@mariozechner/pi-ai 0.37.7 → 0.38.0

package/dist/stream.js.map

@@ -1 +1 @@
-{"version":3,"file":"stream.js","sourceRoot":"","sources":["../src/stream.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,UAAU,EAAE,MAAM,SAAS,CAAC;AACrC,OAAO,EAAE,OAAO,EAAE,MAAM,SAAS,CAAC;AAClC,OAAO,EAAE,IAAI,EAAE,MAAM,WAAW,CAAC;AACjC,OAAO,EAAE,aAAa,EAAE,MAAM,aAAa,CAAC;AAC5C,OAAO,EAAyB,eAAe,EAAE,MAAM,0BAA0B,CAAC;AAClF,OAAO,EAAsB,YAAY,EAAE,MAAM,uBAAuB,CAAC;AACzE,OAAO,EAGN,qBAAqB,GACrB,MAAM,kCAAkC,CAAC;AAC1C,OAAO,EAA4B,kBAAkB,EAAE,MAAM,8BAA8B,CAAC;AAC5F,OAAO,EAAoC,0BAA0B,EAAE,MAAM,uCAAuC,CAAC;AACrH,OAAO,EAAiC,uBAAuB,EAAE,MAAM,mCAAmC,CAAC;AAC3G,OAAO,EAA+B,qBAAqB,EAAE,MAAM,iCAAiC,CAAC;AAarG,MAAM,2BAA2B,GAAG,IAAI,CAAC,OAAO,EAAE,EAAE,SAAS,EAAE,QAAQ,EAAE,sCAAsC,CAAC,CAAC;AAEjH,IAAI,gCAAgC,GAAmB,IAAI,CAAC;AAE5D,SAAS,uBAAuB,GAAY;IAC3C,IAAI,gCAAgC,KAAK,IAAI,EAAE,CAAC;QAC/C,gCAAgC,GAAG,UAAU,CAAC,2BAA2B,CAAC,CAAC;IAC5E,CAAC;IACD,OAAO,gCAAgC,CAAC;AAAA,CACxC;AASD,MAAM,UAAU,YAAY,CAAC,QAAa,EAAsB;IAC/D,qCAAqC;IACrC,IAAI,QAAQ,KAAK,gBAAgB,EAAE,CAAC;QACnC,OAAO,OAAO,CAAC,GAAG,CAAC,oBAAoB,IAAI,OAAO,CAAC,GAAG,CAAC,QAAQ,IAAI,OAAO,CAAC,GAAG,CAAC,YAAY,CAAC;IAC7F,CAAC;IAED,gEAAgE;IAChE,IAAI,QAAQ,KAAK,WAAW,EAAE,CAAC;QAC9B,OAAO,OAAO,CAAC,GAAG,CAAC,qBAAqB,IAAI,OAAO,CAAC,GAAG,CAAC,iBAAiB,CAAC;IAC3E,CAAC;IAED,gEAAgE;IAChE,kEAAkE;IAClE,IAAI,QAAQ,KAAK,eAAe,EAAE,CAAC;QAClC,MAAM,cAAc,GAAG,uBAAuB,EAAE,CAAC;QACjD,MAAM,UAAU,GAAG,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,oBAAoB,IAAI,OAAO,CAAC,GAAG,CAAC,cAAc,CAAC,CAAC;QACtF,MAAM,WAAW,GAAG,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,qBAAqB,CAAC;QAExD,IAAI,cAAc,IAAI,UAAU,IAAI,WAAW,EAAE,CAAC;YACjD,OAAO,iBAAiB,CAAC;QAC1B,CAAC;IACF,CAAC;IAED,MAAM,MAAM,GAA2B;QACtC,MAAM,EAAE,gBAAgB;QACxB,MAAM,EAAE,gBAAgB;QACxB,IAAI,EAAE,cAAc;QACpB,QAAQ,EAAE,kBAAkB;QAC5B,GAAG,EAAE,aAAa;QAClB,UAAU,EAAE,oBAAoB;QAChC,GAAG,EAAE,aAAa;QAClB,OAAO,EAAE,iBAAiB;KAC1B,CAAC;IAEF,MAAM,MAAM,GAAG,MAAM,CAAC,QAAQ,CAAC,CAAC;IAChC,OAAO,MAAM,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;AAAA,CAChD;AAED,MAAM,UAAU,MAAM,CACrB,KAAkB,EAClB,OAAgB,EAChB,OAA6B,EACC;IAC9B,+DAA+D;IAC/D,IAAI,KAAK,CAAC,GAAG,KAAK,eAAe,EAAE,CAAC;QACnC,OAAO,kBAAkB,CAAC,KAA+B,EAAE,OAAO,EAAE,OAA8B,CAAC,CAAC;IACrG,CAAC;IAED,MAAM,MAAM,GAAG,OAAO,EAAE,MAAM,IAAI,YAAY,CAAC,KAAK,CAAC,QAAQ,CAAC,CAAC;IAC/D,IAAI,CAAC,MAAM,EAAE,CAAC;QACb,MAAM,IAAI,KAAK,CAAC,4BAA4B,KAAK,CAAC,QAAQ,EAAE,CAAC,CAAC;IAC/D,CAAC;IACD,MAAM,eAAe,GAAG,EAAE,GAAG,OAAO,EAAE,MAAM,EAAE,CAAC;IAE/C,MAAM,GAAG,GAAQ,KAAK,CAAC,GAAG,CAAC;IAC3B,QAAQ,GAAG,EAAE,CAAC;QACb,KAAK,oBAAoB;YACxB,OAAO,eAAe,CAAC,KAAoC,EAAE,OAAO,EAAE,eAAe,CAAC,CAAC;QAExF,KAAK,oBAAoB;YACxB,OAAO,uBAAuB,CAAC,KAAoC,EAAE,OAAO,EAAE,eAAsB,CAAC,CAAC;QAEvG,KAAK,kBAAkB;YACtB,OAAO,qBAAqB,CAAC,KAAkC,EAAE,OAAO,EAAE,eAAsB,CAAC,CAAC;QAEnG,KAAK,wBAAwB;YAC5B,OAAO,0BAA0B,CAAC,KAAwC,EAAE,OAAO,EAAE,eAAsB,CAAC,CAAC;QAE9G,KAAK,sBAAsB;YAC1B,OAAO,YAAY,CAAC,KAAsC,EAAE,OAAO,EAAE,eAAe,CAAC,CAAC;QAEvF,KAAK,mBAAmB;YACvB,OAAO,qBAAqB,CAC3B,KAAmC,EACnC,OAAO,EACP,eAAyC,CACzC,CAAC;QAEH,SAAS,CAAC;YACT,4DAA4D;YAC5D,MAAM,WAAW,GAAU,GAAG,CAAC;YAC/B,MAAM,IAAI,KAAK,CAAC,kBAAkB,WAAW,EAAE,CAAC,CAAC;QAClD,CAAC;IACF,CAAC;AAAA,CACD;AAED,MAAM,CAAC,KAAK,UAAU,QAAQ,CAC7B,KAAkB,EAClB,OAAgB,EAChB,OAA6B,EACD;IAC5B,MAAM,CAAC,GAAG,MAAM,CAAC,KAAK,EAAE,OAAO,EAAE,OAAO,CAAC,CAAC;IAC1C,OAAO,CAAC,CAAC,MAAM,EAAE,CAAC;AAAA,CAClB;AAED,MAAM,UAAU,YAAY,CAC3B,KAAkB,EAClB,OAAgB,EAChB,OAA6B,EACC;IAC9B,+DAA+D;IAC/D,IAAI,KAAK,CAAC,GAAG,KAAK,eAAe,EAAE,CAAC;QACnC,MAAM,eAAe,GAAG,gBAAgB,CAAC,KAAK,EAAE,OAAO,EAAE,SAAS,CAAC,CAAC;QACpE,OAAO,MAAM,CAAC,KAAK,EAAE,OAAO,EAAE,eAAe,CAAC,CAAC;IAChD,CAAC;IAED,MAAM,MAAM,GAAG,OAAO,EAAE,MAAM,IAAI,YAAY,CAAC,KAAK,CAAC,QAAQ,CAAC,CAAC;IAC/D,IAAI,CAAC,MAAM,EAAE,CAAC;QACb,MAAM,IAAI,KAAK,CAAC,4BAA4B,KAAK,CAAC,QAAQ,EAAE,CAAC,CAAC;IAC/D,CAAC;IAED,MAAM,eAAe,GAAG,gBAAgB,CAAC,KAAK,EAAE,OAAO,EAAE,MAAM,CAAC,CAAC;IACjE,OAAO,MAAM,CAAC,KAAK,EAAE,OAAO,EAAE,eAAe,CAAC,CAAC;AAAA,CAC/C;AAED,MAAM,CAAC,KAAK,UAAU,cAAc,CACnC,KAAkB,EAClB,OAAgB,EAChB,OAA6B,EACD;IAC5B,MAAM,CAAC,GAAG,YAAY,CAAC,KAAK,EAAE,OAAO,EAAE,OAAO,CAAC,CAAC;IAChD,OAAO,CAAC,CAAC,MAAM,EAAE,CAAC;AAAA,CAClB;AAED,SAAS,gBAAgB,CACxB,KAAkB,EAClB,OAA6B,EAC7B,MAAe,EACO;IACtB,MAAM,IAAI,GAAG;QACZ,WAAW,EAAE,OAAO,EAAE,WAAW;QACjC,SAAS,EAAE,OAAO,EAAE,SAAS,IAAI,IAAI,CAAC,GAAG,CAAC,KAAK,CAAC,SAAS,EAAE,KAAK,CAAC;QACjE,MAAM,EAAE,OAAO,EAAE,MAAM;QACvB,MAAM,EAAE,MAAM,IAAI,OAAO,EAAE,MAAM;QACjC,SAAS,EAAE,OAAO,EAAE,SAAS;KAC7B,CAAC;IAEF,oEAAoE;IACpE,MAAM,cAAc,GAAG,CAAC,MAAiC,EAAE,EAAE,CAAC,CAAC,MAAM,KAAK,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC;IAErG,QAAQ,KAAK,CAAC,GAAG,EAAE,CAAC;QACnB,KAAK,oBAAoB,EAAE,CAAC;YAC3B,8DAA8D;YAC9D,IAAI,CAAC,OAAO,EAAE,SAAS,EAAE,CAAC;gBACzB,OAAO,EAAE,GAAG,IAAI,EAAE,eAAe,EAAE,KAAK,EAA6B,CAAC;YACvE,CAAC;YAED,sDAAsD;YACtD,uEAAuE;YACvE,MAAM,gBAAgB,GAAG;gBACxB,OAAO,EAAE,IAAI;gBACb,GAAG,EAAE,IAAI;gBACT,MAAM,EAAE,IAAI;gBACZ,IAAI,EAAE,KAAK;aACX,CAAC;YAEF,MAAM,eAAe,GAAG,IAAI,CAAC;YAC7B,IAAI,cAAc,GAAG,gBAAgB,CAAC,cAAc,CAAC,OAAO,CAAC,SAAS,CAAE,CAAC,CAAC;YAC1E,8FAA8F;YAC9F,MAAM,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,IAAI,CAAC,SAAS,IAAI,CAAC,CAAC,GAAG,cAAc,EAAE,KAAK,CAAC,SAAS,CAAC,CAAC;YAEpF,mEAAmE;YACnE,IAAI,SAAS,IAAI,cAAc,EAAE,CAAC;gBACjC,cAAc,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,SAAS,GAAG,eAAe,CAAC,CAAC;YAC3D,CAAC;YAED,OAAO;gBACN,GAAG,IAAI;gBACP,SAAS;gBACT,eAAe,EAAE,IAAI;gBACrB,oBAAoB,EAAE,cAAc;aACT,CAAC;QAC9B,CAAC;QAED,KAAK,oBAAoB;YACxB,OAAO;gBACN,GAAG,IAAI;gBACP,eAAe,EAAE,aAAa,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,OAAO,EAAE,SAAS,CAAC,CAAC,CAAC,cAAc,CAAC,OAAO,EAAE,SAAS,CAAC;aAC5D,CAAC;QAEtC,KAAK,kBAAkB;YACtB,OAAO;gBACN,GAAG,IAAI;gBACP,eAAe,EAAE,aAAa,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,OAAO,EAAE,SAAS,CAAC,CAAC,CAAC,cAAc,CAAC,OAAO,EAAE,SAAS,CAAC;aAC9D,CAAC;QAEpC,KAAK,wBAAwB;YAC5B,OAAO;gBACN,GAAG,IAAI;gBACP,eAAe,EAAE,aAAa,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,OAAO,EAAE,SAAS,CAAC,CAAC,CAAC,cAAc,CAAC,OAAO,EAAE,SAAS,CAAC;aACzD,CAAC;QAEzC,KAAK,sBAAsB,EAAE,CAAC;YAC7B,8DAA8D;YAC9D,0EAA0E;YAC1E,IAAI,CAAC,OAAO,EAAE,SAAS,EAAE,CAAC;gBACzB,OAAO,EAAE,GAAG,IAAI,EAAE,QAAQ,EAAE,EAAE,OAAO,EAAE,KAAK,EAAE,EAA0B,CAAC;YAC1E,CAAC;YAED,MAAM,WAAW,GAAG,KAAsC,CAAC;YAC3D,MAAM,MAAM,GAAG,cAAc,CAAC,OAAO,CAAC,SAAS,CAAE,CAAC;YAElD,2EAA2E;YAC3E,4DAA4D;YAC5D,IAAI,iBAAiB,CAAC,WAAW,CAAC,IAAI,mBAAmB,CAAC,WAAW,CAAC,EAAE,CAAC;gBACxE,OAAO;oBACN,GAAG,IAAI;oBACP,QAAQ,EAAE;wBACT,OAAO,EAAE,IAAI;wBACb,KAAK,EAAE,uBAAuB,CAAC,MAAM,EAAE,WAAW,CAAC;qBACnD;iBACuB,CAAC;YAC3B,CAAC;YAED,OAAO;gBACN,GAAG,IAAI;gBACP,QAAQ,EAAE;oBACT,OAAO,EAAE,IAAI;oBACb,YAAY,EAAE,eAAe,CAAC,WAAW,EAAE,MAAM,CAAC;iBAClD;aACuB,CAAC;QAC3B,CAAC;QAED,KAAK,mBAAmB,EAAE,CAAC;YAC1B,IAAI,CAAC,OAAO,EAAE,SAAS,EAAE,CAAC;gBACzB,OAAO,EAAE,GAAG,IAAI,EAAE,QAAQ,EAAE,EAAE,OAAO,EAAE,KAAK,EAAE,EAAmC,CAAC;YACnF,CAAC;YAED,MAAM,MAAM,GAAG,cAAc,CAAC,OAAO,CAAC,SAAS,CAAE,CAAC;YAElD,8DAA8D;YAC9D,IAAI,KAAK,CAAC,EAAE,CAAC,QAAQ,CAAC,OAAO,CAAC,IAAI,KAAK,CAAC,EAAE,CAAC,QAAQ,CAAC,SAAS,CAAC,EAAE,CAAC;gBAChE,OAAO;oBACN,GAAG,IAAI;oBACP,QAAQ,EAAE;wBACT,OAAO,EAAE,IAAI;wBACb,KAAK,EAAE,yBAAyB,CAAC,MAAM,EAAE,KAAK,CAAC,EAAE,CAAC;qBAClD;iBACgC,CAAC;YACpC,CAAC;YAED,mEAAmE;YACnE,sDAAsD;YACtD,uEAAuE;YACvE,MAAM,OAAO,GAAyC;gBACrD,OAAO,EAAE,IAAI;gBACb,GAAG,EAAE,IAAI;gBACT,MAAM,EAAE,IAAI;gBACZ,IAAI,EAAE,KAAK;aACX,CAAC;YAEF,MAAM,eAAe,GAAG,IAAI,CAAC;YAC7B,IAAI,cAAc,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC;YACrC,8FAA8F;YAC9F,MAAM,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,IAAI,CAAC,SAAS,IAAI,CAAC,CAAC,GAAG,cAAc,EAAE,KAAK,CAAC,SAAS,CAAC,CAAC;YAEpF,mEAAmE;YACnE,IAAI,SAAS,IAAI,cAAc,EAAE,CAAC;gBACjC,cAAc,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,SAAS,GAAG,eAAe,CAAC,CAAC;YAC3D,CAAC;YAED,OAAO;gBACN,GAAG,IAAI;gBACP,SAAS;gBACT,QAAQ,EAAE;oBACT,OAAO,EAAE,IAAI;oBACb,YAAY,EAAE,cAAc;iBAC5B;aACgC,CAAC;QACpC,CAAC;QAED,KAAK,eAAe,EAAE,CAAC;YACtB,8DAA8D;YAC9D,IAAI,CAAC,OAAO,EAAE,SAAS,EAAE,CAAC;gBACzB,OAAO,EAAE,GAAG,IAAI,EAAE,QAAQ,EAAE,EAAE,OAAO,EAAE,KAAK,EAAE,EAAgC,CAAC;YAChF,CAAC;YAED,MAAM,WAAW,GAAG,KAA+B,CAAC;YACpD,MAAM,MAAM,GAAG,cAAc,CAAC,OAAO,CAAC,SAAS,CAAE,CAAC;YAClD,MAAM,WAAW,GAAG,WAAuD,CAAC;YAE5E,IAAI,iBAAiB,CAAC,WAAW,CAAC,IAAI,mBAAmB,CAAC,WAAW,CAAC,EAAE,CAAC;gBACxE,OAAO;oBACN,GAAG,IAAI;oBACP,QAAQ,EAAE;wBACT,OAAO,EAAE,IAAI;wBACb,KAAK,EAAE,uBAAuB,CAAC,MAAM,EAAE,WAAW,CAAC;qBACnD;iBAC6B,CAAC;YACjC,CAAC;YAED,OAAO;gBACN,GAAG,IAAI;gBACP,QAAQ,EAAE;oBACT,OAAO,EAAE,IAAI;oBACb,YAAY,EAAE,eAAe,CAAC,WAAW,EAAE,MAAM,CAAC;iBAClD;aAC6B,CAAC;QACjC,CAAC;QAED,SAAS,CAAC;YACT,uBAAuB;YACvB,MAAM,WAAW,GAAU,KAAK,CAAC,GAAG,CAAC;YACrC,MAAM,IAAI,KAAK,CAAC,sCAAsC,WAAW,EAAE,CAAC,CAAC;QACtE,CAAC;IACF,CAAC;AAAA,CACD;AAID,SAAS,iBAAiB,CAAC,KAAoC,EAAW;IACzE,2FAA2F;IAC3F,OAAO,KAAK,CAAC,EAAE,CAAC,QAAQ,CAAC,OAAO,CAAC,CAAC;AAAA,CAClC;AAED,SAAS,mBAAmB,CAAC,KAAoC,EAAW;IAC3E,+FAA+F;IAC/F,OAAO,KAAK,CAAC,EAAE,CAAC,QAAQ,CAAC,SAAS,CAAC,CAAC;AAAA,CACpC;AAED,SAAS,uBAAuB,CAC/B,MAA4B,EAC5B,KAAoC,EACd;IACtB,IAAI,iBAAiB,CAAC,KAAK,CAAC,EAAE,CAAC;QAC9B,gDAAgD;QAChD,QAAQ,MAAM,EAAE,CAAC;YAChB,KAAK,SAAS,CAAC;YACf,KAAK,KAAK;gBACT,OAAO,KAAK,CAAC;YACd,KAAK,QAAQ,CAAC;YACd,KAAK,MAAM;gBACV,OAAO,MAAM,CAAC;QAChB,CAAC;IACF,CAAC;IACD,0CAA0C;IAC1C,QAAQ,MAAM,EAAE,CAAC;QAChB,KAAK,SAAS;YACb,OAAO,SAAS,CAAC;QAClB,KAAK,KAAK;YACT,OAAO,KAAK,CAAC;QACd,KAAK,QAAQ;YACZ,OAAO,QAAQ,CAAC;QACjB,KAAK,MAAM;YACV,OAAO,MAAM,CAAC;IAChB,CAAC;AAAA,CACD;AAED,SAAS,yBAAyB,CAAC,MAA4B,EAAE,OAAe,EAAuB;IACtG,IAAI,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAC,EAAE,CAAC;QAC/B,gDAAgD;QAChD,QAAQ,MAAM,EAAE,CAAC;YAChB,KAAK,SAAS,CAAC;YACf,KAAK,KAAK;gBACT,OAAO,KAAK,CAAC;YACd,KAAK,QAAQ,CAAC;YACd,KAAK,MAAM;gBACV,OAAO,MAAM,CAAC;QAChB,CAAC;IACF,CAAC;IACD,0CAA0C;IAC1C,QAAQ,MAAM,EAAE,CAAC;QAChB,KAAK,SAAS;YACb,OAAO,SAAS,CAAC;QAClB,KAAK,KAAK;YACT,OAAO,KAAK,CAAC;QACd,KAAK,QAAQ;YACZ,OAAO,QAAQ,CAAC;QACjB,KAAK,MAAM;YACV,OAAO,MAAM,CAAC;IAChB,CAAC;AAAA,CACD;AAED,SAAS,eAAe,CAAC,KAAoC,EAAE,MAA4B,EAAU;IACpG,gEAAgE;IAChE,IAAI,KAAK,CAAC,EAAE,CAAC,QAAQ,CAAC,SAAS,CAAC,EAAE,CAAC;QAClC,MAAM,OAAO,GAAyC;YACrD,OAAO,EAAE,GAAG;YACZ,GAAG,EAAE,IAAI;YACT,MAAM,EAAE,IAAI;YACZ,IAAI,EAAE,KAAK;SACX,CAAC;QACF,OAAO,OAAO,CAAC,MAAM,CAAC,CAAC;IACxB,CAAC;IAED,IAAI,KAAK,CAAC,EAAE,CAAC,QAAQ,CAAC,WAAW,CAAC,EAAE,CAAC;QACpC,gCAAgC;QAChC,MAAM,OAAO,GAAyC;YACrD,OAAO,EAAE,GAAG;YACZ,GAAG,EAAE,IAAI;YACT,MAAM,EAAE,IAAI;YACZ,IAAI,EAAE,KAAK;SACX,CAAC;QACF,OAAO,OAAO,CAAC,MAAM,CAAC,CAAC;IACxB,CAAC;IAED,8BAA8B;IAC9B,OAAO,CAAC,CAAC,CAAC;AAAA,CACV","sourcesContent":["import { existsSync } from \"node:fs\";\nimport { homedir } from \"node:os\";\nimport { join } from \"node:path\";\nimport { supportsXhigh } from \"./models.js\";\nimport { type AnthropicOptions, streamAnthropic } from \"./providers/anthropic.js\";\nimport { type GoogleOptions, streamGoogle } from \"./providers/google.js\";\nimport {\n\ttype GoogleGeminiCliOptions,\n\ttype GoogleThinkingLevel,\n\tstreamGoogleGeminiCli,\n} from \"./providers/google-gemini-cli.js\";\nimport { type GoogleVertexOptions, streamGoogleVertex } from \"./providers/google-vertex.js\";\nimport { type OpenAICodexResponsesOptions, streamOpenAICodexResponses } from \"./providers/openai-codex-responses.js\";\nimport { type OpenAICompletionsOptions, streamOpenAICompletions } from \"./providers/openai-completions.js\";\nimport { type OpenAIResponsesOptions, streamOpenAIResponses } from \"./providers/openai-responses.js\";\nimport type {\n\tApi,\n\tAssistantMessage,\n\tAssistantMessageEventStream,\n\tContext,\n\tKnownProvider,\n\tModel,\n\tOptionsForApi,\n\tSimpleStreamOptions,\n\tThinkingLevel,\n} from \"./types.js\";\n\nconst VERTEX_ADC_CREDENTIALS_PATH = join(homedir(), \".config\", \"gcloud\", \"application_default_credentials.json\");\n\nlet cachedVertexAdcCredentialsExists: boolean | null = null;\n\nfunction hasVertexAdcCredentials(): boolean {\n\tif (cachedVertexAdcCredentialsExists === null) {\n\t\tcachedVertexAdcCredentialsExists = existsSync(VERTEX_ADC_CREDENTIALS_PATH);\n\t}\n\treturn cachedVertexAdcCredentialsExists;\n}\n\n/**\n * Get API key for provider from known environment variables, e.g. OPENAI_API_KEY.\n *\n * Will not return API keys for providers that require OAuth tokens.\n */\nexport function getEnvApiKey(provider: KnownProvider): string | undefined;\nexport function getEnvApiKey(provider: string): string | undefined;\nexport function getEnvApiKey(provider: any): string | undefined {\n\t// Fall back to environment variables\n\tif (provider === \"github-copilot\") {\n\t\treturn process.env.COPILOT_GITHUB_TOKEN || process.env.GH_TOKEN || process.env.GITHUB_TOKEN;\n\t}\n\n\t// ANTHROPIC_OAUTH_TOKEN takes precedence over ANTHROPIC_API_KEY\n\tif (provider === \"anthropic\") {\n\t\treturn process.env.ANTHROPIC_OAUTH_TOKEN || process.env.ANTHROPIC_API_KEY;\n\t}\n\n\t// Vertex AI uses Application Default Credentials, not API keys.\n\t// Auth is configured via `gcloud auth application-default login`.\n\tif (provider === \"google-vertex\") {\n\t\tconst hasCredentials = hasVertexAdcCredentials();\n\t\tconst hasProject = !!(process.env.GOOGLE_CLOUD_PROJECT || process.env.GCLOUD_PROJECT);\n\t\tconst hasLocation = !!process.env.GOOGLE_CLOUD_LOCATION;\n\n\t\tif (hasCredentials && hasProject && hasLocation) {\n\t\t\treturn \"<authenticated>\";\n\t\t}\n\t}\n\n\tconst envMap: Record<string, string> = {\n\t\topenai: \"OPENAI_API_KEY\",\n\t\tgoogle: \"GEMINI_API_KEY\",\n\t\tgroq: \"GROQ_API_KEY\",\n\t\tcerebras: \"CEREBRAS_API_KEY\",\n\t\txai: \"XAI_API_KEY\",\n\t\topenrouter: \"OPENROUTER_API_KEY\",\n\t\tzai: \"ZAI_API_KEY\",\n\t\tmistral: \"MISTRAL_API_KEY\",\n\t};\n\n\tconst envVar = envMap[provider];\n\treturn envVar ? process.env[envVar] : undefined;\n}\n\nexport function stream<TApi extends Api>(\n\tmodel: Model<TApi>,\n\tcontext: Context,\n\toptions?: OptionsForApi<TApi>,\n): AssistantMessageEventStream {\n\t// Vertex AI uses Application Default Credentials, not API keys\n\tif (model.api === \"google-vertex\") {\n\t\treturn streamGoogleVertex(model as Model<\"google-vertex\">, context, options as GoogleVertexOptions);\n\t}\n\n\tconst apiKey = options?.apiKey || getEnvApiKey(model.provider);\n\tif (!apiKey) {\n\t\tthrow new Error(`No API key for provider: ${model.provider}`);\n\t}\n\tconst providerOptions = { ...options, apiKey };\n\n\tconst api: Api = model.api;\n\tswitch (api) {\n\t\tcase \"anthropic-messages\":\n\t\t\treturn streamAnthropic(model as Model<\"anthropic-messages\">, context, providerOptions);\n\n\t\tcase \"openai-completions\":\n\t\t\treturn streamOpenAICompletions(model as Model<\"openai-completions\">, context, providerOptions as any);\n\n\t\tcase \"openai-responses\":\n\t\t\treturn streamOpenAIResponses(model as Model<\"openai-responses\">, context, providerOptions as any);\n\n\t\tcase \"openai-codex-responses\":\n\t\t\treturn streamOpenAICodexResponses(model as Model<\"openai-codex-responses\">, context, providerOptions as any);\n\n\t\tcase \"google-generative-ai\":\n\t\t\treturn streamGoogle(model as Model<\"google-generative-ai\">, context, providerOptions);\n\n\t\tcase \"google-gemini-cli\":\n\t\t\treturn streamGoogleGeminiCli(\n\t\t\t\tmodel as Model<\"google-gemini-cli\">,\n\t\t\t\tcontext,\n\t\t\t\tproviderOptions as GoogleGeminiCliOptions,\n\t\t\t);\n\n\t\tdefault: {\n\t\t\t// This should never be reached if all Api cases are handled\n\t\t\tconst _exhaustive: never = api;\n\t\t\tthrow new Error(`Unhandled API: ${_exhaustive}`);\n\t\t}\n\t}\n}\n\nexport async function complete<TApi extends Api>(\n\tmodel: Model<TApi>,\n\tcontext: Context,\n\toptions?: OptionsForApi<TApi>,\n): Promise<AssistantMessage> {\n\tconst s = stream(model, context, options);\n\treturn s.result();\n}\n\nexport function streamSimple<TApi extends Api>(\n\tmodel: Model<TApi>,\n\tcontext: Context,\n\toptions?: SimpleStreamOptions,\n): AssistantMessageEventStream {\n\t// Vertex AI uses Application Default Credentials, not API keys\n\tif (model.api === \"google-vertex\") {\n\t\tconst providerOptions = mapOptionsForApi(model, options, undefined);\n\t\treturn stream(model, context, providerOptions);\n\t}\n\n\tconst apiKey = options?.apiKey || getEnvApiKey(model.provider);\n\tif (!apiKey) {\n\t\tthrow new Error(`No API key for provider: ${model.provider}`);\n\t}\n\n\tconst providerOptions = mapOptionsForApi(model, options, apiKey);\n\treturn stream(model, context, providerOptions);\n}\n\nexport async function completeSimple<TApi extends Api>(\n\tmodel: Model<TApi>,\n\tcontext: Context,\n\toptions?: SimpleStreamOptions,\n): Promise<AssistantMessage> {\n\tconst s = streamSimple(model, context, options);\n\treturn s.result();\n}\n\nfunction mapOptionsForApi<TApi extends Api>(\n\tmodel: Model<TApi>,\n\toptions?: SimpleStreamOptions,\n\tapiKey?: string,\n): OptionsForApi<TApi> {\n\tconst base = {\n\t\ttemperature: options?.temperature,\n\t\tmaxTokens: options?.maxTokens || Math.min(model.maxTokens, 32000),\n\t\tsignal: options?.signal,\n\t\tapiKey: apiKey || options?.apiKey,\n\t\tsessionId: options?.sessionId,\n\t};\n\n\t// Helper to clamp xhigh to high for providers that don't support it\n\tconst clampReasoning = (effort: ThinkingLevel | undefined) => (effort === \"xhigh\" ? \"high\" : effort);\n\n\tswitch (model.api) {\n\t\tcase \"anthropic-messages\": {\n\t\t\t// Explicitly disable thinking when reasoning is not specified\n\t\t\tif (!options?.reasoning) {\n\t\t\t\treturn { ...base, thinkingEnabled: false } satisfies AnthropicOptions;\n\t\t\t}\n\n\t\t\t// Claude requires max_tokens > thinking.budget_tokens\n\t\t\t// So we need to ensure maxTokens accounts for both thinking and output\n\t\t\tconst anthropicBudgets = {\n\t\t\t\tminimal: 1024,\n\t\t\t\tlow: 2048,\n\t\t\t\tmedium: 8192,\n\t\t\t\thigh: 16384,\n\t\t\t};\n\n\t\t\tconst minOutputTokens = 1024;\n\t\t\tlet thinkingBudget = anthropicBudgets[clampReasoning(options.reasoning)!];\n\t\t\t// Caller's maxTokens is the desired output; add thinking budget on top, capped at model limit\n\t\t\tconst maxTokens = Math.min((base.maxTokens || 0) + thinkingBudget, model.maxTokens);\n\n\t\t\t// If not enough room for thinking + output, reduce thinking budget\n\t\t\tif (maxTokens <= thinkingBudget) {\n\t\t\t\tthinkingBudget = Math.max(0, maxTokens - minOutputTokens);\n\t\t\t}\n\n\t\t\treturn {\n\t\t\t\t...base,\n\t\t\t\tmaxTokens,\n\t\t\t\tthinkingEnabled: true,\n\t\t\t\tthinkingBudgetTokens: thinkingBudget,\n\t\t\t} satisfies AnthropicOptions;\n\t\t}\n\n\t\tcase \"openai-completions\":\n\t\t\treturn {\n\t\t\t\t...base,\n\t\t\t\treasoningEffort: supportsXhigh(model) ? options?.reasoning : clampReasoning(options?.reasoning),\n\t\t\t} satisfies OpenAICompletionsOptions;\n\n\t\tcase \"openai-responses\":\n\t\t\treturn {\n\t\t\t\t...base,\n\t\t\t\treasoningEffort: supportsXhigh(model) ? options?.reasoning : clampReasoning(options?.reasoning),\n\t\t\t} satisfies OpenAIResponsesOptions;\n\n\t\tcase \"openai-codex-responses\":\n\t\t\treturn {\n\t\t\t\t...base,\n\t\t\t\treasoningEffort: supportsXhigh(model) ? options?.reasoning : clampReasoning(options?.reasoning),\n\t\t\t} satisfies OpenAICodexResponsesOptions;\n\n\t\tcase \"google-generative-ai\": {\n\t\t\t// Explicitly disable thinking when reasoning is not specified\n\t\t\t// This is needed because Gemini has \"dynamic thinking\" enabled by default\n\t\t\tif (!options?.reasoning) {\n\t\t\t\treturn { ...base, thinking: { enabled: false } } satisfies GoogleOptions;\n\t\t\t}\n\n\t\t\tconst googleModel = model as Model<\"google-generative-ai\">;\n\t\t\tconst effort = clampReasoning(options.reasoning)!;\n\n\t\t\t// Gemini 3 models use thinkingLevel exclusively instead of thinkingBudget.