`IG5
z`QobNT+M+y&9b2_KYKMdjcpT>wt5xh%wQCFc%H`t%9LYg|m$)CXI?x2R<;%-OuoLSXo5<-o?Wz<$G}2kR96r#nP1L(hkq
zj_4k7Wx|oyvp;4&e?aftv~2G5>++8WSpr+~t@=mM3@d$)^Y#~gE0_J5(tGDd`dbfI
z!+kEp=06&9VdLpoX4)2oifv)D{zs(dd|X-Ud1?IXsXH!(^m1w+6m3(CKD~kaXJ*{F
z9b<1hh4wKV@CD8DOO1#rc%Zyq>XW#E`^