\n\t\t\t// https://ai.google.dev/gemini-api/docs/thinking#set-budget\n\t\t\tif (isGemini3ProModel(googleModel) || isGemini3FlashModel(googleModel)) {\n\t\t\t\treturn {\n\t\t\t\t\t...base,\n\t\t\t\t\tthinking: {\n\t\t\t\t\t\tenabled: true,\n\t\t\t\t\t\tlevel: getGemini3ThinkingLevel(effort, googleModel),\n\t\t\t\t\t},\n\t\t\t\t} satisfies GoogleOptions;\n\t\t\t}\n\n\t\t\treturn {\n\t\t\t\t...base,\n\t\t\t\tthinking: {\n\t\t\t\t\tenabled: true,\n\t\t\t\t\tbudgetTokens: getGoogleBudget(googleModel, effort),\n\t\t\t\t},\n\t\t\t} satisfies GoogleOptions;\n\t\t}\n\n\t\tcase \"google-gemini-cli\": {\n\t\t\tif (!options?.reasoning) {\n\t\t\t\treturn { ...base, thinking: { enabled: false } } satisfies GoogleGeminiCliOptions;\n\t\t\t}\n\n\t\t\tconst effort = clampReasoning(options.reasoning)!;\n\n\t\t\t// Gemini 3 models use thinkingLevel instead of thinkingBudget\n\t\t\tif (model.id.includes(\"3-pro\") || model.id.includes(\"3-flash\")) {\n\t\t\t\treturn {\n\t\t\t\t\t...base,\n\t\t\t\t\tthinking: {\n\t\t\t\t\t\tenabled: true,\n\t\t\t\t\t\tlevel: getGeminiCliThinkingLevel(effort, model.id),\n\t\t\t\t\t},\n\t\t\t\t} satisfies GoogleGeminiCliOptions;\n\t\t\t}\n\n\t\t\t// Models using thinkingBudget (Gemini 2.x, Claude via Antigravity)\n\t\t\t// Claude requires max_tokens > thinking.budget_tokens\n\t\t\t// So we need to ensure maxTokens accounts for both thinking and output\n\t\t\tconst budgets: Record<ClampedThinkingLevel, number> = {\n\t\t\t\tminimal: 1024,\n\t\t\t\tlow: 2048,\n\t\t\t\tmedium: 8192,\n\t\t\t\thigh: 16384,\n\t\t\t};\n\n\t\t\tconst minOutputTokens = 1024;\n\t\t\tlet thinkingBudget = budgets[effort];\n\t\t\t// Caller's maxTokens is the desired output; add thinking budget on top, capped at model limit\n\t\t\tconst maxTokens = Math.min((base.maxTokens || 0) + thinkingBudget, model.maxTokens);\n\n\t\t\t// If not enough room for thinking + output, reduce thinking budget\n\t\t\tif (maxTokens <= thinkingBudget) {\n\t\t\t\tthinkingBudget = Math.max(0, maxTokens - minOutputTokens);\n\t\t\t}\n\n\t\t\treturn {\n\t\t\t\t...base,\n\t\t\t\tmaxTokens,\n\t\t\t\tthinking: {\n\t\t\t\t\tenabled: true,\n\t\t\t\t\tbudgetTokens: thinkingBudget,\n\t\t\t\t},\n\t\t\t} satisfies GoogleGeminiCliOptions;\n\t\t}\n\n\t\tcase \"google-vertex\": {\n\t\t\t// Explicitly disable thinking when reasoning is not specified\n\t\t\tif (!options?.reasoning) {\n\t\t\t\treturn { ...base, thinking: { enabled: false } } satisfies GoogleVertexOptions;\n\t\t\t}\n\n\t\t\tconst vertexModel = model as Model<\"google-vertex\">;\n\t\t\tconst effort = clampReasoning(options.reasoning)!;\n\t\t\tconst geminiModel = vertexModel as unknown as Model<\"google-generative-ai\">;\n\n\t\t\tif (isGemini3ProModel(geminiModel) || isGemini3FlashModel(geminiModel)) {\n\t\t\t\treturn {\n\t\t\t\t\t...base,\n\t\t\t\t\tthinking: {\n\t\t\t\t\t\tenabled: true,\n\t\t\t\t\t\tlevel: getGemini3ThinkingLevel(effort, geminiModel),\n\t\t\t\t\t},\n\t\t\t\t} satisfies GoogleVertexOptions;\n\t\t\t}\n\n\t\t\treturn {\n\t\t\t\t...base,\n\t\t\t\tthinking: {\n\t\t\t\t\tenabled: true,\n\t\t\t\t\tbudgetTokens: getGoogleBudget(geminiModel, effort),\n\t\t\t\t},\n\t\t\t} satisfies GoogleVertexOptions;\n\t\t}\n\n\t\tdefault: {\n\t\t\t// Exhaustiveness check\n\t\t\tconst _exhaustive: never = model.api;\n\t\t\tthrow new Error(`Unhandled API in mapOptionsForApi: ${_exhaustive}`);\n\t\t}\n\t}\n}\n\ntype ClampedThinkingLevel = Exclude<ThinkingLevel, \"xhigh\">;\n\nfunction isGemini3ProModel(model: Model<\"google-generative-ai\">): boolean {\n\t// Covers gemini-3-pro, gemini-3-pro-preview, and possible other prefixed ids in the future\n\treturn model.id.includes(\"3-pro\");\n}\n\nfunction isGemini3FlashModel(model: Model<\"google-generative-ai\">): boolean {\n\t// Covers gemini-3-flash, gemini-3-flash-preview, and possible other prefixed ids in the future\n\treturn model.id.includes(\"3-flash\");\n}\n\nfunction getGemini3ThinkingLevel(\n\teffort: ClampedThinkingLevel,\n\tmodel: Model<\"google-generative-ai\">,\n): GoogleThinkingLevel {\n\tif (isGemini3ProModel(model)) {\n\t\t// Gemini 3 Pro only supports LOW/HIGH (for now)\n\t\tswitch (effort) {\n\t\t\tcase \"minimal\":\n\t\t\tcase \"low\":\n\t\t\t\treturn \"LOW\";\n\t\t\tcase \"medium\":\n\t\t\tcase \"high\":\n\t\t\t\treturn \"HIGH\";\n\t\t}\n\t}\n\t// Gemini 3 Flash supports all four levels\n\tswitch (effort) {\n\t\tcase \"minimal\":\n\t\t\treturn \"MINIMAL\";\n\t\tcase \"low\":\n\t\t\treturn \"LOW\";\n\t\tcase \"medium\":\n\t\t\treturn \"MEDIUM\";\n\t\tcase \"high\":\n\t\t\treturn \"HIGH\";\n\t}\n}\n\nfunction getGeminiCliThinkingLevel(effort: ClampedThinkingLevel, modelId: string): GoogleThinkingLevel {\n\tif (modelId.includes(\"3-pro\")) {\n\t\t// Gemini 3 Pro only supports LOW/HIGH (for now)\n\t\tswitch (effort) {\n\t\t\tcase \"minimal\":\n\t\t\tcase \"low\":\n\t\t\t\treturn \"LOW\";\n\t\t\tcase \"medium\":\n\t\t\tcase \"high\":\n\t\t\t\treturn \"HIGH\";\n\t\t}\n\t}\n\t// Gemini 3 Flash supports all four levels\n\tswitch (effort) {\n\t\tcase \"minimal\":\n\t\t\treturn \"MINIMAL\";\n\t\tcase \"low\":\n\t\t\treturn \"LOW\";\n\t\tcase \"medium\":\n\t\t\treturn \"MEDIUM\";\n\t\tcase \"high\":\n\t\t\treturn \"HIGH\";\n\t}\n}\n\nfunction getGoogleBudget(model: Model<\"google-generative-ai\">, effort: ClampedThinkingLevel): number {\n\t// See https://ai.google.dev/gemini-api/docs/thinking#set-budget\n\tif (model.id.includes(\"2.5-pro\")) {\n\t\tconst budgets: Record<ClampedThinkingLevel, number> = {\n\t\t\tminimal: 128,\n\t\t\tlow: 2048,\n\t\t\tmedium: 8192,\n\t\t\thigh: 32768,\n\t\t};\n\t\treturn budgets[effort];\n\t}\n\n\tif (model.id.includes(\"2.5-flash\")) {\n\t\t// Covers 2.5-flash-lite as well\n\t\tconst budgets: Record<ClampedThinkingLevel, number> = {\n\t\t\tminimal: 128,\n\t\t\tlow: 2048,\n\t\t\tmedium: 8192,\n\t\t\thigh: 24576,\n\t\t};\n\t\treturn budgets[effort];\n\t}\n\n\t// Unknown model - use dynamic\n\treturn -1;\n}\n"]}
+{"version":3,"file":"stream.js","sourceRoot":"","sources":["../src/stream.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,UAAU,EAAE,MAAM,SAAS,CAAC;AACrC,OAAO,EAAE,OAAO,EAAE,MAAM,SAAS,CAAC;AAClC,OAAO,EAAE,IAAI,EAAE,MAAM,WAAW,CAAC;AACjC,OAAO,EAAE,aAAa,EAAE,MAAM,aAAa,CAAC;AAC5C,OAAO,EAAyB,eAAe,EAAE,MAAM,0BAA0B,CAAC;AAClF,OAAO,EAAsB,YAAY,EAAE,MAAM,uBAAuB,CAAC;AACzE,OAAO,EAGN,qBAAqB,GACrB,MAAM,kCAAkC,CAAC;AAC1C,OAAO,EAA4B,kBAAkB,EAAE,MAAM,8BAA8B,CAAC;AAC5F,OAAO,EAAoC,0BAA0B,EAAE,MAAM,uCAAuC,CAAC;AACrH,OAAO,EAAiC,uBAAuB,EAAE,MAAM,mCAAmC,CAAC;AAC3G,OAAO,EAA+B,qBAAqB,EAAE,MAAM,iCAAiC,CAAC;AAcrG,MAAM,2BAA2B,GAAG,IAAI,CAAC,OAAO,EAAE,EAAE,SAAS,EAAE,QAAQ,EAAE,sCAAsC,CAAC,CAAC;AAEjH,IAAI,gCAAgC,GAAmB,IAAI,CAAC;AAE5D,SAAS,uBAAuB,GAAY;IAC3C,IAAI,gCAAgC,KAAK,IAAI,EAAE,CAAC;QAC/C,gCAAgC,GAAG,UAAU,CAAC,2BAA2B,CAAC,CAAC;IAC5E,CAAC;IACD,OAAO,gCAAgC,CAAC;AAAA,CACxC;AASD,MAAM,UAAU,YAAY,CAAC,QAAa,EAAsB;IAC/D,qCAAqC;IACrC,IAAI,QAAQ,KAAK,gBAAgB,EAAE,CAAC;QACnC,OAAO,OAAO,CAAC,GAAG,CAAC,oBAAoB,IAAI,OAAO,CAAC,GAAG,CAAC,QAAQ,IAAI,OAAO,CAAC,GAAG,CAAC,YAAY,CAAC;IAC7F,CAAC;IAED,gEAAgE;IAChE,IAAI,QAAQ,KAAK,WAAW,EAAE,CAAC;QAC9B,OAAO,OAAO,CAAC,GAAG,CAAC,qBAAqB,IAAI,OAAO,CAAC,GAAG,CAAC,iBAAiB,CAAC;IAC3E,CAAC;IAED,gEAAgE;IAChE,kEAAkE;IAClE,IAAI,QAAQ,KAAK,eAAe,EAAE,CAAC;QAClC,MAAM,cAAc,GAAG,uBAAuB,EAAE,CAAC;QACjD,MAAM,UAAU,GAAG,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,oBAAoB,IAAI,OAAO,CAAC,GAAG,CAAC,cAAc,CAAC,CAAC;QACtF,MAAM,WAAW,GAAG,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,qBAAqB,CAAC;QAExD,IAAI,cAAc,IAAI,UAAU,IAAI,WAAW,EAAE,CAAC;YACjD,OAAO,iBAAiB,CAAC;QAC1B,CAAC;IACF,CAAC;IAED,MAAM,MAAM,GAA2B;QACtC,MAAM,EAAE,gBAAgB;QACxB,MAAM,EAAE,gBAAgB;QACxB,IAAI,EAAE,cAAc;QACpB,QAAQ,EAAE,kBAAkB;QAC5B,GAAG,EAAE,aAAa;QAClB,UAAU,EAAE,oBAAoB;QAChC,GAAG,EAAE,aAAa;QAClB,OAAO,EAAE,iBAAiB;KAC1B,CAAC;IAEF,MAAM,MAAM,GAAG,MAAM,CAAC,QAAQ,CAAC,CAAC;IAChC,OAAO,MAAM,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;AAAA,CAChD;AAED,MAAM,UAAU,MAAM,CACrB,KAAkB,EAClB,OAAgB,EAChB,OAA6B,EACC;IAC9B,+DAA+D;IAC/D,IAAI,KAAK,CAAC,GAAG,KAAK,eAAe,EAAE,CAAC;QACnC,OAAO,kBAAkB,CAAC,KAA+B,EAAE,OAAO,EAAE,OAA8B,CAAC,CAAC;IACrG,CAAC;IAED,MAAM,MAAM,GAAG,OAAO,EAAE,MAAM,IAAI,YAAY,CAAC,KAAK,CAAC,QAAQ,CAAC,CAAC;IAC/D,IAAI,CAAC,MAAM,EAAE,CAAC;QACb,MAAM,IAAI,KAAK,CAAC,4BAA4B,KAAK,CAAC,QAAQ,EAAE,CAAC,CAAC;IAC/D,CAAC;IACD,MAAM,eAAe,GAAG,EAAE,GAAG,OAAO,EAAE,MAAM,EAAE,CAAC;IAE/C,MAAM,GAAG,GAAQ,KAAK,CAAC,GAAG,CAAC;IAC3B,QAAQ,GAAG,EAAE,CAAC;QACb,KAAK,oBAAoB;YACxB,OAAO,eAAe,CAAC,KAAoC,EAAE,OAAO,EAAE,eAAe,CAAC,CAAC;QAExF,KAAK,oBAAoB;YACxB,OAAO,uBAAuB,CAAC,KAAoC,EAAE,OAAO,EAAE,eAAsB,CAAC,CAAC;QAEvG,KAAK,kBAAkB;YACtB,OAAO,qBAAqB,CAAC,KAAkC,EAAE,OAAO,EAAE,eAAsB,CAAC,CAAC;QAEnG,KAAK,wBAAwB;YAC5B,OAAO,0BAA0B,CAAC,KAAwC,EAAE,OAAO,EAAE,eAAsB,CAAC,CAAC;QAE9G,KAAK,sBAAsB;YAC1B,OAAO,YAAY,CAAC,KAAsC,EAAE,OAAO,EAAE,eAAe,CAAC,CAAC;QAEvF,KAAK,mBAAmB;YACvB,OAAO,qBAAqB,CAC3B,KAAmC,EACnC,OAAO,EACP,eAAyC,CACzC,CAAC;QAEH,SAAS,CAAC;YACT,4DAA4D;YAC5D,MAAM,WAAW,GAAU,GAAG,CAAC;YAC/B,MAAM,IAAI,KAAK,CAAC,kBAAkB,WAAW,EAAE,CAAC,CAAC;QAClD,CAAC;IACF,CAAC;AAAA,CACD;AAED,MAAM,CAAC,KAAK,UAAU,QAAQ,CAC7B,KAAkB,EAClB,OAAgB,EAChB,OAA6B,EACD;IAC5B,MAAM,CAAC,GAAG,MAAM,CAAC,KAAK,EAAE,OAAO,EAAE,OAAO,CAAC,CAAC;IAC1C,OAAO,CAAC,CAAC,MAAM,EAAE,CAAC;AAAA,CAClB;AAED,MAAM,UAAU,YAAY,CAC3B,KAAkB,EAClB,OAAgB,EAChB,OAA6B,EACC;IAC9B,+DAA+D;IAC/D,IAAI,KAAK,CAAC,GAAG,KAAK,eAAe,EAAE,CAAC;QACnC,MAAM,eAAe,GAAG,gBAAgB,CAAC,KAAK,EAAE,OAAO,EAAE,SAAS,CAAC,CAAC;QACpE,OAAO,MAAM,CAAC,KAAK,EAAE,OAAO,EAAE,eAAe,CAAC,CAAC;IAChD,CAAC;IAED,MAAM,MAAM,GAAG,OAAO,EAAE,MAAM,IAAI,YAAY,CAAC,KAAK,CAAC,QAAQ,CAAC,CAAC;IAC/D,IAAI,CAAC,MAAM,EAAE,CAAC;QACb,MAAM,IAAI,KAAK,CAAC,4BAA4B,KAAK,CAAC,QAAQ,EAAE,CAAC,CAAC;IAC/D,CAAC;IAED,MAAM,eAAe,GAAG,gBAAgB,CAAC,KAAK,EAAE,OAAO,EAAE,MAAM,CAAC,CAAC;IACjE,OAAO,MAAM,CAAC,KAAK,EAAE,OAAO,EAAE,eAAe,CAAC,CAAC;AAAA,CAC/C;AAED,MAAM,CAAC,KAAK,UAAU,cAAc,CACnC,KAAkB,EAClB,OAAgB,EAChB,OAA6B,EACD;IAC5B,MAAM,CAAC,GAAG,YAAY,CAAC,KAAK,EAAE,OAAO,EAAE,OAAO,CAAC,CAAC;IAChD,OAAO,CAAC,CAAC,MAAM,EAAE,CAAC;AAAA,CAClB;AAED,SAAS,gBAAgB,CACxB,KAAkB,EAClB,OAA6B,EAC7B,MAAe,EACO;IACtB,MAAM,IAAI,GAAG;QACZ,WAAW,EAAE,OAAO,EAAE,WAAW;QACjC,SAAS,EAAE,OAAO,EAAE,SAAS,IAAI,IAAI,CAAC,GAAG,CAAC,KAAK,CAAC,SAAS,EAAE,KAAK,CAAC;QACjE,MAAM,EAAE,OAAO,EAAE,MAAM;QACvB,MAAM,EAAE,MAAM,IAAI,OAAO,EAAE,MAAM;QACjC,SAAS,EAAE,OAAO,EAAE,SAAS;KAC7B,CAAC;IAEF,oEAAoE;IACpE,MAAM,cAAc,GAAG,CAAC,MAAiC,EAAE,EAAE,CAAC,CAAC,MAAM,KAAK,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC;IAErG,QAAQ,KAAK,CAAC,GAAG,EAAE,CAAC;QACnB,KAAK,oBAAoB,EAAE,CAAC;YAC3B,8DAA8D;YAC9D,IAAI,CAAC,OAAO,EAAE,SAAS,EAAE,CAAC;gBACzB,OAAO,EAAE,GAAG,IAAI,EAAE,eAAe,EAAE,KAAK,EAA6B,CAAC;YACvE,CAAC;YAED,sDAAsD;YACtD,uEAAuE;YACvE,MAAM,cAAc,GAAoB;gBACvC,OAAO,EAAE,IAAI;gBACb,GAAG,EAAE,IAAI;gBACT,MAAM,EAAE,IAAI;gBACZ,IAAI,EAAE,KAAK;aACX,CAAC;YACF,MAAM,OAAO,GAAG,EAAE,GAAG,cAAc,EAAE,GAAG,OAAO,EAAE,eAAe,EAAE,CAAC;YAEnE,MAAM,eAAe,GAAG,IAAI,CAAC;YAC7B,MAAM,KAAK,GAAG,cAAc,CAAC,OAAO,CAAC,SAAS,CAAE,CAAC;YACjD,IAAI,cAAc,GAAG,OAAO,CAAC,KAAK,CAAE,CAAC;YACrC,8FAA8F;YAC9F,MAAM,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,IAAI,CAAC,SAAS,IAAI,CAAC,CAAC,GAAG,cAAc,EAAE,KAAK,CAAC,SAAS,CAAC,CAAC;YAEpF,mEAAmE;YACnE,IAAI,SAAS,IAAI,cAAc,EAAE,CAAC;gBACjC,cAAc,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,SAAS,GAAG,eAAe,CAAC,CAAC;YAC3D,CAAC;YAED,OAAO;gBACN,GAAG,IAAI;gBACP,SAAS;gBACT,eAAe,EAAE,IAAI;gBACrB,oBAAoB,EAAE,cAAc;aACT,CAAC;QAC9B,CAAC;QAED,KAAK,oBAAoB;YACxB,OAAO;gBACN,GAAG,IAAI;gBACP,eAAe,EAAE,aAAa,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,OAAO,EAAE,SAAS,CAAC,CAAC,CAAC,cAAc,CAAC,OAAO,EAAE,SAAS,CAAC;aAC5D,CAAC;QAEtC,KAAK,kBAAkB;YACtB,OAAO;gBACN,GAAG,IAAI;gBACP,eAAe,EAAE,aAAa,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,OAAO,EAAE,SAAS,CAAC,CAAC,CAAC,cAAc,CAAC,OAAO,EAAE,SAAS,CAAC;aAC9D,CAAC;QAEpC,KAAK,wBAAwB;YAC5B,OAAO;gBACN,GAAG,IAAI;gBACP,eAAe,EAAE,aAAa,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,OAAO,EAAE,SAAS,CAAC,CAAC,CAAC,cAAc,CAAC,OAAO,EAAE,SAAS,CAAC;aACzD,CAAC;QAEzC,KAAK,sBAAsB,EAAE,CAAC;YAC7B,8DAA8D;YAC9D,0EAA0E;YAC1E,IAAI,CAAC,OAAO,EAAE,SAAS,EAAE,CAAC;gBACzB,OAAO,EAAE,GAAG,IAAI,EAAE,QAAQ,EAAE,EAAE,OAAO,EAAE,KAAK,EAAE,EAA0B,CAAC;YAC1E,CAAC;YAED,MAAM,WAAW,GAAG,KAAsC,CAAC;YAC3D,MAAM,MAAM,GAAG,cAAc,CAAC,OAAO,CAAC,SAAS,CAAE,CAAC;YAElD,2EAA2E;YAC3E,4DAA4D;YAC5D,IAAI,iBAAiB,CAAC,WAAW,CAAC,IAAI,mBAAmB,CAAC,WAAW,CAAC,EAAE,CAAC;gBACxE,OAAO;oBACN,GAAG,IAAI;oBACP,QAAQ,EAAE;wBACT,OAAO,EAAE,IAAI;wBACb,KAAK,EAAE,uBAAuB,CAAC,MAAM,EAAE,WAAW,CAAC;qBACnD;iBACuB,CAAC;YAC3B,CAAC;YAED,OAAO;gBACN,GAAG,IAAI;gBACP,QAAQ,EAAE;oBACT,OAAO,EAAE,IAAI;oBACb,YAAY,EAAE,eAAe,CAAC,WAAW,EAAE,MAAM,EAAE,OAAO,EAAE,eAAe,CAAC;iBAC5E;aACuB,CAAC;QAC3B,CAAC;QAED,KAAK,mBAAmB,EAAE,CAAC;YAC1B,IAAI,CAAC,OAAO,EAAE,SAAS,EAAE,CAAC;gBACzB,OAAO,EAAE,GAAG,IAAI,EAAE,QAAQ,EAAE,EAAE,OAAO,EAAE,KAAK,EAAE,EAAmC,CAAC;YACnF,CAAC;YAED,MAAM,MAAM,GAAG,cAAc,CAAC,OAAO,CAAC,SAAS,CAAE,CAAC;YAElD,8DAA8D;YAC9D,IAAI,KAAK,CAAC,EAAE,CAAC,QAAQ,CAAC,OAAO,CAAC,IAAI,KAAK,CAAC,EAAE,CAAC,QAAQ,CAAC,SAAS,CAAC,EAAE,CAAC;gBAChE,OAAO;oBACN,GAAG,IAAI;oBACP,QAAQ,EAAE;wBACT,OAAO,EAAE,IAAI;wBACb,KAAK,EAAE,yBAAyB,CAAC,MAAM,EAAE,KAAK,CAAC,EAAE,CAAC;qBAClD;iBACgC,CAAC;YACpC,CAAC;YAED,mEAAmE;YACnE,sDAAsD;YACtD,uEAAuE;YACvE,MAAM,cAAc,GAAoB;gBACvC,OAAO,EAAE,IAAI;gBACb,GAAG,EAAE,IAAI;gBACT,MAAM,EAAE,IAAI;gBACZ,IAAI,EAAE,KAAK;aACX,CAAC;YACF,MAAM,OAAO,GAAG,EAAE,GAAG,cAAc,EAAE,GAAG,OAAO,EAAE,eAAe,EAAE,CAAC;YAEnE,MAAM,eAAe,GAAG,IAAI,CAAC;YAC7B,IAAI,cAAc,GAAG,OAAO,CAAC,MAAM,CAAE,CAAC;YACtC,8FAA8F;YAC9F,MAAM,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,IAAI,CAAC,SAAS,IAAI,CAAC,CAAC,GAAG,cAAc,EAAE,KAAK,CAAC,SAAS,CAAC,CAAC;YAEpF,mEAAmE;YACnE,IAAI,SAAS,IAAI,cAAc,EAAE,CAAC;gBACjC,cAAc,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,SAAS,GAAG,eAAe,CAAC,CAAC;YAC3D,CAAC;YAED,OAAO;gBACN,GAAG,IAAI;gBACP,SAAS;gBACT,QAAQ,EAAE;oBACT,OAAO,EAAE,IAAI;oBACb,YAAY,EAAE,cAAc;iBAC5B;aACgC,CAAC;QACpC,CAAC;QAED,KAAK,eAAe,EAAE,CAAC;YACtB,8DAA8D;YAC9D,IAAI,CAAC,OAAO,EAAE,SAAS,EAAE,CAAC;gBACzB,OAAO,EAAE,GAAG,IAAI,EAAE,QAAQ,EAAE,EAAE,OAAO,EAAE,KAAK,EAAE,EAAgC,CAAC;YAChF,CAAC;YAED,MAAM,WAAW,GAAG,KAA+B,CAAC;YACpD,MAAM,MAAM,GAAG,cAAc,CAAC,OAAO,CAAC,SAAS,CAAE,CAAC;YAClD,MAAM,WAAW,GAAG,WAAuD,CAAC;YAE5E,IAAI,iBAAiB,CAAC,WAAW,CAAC,IAAI,mBAAmB,CAAC,WAAW,CAAC,EAAE,CAAC;gBACxE,OAAO;oBACN,GAAG,IAAI;oBACP,QAAQ,EAAE;wBACT,OAAO,EAAE,IAAI;wBACb,KAAK,EAAE,uBAAuB,CAAC,MAAM,EAAE,WAAW,CAAC;qBACnD;iBAC6B,CAAC;YACjC,CAAC;YAED,OAAO;gBACN,GAAG,IAAI;gBACP,QAAQ,EAAE;oBACT,OAAO,EAAE,IAAI;oBACb,YAAY,EAAE,eAAe,CAAC,WAAW,EAAE,MAAM,EAAE,OAAO,EAAE,eAAe,CAAC;iBAC5E;aAC6B,CAAC;QACjC,CAAC;QAED,SAAS,CAAC;YACT,uBAAuB;YACvB,MAAM,WAAW,GAAU,KAAK,CAAC,GAAG,CAAC;YACrC,MAAM,IAAI,KAAK,CAAC,sCAAsC,WAAW,EAAE,CAAC,CAAC;QACtE,CAAC;IACF,CAAC;AAAA,CACD;AAID,SAAS,iBAAiB,CAAC,KAAoC,EAAW;IACzE,2FAA2F;IAC3F,OAAO,KAAK,CAAC,EAAE,CAAC,QAAQ,CAAC,OAAO,CAAC,CAAC;AAAA,CAClC;AAED,SAAS,mBAAmB,CAAC,KAAoC,EAAW;IAC3E,+FAA+F;IAC/F,OAAO,KAAK,CAAC,EAAE,CAAC,QAAQ,CAAC,SAAS,CAAC,CAAC;AAAA,CACpC;AAED,SAAS,uBAAuB,CAC/B,MAA4B,EAC5B,KAAoC,EACd;IACtB,IAAI,iBAAiB,CAAC,KAAK,CAAC,EAAE,CAAC;QAC9B,gDAAgD;QAChD,QAAQ,MAAM,EAAE,CAAC;YAChB,KAAK,SAAS,CAAC;YACf,KAAK,KAAK;gBACT,OAAO,KAAK,CAAC;YACd,KAAK,QAAQ,CAAC;YACd,KAAK,MAAM;gBACV,OAAO,MAAM,CAAC;QAChB,CAAC;IACF,CAAC;IACD,0CAA0C;IAC1C,QAAQ,MAAM,EAAE,CAAC;QAChB,KAAK,SAAS;YACb,OAAO,SAAS,CAAC;QAClB,KAAK,KAAK;YACT,OAAO,KAAK,CAAC;QACd,KAAK,QAAQ;YACZ,OAAO,QAAQ,CAAC;QACjB,KAAK,MAAM;YACV,OAAO,MAAM,CAAC;IAChB,CAAC;AAAA,CACD;AAED,SAAS,yBAAyB,CAAC,MAA4B,EAAE,OAAe,EAAuB;IACtG,IAAI,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAC,EAAE,CAAC;QAC/B,gDAAgD;QAChD,QAAQ,MAAM,EAAE,CAAC;YAChB,KAAK,SAAS,CAAC;YACf,KAAK,KAAK;gBACT,OAAO,KAAK,CAAC;YACd,KAAK,QAAQ,CAAC;YACd,KAAK,MAAM;gBACV,OAAO,MAAM,CAAC;QAChB,CAAC;IACF,CAAC;IACD,0CAA0C;IAC1C,QAAQ,MAAM,EAAE,CAAC;QAChB,KAAK,SAAS;YACb,OAAO,SAAS,CAAC;QAClB,KAAK,KAAK;YACT,OAAO,KAAK,CAAC;QACd,KAAK,QAAQ;YACZ,OAAO,QAAQ,CAAC;QACjB,KAAK,MAAM;YACV,OAAO,MAAM,CAAC;IAChB,CAAC;AAAA,CACD;AAED,SAAS,eAAe,CACvB,KAAoC,EACpC,MAA4B,EAC5B,aAA+B,EACtB;IACT,4DAA4D;IAC5D,IAAI,aAAa,EAAE,CAAC,MAAM,CAAC,KAAK,SAAS,EAAE,CAAC;QAC3C,OAAO,aAAa,CAAC,MAAM,CAAE,CAAC;IAC/B,CAAC;IAED,gEAAgE;IAChE,IAAI,KAAK,CAAC,EAAE,CAAC,QAAQ,CAAC,SAAS,CAAC,EAAE,CAAC;QAClC,MAAM,OAAO,GAAyC;YACrD,OAAO,EAAE,GAAG;YACZ,GAAG,EAAE,IAAI;YACT,MAAM,EAAE,IAAI;YACZ,IAAI,EAAE,KAAK;SACX,CAAC;QACF,OAAO,OAAO,CAAC,MAAM,CAAC,CAAC;IACxB,CAAC;IAED,IAAI,KAAK,CAAC,EAAE,CAAC,QAAQ,CAAC,WAAW,CAAC,EAAE,CAAC;QACpC,gCAAgC;QAChC,MAAM,OAAO,GAAyC;YACrD,OAAO,EAAE,GAAG;YACZ,GAAG,EAAE,IAAI;YACT,MAAM,EAAE,IAAI;YACZ,IAAI,EAAE,KAAK;SACX,CAAC;QACF,OAAO,OAAO,CAAC,MAAM,CAAC,CAAC;IACxB,CAAC;IAED,8BAA8B;IAC9B,OAAO,CAAC,CAAC,CAAC;AAAA,CACV","sourcesContent":["import { existsSync } from \"node:fs\";\nimport { homedir } from \"node:os\";\nimport { join } from \"node:path\";\nimport { supportsXhigh } from \"./models.js\";\nimport { type AnthropicOptions, streamAnthropic } from \"./providers/anthropic.js\";\nimport { type GoogleOptions, streamGoogle } from \"./providers/google.js\";\nimport {\n\ttype GoogleGeminiCliOptions,\n\ttype GoogleThinkingLevel,\n\tstreamGoogleGeminiCli,\n} from \"./providers/google-gemini-cli.js\";\nimport { type GoogleVertexOptions, streamGoogleVertex } from \"./providers/google-vertex.js\";\nimport { type OpenAICodexResponsesOptions, streamOpenAICodexResponses } from \"./providers/openai-codex-responses.js\";\nimport { type OpenAICompletionsOptions, streamOpenAICompletions } from \"./providers/openai-completions.js\";\nimport { type OpenAIResponsesOptions, streamOpenAIResponses } from \"./providers/openai-responses.js\";\nimport type {\n\tApi,\n\tAssistantMessage,\n\tAssistantMessageEventStream,\n\tContext,\n\tKnownProvider,\n\tModel,\n\tOptionsForApi,\n\tSimpleStreamOptions,\n\tThinkingBudgets,\n\tThinkingLevel,\n} from \"./types.js\";\n\nconst VERTEX_ADC_CREDENTIALS_PATH = join(homedir(), \".config\", \"gcloud\", \"application_default_credentials.json\");\n\nlet cachedVertexAdcCredentialsExists: boolean | null = null;\n\nfunction hasVertexAdcCredentials(): boolean {\n\tif (cachedVertexAdcCredentialsExists === null) {\n\t\tcachedVertexAdcCredentialsExists = existsSync(VERTEX_ADC_CREDENTIALS_PATH);\n\t}\n\treturn cachedVertexAdcCredentialsExists;\n}\n\n/**\n * Get API key for provider from known environment variables, e.g. OPENAI_API_KEY.\n *\n * Will not return API keys for providers that require OAuth tokens.\n */\nexport function getEnvApiKey(provider: KnownProvider): string | undefined;\nexport function getEnvApiKey(provider: string): string | undefined;\nexport function getEnvApiKey(provider: any): string | undefined {\n\t// Fall back to environment variables\n\tif (provider === \"github-copilot\") {\n\t\treturn process.env.COPILOT_GITHUB_TOKEN || process.env.GH_TOKEN || process.env.GITHUB_TOKEN;\n\t}\n\n\t// ANTHROPIC_OAUTH_TOKEN takes precedence over ANTHROPIC_API_KEY\n\tif (provider === \"anthropic\") {\n\t\treturn process.env.ANTHROPIC_OAUTH_TOKEN || process.env.ANTHROPIC_API_KEY;\n\t}\n\n\t// Vertex AI uses Application Default Credentials, not API keys.\n\t// Auth is configured via `gcloud auth application-default login`.\n\tif (provider === \"google-vertex\") {\n\t\tconst hasCredentials = hasVertexAdcCredentials();\n\t\tconst hasProject = !!(process.env.GOOGLE_CLOUD_PROJECT || process.env.GCLOUD_PROJECT);\n\t\tconst hasLocation = !!process.env.GOOGLE_CLOUD_LOCATION;\n\n\t\tif (hasCredentials && hasProject && hasLocation) {\n\t\t\treturn \"<authenticated>\";\n\t\t}\n\t}\n\n\tconst envMap: Record<string, string> = {\n\t\topenai: \"OPENAI_API_KEY\",\n\t\tgoogle: \"GEMINI_API_KEY\",\n\t\tgroq: \"GROQ_API_KEY\",\n\t\tcerebras: \"CEREBRAS_API_KEY\",\n\t\txai: \"XAI_API_KEY\",\n\t\topenrouter: \"OPENROUTER_API_KEY\",\n\t\tzai: \"ZAI_API_KEY\",\n\t\tmistral: \"MISTRAL_API_KEY\",\n\t};\n\n\tconst envVar = envMap[provider];\n\treturn envVar ? process.env[envVar] : undefined;\n}\n\nexport function stream<TApi extends Api>(\n\tmodel: Model<TApi>,\n\tcontext: Context,\n\toptions?: OptionsForApi<TApi>,\n): AssistantMessageEventStream {\n\t// Vertex AI uses Application Default Credentials, not API keys\n\tif (model.api === \"google-vertex\") {\n\t\treturn streamGoogleVertex(model as Model<\"google-vertex\">, context, options as GoogleVertexOptions);\n\t}\n\n\tconst apiKey = options?.apiKey || getEnvApiKey(model.provider);\n\tif (!apiKey) {\n\t\tthrow new Error(`No API key for provider: ${model.provider}`);\n\t}\n\tconst providerOptions = { ...options, apiKey };\n\n\tconst api: Api = model.api;\n\tswitch (api) {\n\t\tcase \"anthropic-messages\":\n\t\t\treturn streamAnthropic(model as Model<\"anthropic-messages\">, context, providerOptions);\n\n\t\tcase \"openai-completions\":\n\t\t\treturn streamOpenAICompletions(model as Model<\"openai-completions\">, context, providerOptions as any);\n\n\t\tcase \"openai-responses\":\n\t\t\treturn streamOpenAIResponses(model as Model<\"openai-responses\">, context, providerOptions as any);\n\n\t\tcase \"openai-codex-responses\":\n\t\t\treturn streamOpenAICodexResponses(model as Model<\"openai-codex-responses\">, context, providerOptions as any);\n\n\t\tcase \"google-generative-ai\":\n\t\t\treturn streamGoogle(model as Model<\"google-generative-ai\">, context, providerOptions);\n\n\t\tcase \"google-gemini-cli\":\n\t\t\treturn streamGoogleGeminiCli(\n\t\t\t\tmodel as Model<\"google-gemini-cli\">,\n\t\t\t\tcontext,\n\t\t\t\tproviderOptions as GoogleGeminiCliOptions,\n\t\t\t);\n\n\t\tdefault: {\n\t\t\t// This should never be reached if all Api cases are handled\n\t\t\tconst _exhaustive: never = api;\n\t\t\tthrow new Error(`Unhandled API: ${_exhaustive}`);\n\t\t}\n\t}\n}\n\nexport async function complete<TApi extends Api>(\n\tmodel: Model<TApi>,\n\tcontext: Context,\n\toptions?: OptionsForApi<TApi>,\n): Promise<AssistantMessage> {\n\tconst s = stream(model, context, options);\n\treturn s.result();\n}\n\nexport function streamSimple<TApi extends Api>(\n\tmodel: Model<TApi>,\n\tcontext: Context,\n\toptions?: SimpleStreamOptions,\n): AssistantMessageEventStream {\n\t// Vertex AI uses Application Default Credentials, not API keys\n\tif (model.api === \"google-vertex\") {\n\t\tconst providerOptions = mapOptionsForApi(model, options, undefined);\n\t\treturn stream(model, context, providerOptions);\n\t}\n\n\tconst apiKey = options?.apiKey || getEnvApiKey(model.provider);\n\tif (!apiKey) {\n\t\tthrow new Error(`No API key for provider: ${model.provider}`);\n\t}\n\n\tconst providerOptions = mapOptionsForApi(model, options, apiKey);\n\treturn stream(model, context, providerOptions);\n}\n\nexport async function completeSimple<TApi extends Api>(\n\tmodel: Model<TApi>,\n\tcontext: Context,\n\toptions?: SimpleStreamOptions,\n): Promise<AssistantMessage> {\n\tconst s = streamSimple(model, context, options);\n\treturn s.result();\n}\n\nfunction mapOptionsForApi<TApi extends Api>(\n\tmodel: Model<TApi>,\n\toptions?: SimpleStreamOptions,\n\tapiKey?: string,\n): OptionsForApi<TApi> {\n\tconst base = {\n\t\ttemperature: options?.temperature,\n\t\tmaxTokens: options?.maxTokens || Math.min(model.maxTokens, 32000),\n\t\tsignal: options?.signal,\n\t\tapiKey: apiKey || options?.apiKey,\n\t\tsessionId: options?.sessionId,\n\t};\n\n\t// Helper to clamp xhigh to high for providers that don't support it\n\tconst clampReasoning = (effort: ThinkingLevel | undefined) => (effort === \"xhigh\" ? \"high\" : effort);\n\n\tswitch (model.api) {\n\t\tcase \"anthropic-messages\": {\n\t\t\t// Explicitly disable thinking when reasoning is not specified\n\t\t\tif (!options?.reasoning) {\n\t\t\t\treturn { ...base, thinkingEnabled: false } satisfies AnthropicOptions;\n\t\t\t}\n\n\t\t\t// Claude requires max_tokens > thinking.budget_tokens\n\t\t\t// So we need to ensure maxTokens accounts for both thinking and output\n\t\t\tconst defaultBudgets: ThinkingBudgets = {\n\t\t\t\tminimal: 1024,\n\t\t\t\tlow: 2048,\n\t\t\t\tmedium: 8192,\n\t\t\t\thigh: 16384,\n\t\t\t};\n\t\t\tconst budgets = { ...defaultBudgets, ...options?.thinkingBudgets };\n\n\t\t\tconst minOutputTokens = 1024;\n\t\t\tconst level = clampReasoning(options.reasoning)!;\n\t\t\tlet thinkingBudget = budgets[level]!;\n\t\t\t// Caller's maxTokens is the desired output; add thinking budget on top, capped at model limit\n\t\t\tconst maxTokens = Math.min((base.maxTokens || 0) + thinkingBudget, model.maxTokens);\n\n\t\t\t// If not enough room for thinking + output, reduce thinking budget\n\t\t\tif (maxTokens <= thinkingBudget) {\n\t\t\t\tthinkingBudget = Math.max(0, maxTokens - minOutputTokens);\n\t\t\t}\n\n\t\t\treturn {\n\t\t\t\t...base,\n\t\t\t\tmaxTokens,\n\t\t\t\tthinkingEnabled: true,\n\t\t\t\tthinkingBudgetTokens: thinkingBudget,\n\t\t\t} satisfies AnthropicOptions;\n\t\t}\n\n\t\tcase \"openai-completions\":\n\t\t\treturn {\n\t\t\t\t...base,\n\t\t\t\treasoningEffort: supportsXhigh(model) ? options?.reasoning : clampReasoning(options?.reasoning),\n\t\t\t} satisfies OpenAICompletionsOptions;\n\n\t\tcase \"openai-responses\":\n\t\t\treturn {\n\t\t\t\t...base,\n\t\t\t\treasoningEffort: supportsXhigh(model) ? options?.reasoning : clampReasoning(options?.reasoning),\n\t\t\t} satisfies OpenAIResponsesOptions;\n\n\t\tcase \"openai-codex-responses\":\n\t\t\treturn {\n\t\t\t\t...base,\n\t\t\t\treasoningEffort: supportsXhigh(model) ? options?.reasoning : clampReasoning(options?.reasoning),\n\t\t\t} satisfies OpenAICodexResponsesOptions;\n\n\t\tcase \"google-generative-ai\": {\n\t\t\t// Explicitly disable thinking when reasoning is not specified\n\t\t\t// This is needed because Gemini has \"dynamic thinking\" enabled by default\n\t\t\tif (!options?.reasoning) {\n\t\t\t\treturn { ...base, thinking: { enabled: false } } satisfies GoogleOptions;\n\t\t\t}\n\n\t\t\tconst googleModel = model as Model<\"google-generative-ai\">;\n\t\t\tconst effort = clampReasoning(options.reasoning)!;\n\n\t\t\t// Gemini 3 models use thinkingLevel exclusively instead of thinkingBudget.\n\t\t\t// https://ai.google.dev/gemini-api/docs/thinking#set-budget\n\t\t\tif (isGemini3ProModel(googleModel) || isGemini3FlashModel(googleModel)) {\n\t\t\t\treturn {\n\t\t\t\t\t...base,\n\t\t\t\t\tthinking: {\n\t\t\t\t\t\tenabled: true,\n\t\t\t\t\t\tlevel: getGemini3ThinkingLevel(effort, googleModel),\n\t\t\t\t\t},\n\t\t\t\t} satisfies GoogleOptions;\n\t\t\t}\n\n\t\t\treturn {\n\t\t\t\t...base,\n\t\t\t\tthinking: {\n\t\t\t\t\tenabled: true,\n\t\t\t\t\tbudgetTokens: getGoogleBudget(googleModel, effort, options?.thinkingBudgets),\n\t\t\t\t},\n\t\t\t} satisfies GoogleOptions;\n\t\t}\n\n\t\tcase \"google-gemini-cli\": {\n\t\t\tif (!options?.reasoning) {\n\t\t\t\treturn { ...base, thinking: { enabled: false } } satisfies GoogleGeminiCliOptions;\n\t\t\t}\n\n\t\t\tconst effort = clampReasoning(options.reasoning)!;\n\n\t\t\t// Gemini 3 models use thinkingLevel instead of thinkingBudget\n\t\t\tif (model.id.includes(\"3-pro\") || model.id.includes(\"3-flash\")) {\n\t\t\t\treturn {\n\t\t\t\t\t...base,\n\t\t\t\t\tthinking: {\n\t\t\t\t\t\tenabled: true,\n\t\t\t\t\t\tlevel: getGeminiCliThinkingLevel(effort, model.id),\n\t\t\t\t\t},\n\t\t\t\t} satisfies GoogleGeminiCliOptions;\n\t\t\t}\n\n\t\t\t// Models using thinkingBudget (Gemini 2.x, Claude via Antigravity)\n\t\t\t// Claude requires max_tokens > thinking.budget_tokens\n\t\t\t// So we need to ensure maxTokens accounts for both thinking and output\n\t\t\tconst defaultBudgets: ThinkingBudgets = {\n\t\t\t\tminimal: 1024,\n\t\t\t\tlow: 2048,\n\t\t\t\tmedium: 8192,\n\t\t\t\thigh: 16384,\n\t\t\t};\n\t\t\tconst budgets = { ...defaultBudgets, ...options?.thinkingBudgets };\n\n\t\t\tconst minOutputTokens = 1024;\n\t\t\tlet thinkingBudget = budgets[effort]!;\n\t\t\t// Caller's maxTokens is the desired output; add thinking budget on top, capped at model limit\n\t\t\tconst maxTokens = Math.min((base.maxTokens || 0) + thinkingBudget, model.maxTokens);\n\n\t\t\t// If not enough room for thinking + output, reduce thinking budget\n\t\t\tif (maxTokens <= thinkingBudget) {\n\t\t\t\tthinkingBudget = Math.max(0, maxTokens - minOutputTokens);\n\t\t\t}\n\n\t\t\treturn {\n\t\t\t\t...base,\n\t\t\t\tmaxTokens,\n\t\t\t\tthinking: {\n\t\t\t\t\tenabled: true,\n\t\t\t\t\tbudgetTokens: thinkingBudget,\n\t\t\t\t},\n\t\t\t} satisfies GoogleGeminiCliOptions;\n\t\t}\n\n\t\tcase \"google-vertex\": {\n\t\t\t// Explicitly disable thinking when reasoning is not specified\n\t\t\tif (!options?.reasoning) {\n\t\t\t\treturn { ...base, thinking: { enabled: false } } satisfies GoogleVertexOptions;\n\t\t\t}\n\n\t\t\tconst vertexModel = model as Model<\"google-vertex\">;\n\t\t\tconst effort = clampReasoning(options.reasoning)!;\n\t\t\tconst geminiModel = vertexModel as unknown as Model<\"google-generative-ai\">;\n\n\t\t\tif (isGemini3ProModel(geminiModel) || isGemini3FlashModel(geminiModel)) {\n\t\t\t\treturn {\n\t\t\t\t\t...base,\n\t\t\t\t\tthinking: {\n\t\t\t\t\t\tenabled: true,\n\t\t\t\t\t\tlevel: getGemini3ThinkingLevel(effort, geminiModel),\n\t\t\t\t\t},\n\t\t\t\t} satisfies GoogleVertexOptions;\n\t\t\t}\n\n\t\t\treturn {\n\t\t\t\t...base,\n\t\t\t\tthinking: {\n\t\t\t\t\tenabled: true,\n\t\t\t\t\tbudgetTokens: getGoogleBudget(geminiModel, effort, options?.thinkingBudgets),\n\t\t\t\t},\n\t\t\t} satisfies GoogleVertexOptions;\n\t\t}\n\n\t\tdefault: {\n\t\t\t// Exhaustiveness check\n\t\t\tconst _exhaustive: never = model.api;\n\t\t\tthrow new Error(`Unhandled API in mapOptionsForApi: ${_exhaustive}`);\n\t\t}\n\t}\n}\n\ntype ClampedThinkingLevel = Exclude<ThinkingLevel, \"xhigh\">;\n\nfunction isGemini3ProModel(model: Model<\"google-generative-ai\">): boolean {\n\t// Covers gemini-3-pro, gemini-3-pro-preview, and possible other prefixed ids in the future\n\treturn model.id.includes(\"3-pro\");\n}\n\nfunction isGemini3FlashModel(model: Model<\"google-generative-ai\">): boolean {\n\t// Covers gemini-3-flash, gemini-3-flash-preview, and possible other prefixed ids in the future\n\treturn model.id.includes(\"3-flash\");\n}\n\nfunction getGemini3ThinkingLevel(\n\teffort: ClampedThinkingLevel,\n\tmodel: Model<\"google-generative-ai\">,\n): GoogleThinkingLevel {\n\tif (isGemini3ProModel(model)) {\n\t\t// Gemini 3 Pro only supports LOW/HIGH (for now)\n\t\tswitch (effort) {\n\t\t\tcase \"minimal\":\n\t\t\tcase \"low\":\n\t\t\t\treturn \"LOW\";\n\t\t\tcase \"medium\":\n\t\t\tcase \"high\":\n\t\t\t\treturn \"HIGH\";\n\t\t}\n\t}\n\t// Gemini 3 Flash supports all four levels\n\tswitch (effort) {\n\t\tcase \"minimal\":\n\t\t\treturn \"MINIMAL\";\n\t\tcase \"low\":\n\t\t\treturn \"LOW\";\n\t\tcase \"medium\":\n\t\t\treturn \"MEDIUM\";\n\t\tcase \"high\":\n\t\t\treturn \"HIGH\";\n\t}\n}\n\nfunction getGeminiCliThinkingLevel(effort: ClampedThinkingLevel, modelId: string): GoogleThinkingLevel {\n\tif (modelId.includes(\"3-pro\")) {\n\t\t// Gemini 3 Pro only supports LOW/HIGH (for now)\n\t\tswitch (effort) {\n\t\t\tcase \"minimal\":\n\t\t\tcase \"low\":\n\t\t\t\treturn \"LOW\";\n\t\t\tcase \"medium\":\n\t\t\tcase \"high\":\n\t\t\t\treturn \"HIGH\";\n\t\t}\n\t}\n\t// Gemini 3 Flash supports all four levels\n\tswitch (effort) {\n\t\tcase \"minimal\":\n\t\t\treturn \"MINIMAL\";\n\t\tcase \"low\":\n\t\t\treturn \"LOW\";\n\t\tcase \"medium\":\n\t\t\treturn \"MEDIUM\";\n\t\tcase \"high\":\n\t\t\treturn \"HIGH\";\n\t}\n}\n\nfunction getGoogleBudget(\n\tmodel: Model<\"google-generative-ai\">,\n\teffort: ClampedThinkingLevel,\n\tcustomBudgets?: ThinkingBudgets,\n): number {\n\t// Custom budgets take precedence if provided for this level\n\tif (customBudgets?.[effort] !== undefined) {\n\t\treturn customBudgets[effort]!;\n\t}\n\n\t// See https://ai.google.dev/gemini-api/docs/thinking#set-budget\n\tif (model.id.includes(\"2.5-pro\")) {\n\t\tconst budgets: Record<ClampedThinkingLevel, number> = {\n\t\t\tminimal: 128,\n\t\t\tlow: 2048,\n\t\t\tmedium: 8192,\n\t\t\thigh: 32768,\n\t\t};\n\t\treturn budgets[effort];\n\t}\n\n\tif (model.id.includes(\"2.5-flash\")) {\n\t\t// Covers 2.5-flash-lite as well\n\t\tconst budgets: Record<ClampedThinkingLevel, number> = {\n\t\t\tminimal: 128,\n\t\t\tlow: 2048,\n\t\t\tmedium: 8192,\n\t\t\thigh: 24576,\n\t\t};\n\t\treturn budgets[effort];\n\t}\n\n\t// Unknown model - use dynamic\n\treturn -1;\n}\n"]}

package/dist/types.d.ts

@@ -21,6 +21,13 @@ export type OptionsForApi<TApi extends Api> = ApiOptionsMap[TApi];
 export type KnownProvider = "anthropic" | "google" | "google-gemini-cli" | "google-antigravity" | "google-vertex" | "openai" | "openai-codex" | "github-copilot" | "xai" | "groq" | "cerebras" | "openrouter" | "zai" | "mistral";
 export type Provider = KnownProvider | string;
 export type ThinkingLevel = "minimal" | "low" | "medium" | "high" | "xhigh";
+/** Token budgets for each thinking level (token-based providers only) */
+export interface ThinkingBudgets {
+    minimal?: number;
+    low?: number;
+    medium?: number;
+    high?: number;
+}
 export interface StreamOptions {
     temperature?: number;
     maxTokens?: number;
@@ -35,6 +42,8 @@ export interface StreamOptions {
 }
 export interface SimpleStreamOptions extends StreamOptions {
     reasoning?: ThinkingLevel;
+    /** Custom token budgets for thinking levels (token-based providers only) */
+    thinkingBudgets?: ThinkingBudgets;
 }
 export type StreamFunction<TApi extends Api> = (model: Model<TApi>, context: Context, options: OptionsForApi<TApi>) => AssistantMessageEventStream;
 export interface TextContent {

package/dist/types.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,0BAA0B,CAAC;AACjE,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,uBAAuB,CAAC;AAC3D,OAAO,KAAK,EAAE,sBAAsB,EAAE,MAAM,kCAAkC,CAAC;AAC/E,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,8BAA8B,CAAC;AACxE,OAAO,KAAK,EAAE,2BAA2B,EAAE,MAAM,uCAAuC,CAAC;AACzF,OAAO,KAAK,EAAE,wBAAwB,EAAE,MAAM,mCAAmC,CAAC;AAClF,OAAO,KAAK,EAAE,sBAAsB,EAAE,MAAM,iCAAiC,CAAC;AAC9E,OAAO,KAAK,EAAE,2BAA2B,EAAE,MAAM,yBAAyB,CAAC;AAE3E,YAAY,EAAE,2BAA2B,EAAE,MAAM,yBAAyB,CAAC;AAE3E,MAAM,MAAM,GAAG,GACZ,oBAAoB,GACpB,kBAAkB,GAClB,wBAAwB,GACxB,oBAAoB,GACpB,sBAAsB,GACtB,mBAAmB,GACnB,eAAe,CAAC;AAEnB,MAAM,WAAW,aAAa;IAC7B,oBAAoB,EAAE,gBAAgB,CAAC;IACvC,oBAAoB,EAAE,wBAAwB,CAAC;IAC/C,kBAAkB,EAAE,sBAAsB,CAAC;IAC3C,wBAAwB,EAAE,2BAA2B,CAAC;IACtD,sBAAsB,EAAE,aAAa,CAAC;IACtC,mBAAmB,EAAE,sBAAsB,CAAC;IAC5C,eAAe,EAAE,mBAAmB,CAAC;CACrC;AAWD,MAAM,MAAM,aAAa,CAAC,IAAI,SAAS,GAAG,IAAI,aAAa,CAAC,IAAI,CAAC,CAAC;AAElE,MAAM,MAAM,aAAa,GACtB,WAAW,GACX,QAAQ,GACR,mBAAmB,GACnB,oBAAoB,GACpB,eAAe,GACf,QAAQ,GACR,cAAc,GACd,gBAAgB,GAChB,KAAK,GACL,MAAM,GACN,UAAU,GACV,YAAY,GACZ,KAAK,GACL,SAAS,CAAC;AACb,MAAM,MAAM,QAAQ,GAAG,aAAa,GAAG,MAAM,CAAC;AAE9C,MAAM,MAAM,aAAa,GAAG,SAAS,GAAG,KAAK,GAAG,QAAQ,GAAG,MAAM,GAAG,OAAO,CAAC;AAG5E,MAAM,WAAW,aAAa;IAC7B,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,MAAM,CAAC,EAAE,WAAW,CAAC;IACrB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB;;;;OAIG;IACH,SAAS,CAAC,EAAE,MAAM,CAAC;CACnB;AAGD,MAAM,WAAW,mBAAoB,SAAQ,aAAa;IACzD,SAAS,CAAC,EAAE,aAAa,CAAC;CAC1B;AAGD,MAAM,MAAM,cAAc,CAAC,IAAI,SAAS,GAAG,IAAI,CAC9C,KAAK,EAAE,KAAK,CAAC,IAAI,CAAC,EAClB,OAAO,EAAE,OAAO,EAChB,OAAO,EAAE,aAAa,CAAC,IAAI,CAAC,KACxB,2BAA2B,CAAC;AAEjC,MAAM,WAAW,WAAW;IAC3B,IAAI,EAAE,MAAM,CAAC;IACb,IAAI,EAAE,MAAM,CAAC;IACb,aAAa,CAAC,EAAE,MAAM,CAAC;CACvB;AAED,MAAM,WAAW,eAAe;IAC/B,IAAI,EAAE,UAAU,CAAC;IACjB,QAAQ,EAAE,MAAM,CAAC;IACjB,iBAAiB,CAAC,EAAE,MAAM,CAAC;CAC3B;AAED,MAAM,WAAW,YAAY;IAC5B,IAAI,EAAE,OAAO,CAAC;IACd,IAAI,EAAE,MAAM,CAAC;IACb,QAAQ,EAAE,MAAM,CAAC;CACjB;AAED,MAAM,WAAW,QAAQ;IACxB,IAAI,EAAE,UAAU,CAAC;IACjB,EAAE,EAAE,MAAM,CAAC;IACX,IAAI,EAAE,MAAM,CAAC;IACb,SAAS,EAAE,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC;IAC/B,gBAAgB,CAAC,EAAE,MAAM,CAAC;CAC1B;AAED,MAAM,WAAW,KAAK;IACrB,KAAK,EAAE,MAAM,CAAC;IACd,MAAM,EAAE,MAAM,CAAC;IACf,SAAS,EAAE,MAAM,CAAC;IAClB,UAAU,EAAE,MAAM,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;IACpB,IAAI,EAAE;QACL,KAAK,EAAE,MAAM,CAAC;QACd,MAAM,EAAE,MAAM,CAAC;QACf,SAAS,EAAE,MAAM,CAAC;QAClB,UAAU,EAAE,MAAM,CAAC;QACnB,KAAK,EAAE,MAAM,CAAC;KACd,CAAC;CACF;AAED,MAAM,MAAM,UAAU,GAAG,MAAM,GAAG,QAAQ,GAAG,SAAS,GAAG,OAAO,GAAG,SAAS,CAAC;AAE7E,MAAM,WAAW,WAAW;IAC3B,IAAI,EAAE,MAAM,CAAC;IACb,OAAO,EAAE,MAAM,GAAG,CAAC,WAAW,GAAG,YAAY,CAAC,EAAE,CAAC;IACjD,SAAS,EAAE,MAAM,CAAC;CAClB;AAED,MAAM,WAAW,gBAAgB;IAChC,IAAI,EAAE,WAAW,CAAC;IAClB,OAAO,EAAE,CAAC,WAAW,GAAG,eAAe,GAAG,QAAQ,CAAC,EAAE,CAAC;IACtD,GAAG,EAAE,GAAG,CAAC;IACT,QAAQ,EAAE,QAAQ,CAAC;IACnB,KAAK,EAAE,MAAM,CAAC;IACd,KAAK,EAAE,KAAK,CAAC;IACb,UAAU,EAAE,UAAU,CAAC;IACvB,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,SAAS,EAAE,MAAM,CAAC;CAClB;AAED,MAAM,WAAW,iBAAiB,CAAC,QAAQ,GAAG,GAAG;IAChD,IAAI,EAAE,YAAY,CAAC;IACnB,UAAU,EAAE,MAAM,CAAC;IACnB,QAAQ,EAAE,MAAM,CAAC;IACjB,OAAO,EAAE,CAAC,WAAW,GAAG,YAAY,CAAC,EAAE,CAAC;IACxC,OAAO,CAAC,EAAE,QAAQ,CAAC;IACnB,OAAO,EAAE,OAAO,CAAC;IACjB,SAAS,EAAE,MAAM,CAAC;CAClB;AAED,MAAM,MAAM,OAAO,GAAG,WAAW,GAAG,gBAAgB,GAAG,iBAAiB,CAAC;AAEzE,OAAO,KAAK,EAAE,OAAO,EAAE,MAAM,mBAAmB,CAAC;AAEjD,MAAM,WAAW,IAAI,CAAC,WAAW,SAAS,OAAO,GAAG,OAAO;IAC1D,IAAI,EAAE,MAAM,CAAC;IACb,WAAW,EAAE,MAAM,CAAC;IACpB,UAAU,EAAE,WAAW,CAAC;CACxB;AAED,MAAM,WAAW,OAAO;IACvB,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,QAAQ,EAAE,OAAO,EAAE,CAAC;IACpB,KAAK,CAAC,EAAE,IAAI,EAAE,CAAC;CACf;AAED,MAAM,MAAM,qBAAqB,GAC9B;IAAE,IAAI,EAAE,OAAO,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GAC5C;IAAE,IAAI,EAAE,YAAY,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GACvE;IAAE,IAAI,EAAE,YAAY,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GACtF;IAAE,IAAI,EAAE,UAAU,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GACtF;IAAE,IAAI,EAAE,gBAAgB,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GAC3E;IAAE,IAAI,EAAE,gBAAgB,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GAC1F;IAAE,IAAI,EAAE,cAAc,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GAC1F;IAAE,IAAI,EAAE,gBAAgB,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GAC3E;IAAE,IAAI,EAAE,gBAAgB,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GAC1F;IAAE,IAAI,EAAE,cAAc,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,QAAQ,EAAE,QAAQ,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GAC7F;IAAE,IAAI,EAAE,MAAM,CAAC;IAAC,MAAM,EAAE,OAAO,CAAC,UAAU,EAAE,MAAM,GAAG,QAAQ,GAAG,SAAS,CAAC,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GACvG;IAAE,IAAI,EAAE,OAAO,CAAC;IAAC,MAAM,EAAE,OAAO,CAAC,UAAU,EAAE,SAAS,GAAG,OAAO,CAAC,CAAC;IAAC,KAAK,EAAE,gBAAgB,CAAA;CAAE,CAAC;AAEhG;;;GAGG;AACH,MAAM,WAAW,YAAY;IAC5B,wFAAwF;IACxF,aAAa,CAAC,EAAE,OAAO,CAAC;IACxB,yGAAyG;IACzG,qBAAqB,CAAC,EAAE,OAAO,CAAC;IAChC,yFAAyF;IACzF,uBAAuB,CAAC,EAAE,OAAO,CAAC;IAClC,0EAA0E;IAC1E,cAAc,CAAC,EAAE,uBAAuB,GAAG,YAAY,CAAC;IACxD,sFAAsF;IACtF,sBAAsB,CAAC,EAAE,OAAO,CAAC;IACjC,2HAA2H;IAC3H,gCAAgC,CAAC,EAAE,OAAO,CAAC;IAC3C,4HAA4H;IAC5H,sBAAsB,CAAC,EAAE,OAAO,CAAC;IACjC,kIAAkI;IAClI,sBAAsB,CAAC,EAAE,OAAO,CAAC;CACjC;AAGD,MAAM,WAAW,KAAK,CAAC,IAAI,SAAS,GAAG;IACtC,EAAE,EAAE,MAAM,CAAC;IACX,IAAI,EAAE,MAAM,CAAC;IACb,GAAG,EAAE,IAAI,CAAC;IACV,QAAQ,EAAE,QAAQ,CAAC;IACnB,OAAO,EAAE,MAAM,CAAC;IAChB,SAAS,EAAE,OAAO,CAAC;IACnB,KAAK,EAAE,CAAC,MAAM,GAAG,OAAO,CAAC,EAAE,CAAC;IAC5B,IAAI,EAAE;QACL,KAAK,EAAE,MAAM,CAAC;QACd,MAAM,EAAE,MAAM,CAAC;QACf,SAAS,EAAE,MAAM,CAAC;QAClB,UAAU,EAAE,MAAM,CAAC;KACnB,CAAC;IACF,aAAa,EAAE,MAAM,CAAC;IACtB,SAAS,EAAE,MAAM,CAAC;IAClB,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;IACjC,kGAAkG;IAClG,MAAM,CAAC,EAAE,IAAI,SAAS,oBAAoB,GAAG,YAAY,GAAG,KAAK,CAAC;CAClE","sourcesContent":["import type { AnthropicOptions } from \"./providers/anthropic.js\";\nimport type { GoogleOptions } from \"./providers/google.js\";\nimport type { GoogleGeminiCliOptions } from \"./providers/google-gemini-cli.js\";\nimport type { GoogleVertexOptions } from \"./providers/google-vertex.js\";\nimport type { OpenAICodexResponsesOptions } from \"./providers/openai-codex-responses.js\";\nimport type { OpenAICompletionsOptions } from \"./providers/openai-completions.js\";\nimport type { OpenAIResponsesOptions } from \"./providers/openai-responses.js\";\nimport type { AssistantMessageEventStream } from \"./utils/event-stream.js\";\n\nexport type { AssistantMessageEventStream } from \"./utils/event-stream.js\";\n\nexport type Api =\n\t| \"openai-completions\"\n\t| \"openai-responses\"\n\t| \"openai-codex-responses\"\n\t| \"anthropic-messages\"\n\t| \"google-generative-ai\"\n\t| \"google-gemini-cli\"\n\t| \"google-vertex\";\n\nexport interface ApiOptionsMap {\n\t\"anthropic-messages\": AnthropicOptions;\n\t\"openai-completions\": OpenAICompletionsOptions;\n\t\"openai-responses\": OpenAIResponsesOptions;\n\t\"openai-codex-responses\": OpenAICodexResponsesOptions;\n\t\"google-generative-ai\": GoogleOptions;\n\t\"google-gemini-cli\": GoogleGeminiCliOptions;\n\t\"google-vertex\": GoogleVertexOptions;\n}\n\n// Compile-time exhaustiveness check - this will fail if ApiOptionsMap doesn't have all KnownApi keys\ntype _CheckExhaustive = ApiOptionsMap extends Record<Api, StreamOptions>\n\t? Record<Api, StreamOptions> extends ApiOptionsMap\n\t\t? true\n\t\t: [\"ApiOptionsMap is missing some KnownApi values\", Exclude<Api, keyof ApiOptionsMap>]\n\t: [\"ApiOptionsMap doesn't extend Record<KnownApi, StreamOptions>\"];\nconst _exhaustive: _CheckExhaustive = true;\n\n// Helper type to get options for a specific API\nexport type OptionsForApi<TApi extends Api> = ApiOptionsMap[TApi];\n\nexport type KnownProvider =\n\t| \"anthropic\"\n\t| \"google\"\n\t| \"google-gemini-cli\"\n\t| \"google-antigravity\"\n\t| \"google-vertex\"\n\t| \"openai\"\n\t| \"openai-codex\"\n\t| \"github-copilot\"\n\t| \"xai\"\n\t| \"groq\"\n\t| \"cerebras\"\n\t| \"openrouter\"\n\t| \"zai\"\n\t| \"mistral\";\nexport type Provider = KnownProvider | string;\n\nexport type ThinkingLevel = \"minimal\" | \"low\" | \"medium\" | \"high\" | \"xhigh\";\n\n// Base options all providers share\nexport interface StreamOptions {\n\ttemperature?: number;\n\tmaxTokens?: number;\n\tsignal?: AbortSignal;\n\tapiKey?: string;\n\t/**\n\t * Optional session identifier for providers that support session-based caching.\n\t * Providers can use this to enable prompt caching, request routing, or other\n\t * session-aware features. Ignored by providers that don't support it.\n\t */\n\tsessionId?: string;\n}\n\n// Unified options with reasoning passed to streamSimple() and completeSimple()\nexport interface SimpleStreamOptions extends StreamOptions {\n\treasoning?: ThinkingLevel;\n}\n\n// Generic StreamFunction with typed options\nexport type StreamFunction<TApi extends Api> = (\n\tmodel: Model<TApi>,\n\tcontext: Context,\n\toptions: OptionsForApi<TApi>,\n) => AssistantMessageEventStream;\n\nexport interface TextContent {\n\ttype: \"text\";\n\ttext: string;\n\ttextSignature?: string; // e.g., for OpenAI responses, the message ID\n}\n\nexport interface ThinkingContent {\n\ttype: \"thinking\";\n\tthinking: string;\n\tthinkingSignature?: string; // e.g., for OpenAI responses, the reasoning item ID\n}\n\nexport interface ImageContent {\n\ttype: \"image\";\n\tdata: string; // base64 encoded image data\n\tmimeType: string; // e.g., \"image/jpeg\", \"image/png\"\n}\n\nexport interface ToolCall {\n\ttype: \"toolCall\";\n\tid: string;\n\tname: string;\n\targuments: Record<string, any>;\n\tthoughtSignature?: string; // Google-specific: opaque signature for reusing thought context\n}\n\nexport interface Usage {\n\tinput: number;\n\toutput: number;\n\tcacheRead: number;\n\tcacheWrite: number;\n\ttotalTokens: number;\n\tcost: {\n\t\tinput: number;\n\t\toutput: number;\n\t\tcacheRead: number;\n\t\tcacheWrite: number;\n\t\ttotal: number;\n\t};\n}\n\nexport type StopReason = \"stop\" | \"length\" | \"toolUse\" | \"error\" | \"aborted\";\n\nexport interface UserMessage {\n\trole: \"user\";\n\tcontent: string | (TextContent | ImageContent)[];\n\ttimestamp: number; // Unix timestamp in milliseconds\n}\n\nexport interface AssistantMessage {\n\trole: \"assistant\";\n\tcontent: (TextContent | ThinkingContent | ToolCall)[];\n\tapi: Api;\n\tprovider: Provider;\n\tmodel: string;\n\tusage: Usage;\n\tstopReason: StopReason;\n\terrorMessage?: string;\n\ttimestamp: number; // Unix timestamp in milliseconds\n}\n\nexport interface ToolResultMessage<TDetails = any> {\n\trole: \"toolResult\";\n\ttoolCallId: string;\n\ttoolName: string;\n\tcontent: (TextContent | ImageContent)[]; // Supports text and images\n\tdetails?: TDetails;\n\tisError: boolean;\n\ttimestamp: number; // Unix timestamp in milliseconds\n}\n\nexport type Message = UserMessage | AssistantMessage | ToolResultMessage;\n\nimport type { TSchema } from \"@sinclair/typebox\";\n\nexport interface Tool<TParameters extends TSchema = TSchema> {\n\tname: string;\n\tdescription: string;\n\tparameters: TParameters;\n}\n\nexport interface Context {\n\tsystemPrompt?: string;\n\tmessages: Message[];\n\ttools?: Tool[];\n}\n\nexport type AssistantMessageEvent =\n\t| { type: \"start\"; partial: AssistantMessage }\n\t| { type: \"text_start\"; contentIndex: number; partial: AssistantMessage }\n\t| { type: \"text_delta\"; contentIndex: number; delta: string; partial: AssistantMessage }\n\t| { type: \"text_end\"; contentIndex: number; content: string; partial: AssistantMessage }\n\t| { type: \"thinking_start\"; contentIndex: number; partial: AssistantMessage }\n\t| { type: \"thinking_delta\"; contentIndex: number; delta: string; partial: AssistantMessage }\n\t| { type: \"thinking_end\"; contentIndex: number; content: string; partial: AssistantMessage }\n\t| { type: \"toolcall_start\"; contentIndex: number; partial: AssistantMessage }\n\t| { type: \"toolcall_delta\"; contentIndex: number; delta: string; partial: AssistantMessage }\n\t| { type: \"toolcall_end\"; contentIndex: number; toolCall: ToolCall; partial: AssistantMessage }\n\t| { type: \"done\"; reason: Extract<StopReason, \"stop\" | \"length\" | \"toolUse\">; message: AssistantMessage }\n\t| { type: \"error\"; reason: Extract<StopReason, \"aborted\" | \"error\">; error: AssistantMessage };\n\n/**\n * Compatibility settings for openai-completions API.\n * Use this to override URL-based auto-detection for custom providers.\n */\nexport interface OpenAICompat {\n\t/** Whether the provider supports the `store` field. Default: auto-detected from URL. */\n\tsupportsStore?: boolean;\n\t/** Whether the provider supports the `developer` role (vs `system`). Default: auto-detected from URL. */\n\tsupportsDeveloperRole?: boolean;\n\t/** Whether the provider supports `reasoning_effort`. Default: auto-detected from URL. */\n\tsupportsReasoningEffort?: boolean;\n\t/** Which field to use for max tokens. Default: auto-detected from URL. */\n\tmaxTokensField?: \"max_completion_tokens\" | \"max_tokens\";\n\t/** Whether tool results require the `name` field. Default: auto-detected from URL. */\n\trequiresToolResultName?: boolean;\n\t/** Whether a user message after tool results requires an assistant message in between. Default: auto-detected from URL. */\n\trequiresAssistantAfterToolResult?: boolean;\n\t/** Whether thinking blocks must be converted to text blocks with <thinking> delimiters. Default: auto-detected from URL. */\n\trequiresThinkingAsText?: boolean;\n\t/** Whether tool call IDs must be normalized to Mistral format (exactly 9 alphanumeric chars). Default: auto-detected from URL. */\n\trequiresMistralToolIds?: boolean;\n}\n\n// Model interface for the unified model system\nexport interface Model<TApi extends Api> {\n\tid: string;\n\tname: string;\n\tapi: TApi;\n\tprovider: Provider;\n\tbaseUrl: string;\n\treasoning: boolean;\n\tinput: (\"text\" | \"image\")[];\n\tcost: {\n\t\tinput: number; // $/million tokens\n\t\toutput: number; // $/million tokens\n\t\tcacheRead: number; // $/million tokens\n\t\tcacheWrite: number; // $/million tokens\n\t};\n\tcontextWindow: number;\n\tmaxTokens: number;\n\theaders?: Record<string, string>;\n\t/** Compatibility overrides for openai-completions API. If not set, auto-detected from baseUrl. */\n\tcompat?: TApi extends \"openai-completions\" ? OpenAICompat : never;\n}\n"]}
+{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,0BAA0B,CAAC;AACjE,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,uBAAuB,CAAC;AAC3D,OAAO,KAAK,EAAE,sBAAsB,EAAE,MAAM,kCAAkC,CAAC;AAC/E,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,8BAA8B,CAAC;AACxE,OAAO,KAAK,EAAE,2BAA2B,EAAE,MAAM,uCAAuC,CAAC;AACzF,OAAO,KAAK,EAAE,wBAAwB,EAAE,MAAM,mCAAmC,CAAC;AAClF,OAAO,KAAK,EAAE,sBAAsB,EAAE,MAAM,iCAAiC,CAAC;AAC9E,OAAO,KAAK,EAAE,2BAA2B,EAAE,MAAM,yBAAyB,CAAC;AAE3E,YAAY,EAAE,2BAA2B,EAAE,MAAM,yBAAyB,CAAC;AAE3E,MAAM,MAAM,GAAG,GACZ,oBAAoB,GACpB,kBAAkB,GAClB,wBAAwB,GACxB,oBAAoB,GACpB,sBAAsB,GACtB,mBAAmB,GACnB,eAAe,CAAC;AAEnB,MAAM,WAAW,aAAa;IAC7B,oBAAoB,EAAE,gBAAgB,CAAC;IACvC,oBAAoB,EAAE,wBAAwB,CAAC;IAC/C,kBAAkB,EAAE,sBAAsB,CAAC;IAC3C,wBAAwB,EAAE,2BAA2B,CAAC;IACtD,sBAAsB,EAAE,aAAa,CAAC;IACtC,mBAAmB,EAAE,sBAAsB,CAAC;IAC5C,eAAe,EAAE,mBAAmB,CAAC;CACrC;AAWD,MAAM,MAAM,aAAa,CAAC,IAAI,SAAS,GAAG,IAAI,aAAa,CAAC,IAAI,CAAC,CAAC;AAElE,MAAM,MAAM,aAAa,GACtB,WAAW,GACX,QAAQ,GACR,mBAAmB,GACnB,oBAAoB,GACpB,eAAe,GACf,QAAQ,GACR,cAAc,GACd,gBAAgB,GAChB,KAAK,GACL,MAAM,GACN,UAAU,GACV,YAAY,GACZ,KAAK,GACL,SAAS,CAAC;AACb,MAAM,MAAM,QAAQ,GAAG,aAAa,GAAG,MAAM,CAAC;AAE9C,MAAM,MAAM,aAAa,GAAG,SAAS,GAAG,KAAK,GAAG,QAAQ,GAAG,MAAM,GAAG,OAAO,CAAC;AAE5E,yEAAyE;AACzE,MAAM,WAAW,eAAe;IAC/B,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,GAAG,CAAC,EAAE,MAAM,CAAC;IACb,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,IAAI,CAAC,EAAE,MAAM,CAAC;CACd;AAGD,MAAM,WAAW,aAAa;IAC7B,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,MAAM,CAAC,EAAE,WAAW,CAAC;IACrB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB;;;;OAIG;IACH,SAAS,CAAC,EAAE,MAAM,CAAC;CACnB;AAGD,MAAM,WAAW,mBAAoB,SAAQ,aAAa;IACzD,SAAS,CAAC,EAAE,aAAa,CAAC;IAC1B,4EAA4E;IAC5E,eAAe,CAAC,EAAE,eAAe,CAAC;CAClC;AAGD,MAAM,MAAM,cAAc,CAAC,IAAI,SAAS,GAAG,IAAI,CAC9C,KAAK,EAAE,KAAK,CAAC,IAAI,CAAC,EAClB,OAAO,EAAE,OAAO,EAChB,OAAO,EAAE,aAAa,CAAC,IAAI,CAAC,KACxB,2BAA2B,CAAC;AAEjC,MAAM,WAAW,WAAW;IAC3B,IAAI,EAAE,MAAM,CAAC;IACb,IAAI,EAAE,MAAM,CAAC;IACb,aAAa,CAAC,EAAE,MAAM,CAAC;CACvB;AAED,MAAM,WAAW,eAAe;IAC/B,IAAI,EAAE,UAAU,CAAC;IACjB,QAAQ,EAAE,MAAM,CAAC;IACjB,iBAAiB,CAAC,EAAE,MAAM,CAAC;CAC3B;AAED,MAAM,WAAW,YAAY;IAC5B,IAAI,EAAE,OAAO,CAAC;IACd,IAAI,EAAE,MAAM,CAAC;IACb,QAAQ,EAAE,MAAM,CAAC;CACjB;AAED,MAAM,WAAW,QAAQ;IACxB,IAAI,EAAE,UAAU,CAAC;IACjB,EAAE,EAAE,MAAM,CAAC;IACX,IAAI,EAAE,MAAM,CAAC;IACb,SAAS,EAAE,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC;IAC/B,gBAAgB,CAAC,EAAE,MAAM,CAAC;CAC1B;AAED,MAAM,WAAW,KAAK;IACrB,KAAK,EAAE,MAAM,CAAC;IACd,MAAM,EAAE,MAAM,CAAC;IACf,SAAS,EAAE,MAAM,CAAC;IAClB,UAAU,EAAE,MAAM,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;IACpB,IAAI,EAAE;QACL,KAAK,EAAE,MAAM,CAAC;QACd,MAAM,EAAE,MAAM,CAAC;QACf,SAAS,EAAE,MAAM,CAAC;QAClB,UAAU,EAAE,MAAM,CAAC;QACnB,KAAK,EAAE,MAAM,CAAC;KACd,CAAC;CACF;AAED,MAAM,MAAM,UAAU,GAAG,MAAM,GAAG,QAAQ,GAAG,SAAS,GAAG,OAAO,GAAG,SAAS,CAAC;AAE7E,MAAM,WAAW,WAAW;IAC3B,IAAI,EAAE,MAAM,CAAC;IACb,OAAO,EAAE,MAAM,GAAG,CAAC,WAAW,GAAG,YAAY,CAAC,EAAE,CAAC;IACjD,SAAS,EAAE,MAAM,CAAC;CAClB;AAED,MAAM,WAAW,gBAAgB;IAChC,IAAI,EAAE,WAAW,CAAC;IAClB,OAAO,EAAE,CAAC,WAAW,GAAG,eAAe,GAAG,QAAQ,CAAC,EAAE,CAAC;IACtD,GAAG,EAAE,GAAG,CAAC;IACT,QAAQ,EAAE,QAAQ,CAAC;IACnB,KAAK,EAAE,MAAM,CAAC;IACd,KAAK,EAAE,KAAK,CAAC;IACb,UAAU,EAAE,UAAU,CAAC;IACvB,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,SAAS,EAAE,MAAM,CAAC;CAClB;AAED,MAAM,WAAW,iBAAiB,CAAC,QAAQ,GAAG,GAAG;IAChD,IAAI,EAAE,YAAY,CAAC;IACnB,UAAU,EAAE,MAAM,CAAC;IACnB,QAAQ,EAAE,MAAM,CAAC;IACjB,OAAO,EAAE,CAAC,WAAW,GAAG,YAAY,CAAC,EAAE,CAAC;IACxC,OAAO,CAAC,EAAE,QAAQ,CAAC;IACnB,OAAO,EAAE,OAAO,CAAC;IACjB,SAAS,EAAE,MAAM,CAAC;CAClB;AAED,MAAM,MAAM,OAAO,GAAG,WAAW,GAAG,gBAAgB,GAAG,iBAAiB,CAAC;AAEzE,OAAO,KAAK,EAAE,OAAO,EAAE,MAAM,mBAAmB,CAAC;AAEjD,MAAM,WAAW,IAAI,CAAC,WAAW,SAAS,OAAO,GAAG,OAAO;IAC1D,IAAI,EAAE,MAAM,CAAC;IACb,WAAW,EAAE,MAAM,CAAC;IACpB,UAAU,EAAE,WAAW,CAAC;CACxB;AAED,MAAM,WAAW,OAAO;IACvB,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,QAAQ,EAAE,OAAO,EAAE,CAAC;IACpB,KAAK,CAAC,EAAE,IAAI,EAAE,CAAC;CACf;AAED,MAAM,MAAM,qBAAqB,GAC9B;IAAE,IAAI,EAAE,OAAO,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GAC5C;IAAE,IAAI,EAAE,YAAY,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GACvE;IAAE,IAAI,EAAE,YAAY,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GACtF;IAAE,IAAI,EAAE,UAAU,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GACtF;IAAE,IAAI,EAAE,gBAAgB,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GAC3E;IAAE,IAAI,EAAE,gBAAgB,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GAC1F;IAAE,IAAI,EAAE,cAAc,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GAC1F;IAAE,IAAI,EAAE,gBAAgB,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GAC3E;IAAE,IAAI,EAAE,gBAAgB,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GAC1F;IAAE,IAAI,EAAE,cAAc,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,QAAQ,EAAE,QAAQ,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GAC7F;IAAE,IAAI,EAAE,MAAM,CAAC;IAAC,MAAM,EAAE,OAAO,CAAC,UAAU,EAAE,MAAM,GAAG,QAAQ,GAAG,SAAS,CAAC,CAAC;IAAC,OAAO,EAAE,gBAAgB,CAAA;CAAE,GACvG;IAAE,IAAI,EAAE,OAAO,CAAC;IAAC,MAAM,EAAE,OAAO,CAAC,UAAU,EAAE,SAAS,GAAG,OAAO,CAAC,CAAC;IAAC,KAAK,EAAE,gBAAgB,CAAA;CAAE,CAAC;AAEhG;;;GAGG;AACH,MAAM,WAAW,YAAY;IAC5B,wFAAwF;IACxF,aAAa,CAAC,EAAE,OAAO,CAAC;IACxB,yGAAyG;IACzG,qBAAqB,CAAC,EAAE,OAAO,CAAC;IAChC,yFAAyF;IACzF,uBAAuB,CAAC,EAAE,OAAO,CAAC;IAClC,0EAA0E;IAC1E,cAAc,CAAC,EAAE,uBAAuB,GAAG,YAAY,CAAC;IACxD,sFAAsF;IACtF,sBAAsB,CAAC,EAAE,OAAO,CAAC;IACjC,2HAA2H;IAC3H,gCAAgC,CAAC,EAAE,OAAO,CAAC;IAC3C,4HAA4H;IAC5H,sBAAsB,CAAC,EAAE,OAAO,CAAC;IACjC,kIAAkI;IAClI,sBAAsB,CAAC,EAAE,OAAO,CAAC;CACjC;AAGD,MAAM,WAAW,KAAK,CAAC,IAAI,SAAS,GAAG;IACtC,EAAE,EAAE,MAAM,CAAC;IACX,IAAI,EAAE,MAAM,CAAC;IACb,GAAG,EAAE,IAAI,CAAC;IACV,QAAQ,EAAE,QAAQ,CAAC;IACnB,OAAO,EAAE,MAAM,CAAC;IAChB,SAAS,EAAE,OAAO,CAAC;IACnB,KAAK,EAAE,CAAC,MAAM,GAAG,OAAO,CAAC,EAAE,CAAC;IAC5B,IAAI,EAAE;QACL,KAAK,EAAE,MAAM,CAAC;QACd,MAAM,EAAE,MAAM,CAAC;QACf,SAAS,EAAE,MAAM,CAAC;QAClB,UAAU,EAAE,MAAM,CAAC;KACnB,CAAC;IACF,aAAa,EAAE,MAAM,CAAC;IACtB,SAAS,EAAE,MAAM,CAAC;IAClB,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;IACjC,kGAAkG;IAClG,MAAM,CAAC,EAAE,IAAI,SAAS,oBAAoB,GAAG,YAAY,GAAG,KAAK,CAAC;CAClE","sourcesContent":["import type { AnthropicOptions } from \"./providers/anthropic.js\";\nimport type { GoogleOptions } from \"./providers/google.js\";\nimport type { GoogleGeminiCliOptions } from \"./providers/google-gemini-cli.js\";\nimport type { GoogleVertexOptions } from \"./providers/google-vertex.js\";\nimport type { OpenAICodexResponsesOptions } from \"./providers/openai-codex-responses.js\";\nimport type { OpenAICompletionsOptions } from \"./providers/openai-completions.js\";\nimport type { OpenAIResponsesOptions } from \"./providers/openai-responses.js\";\nimport type { AssistantMessageEventStream } from \"./utils/event-stream.js\";\n\nexport type { AssistantMessageEventStream } from \"./utils/event-stream.js\";\n\nexport type Api =\n\t| \"openai-completions\"\n\t| \"openai-responses\"\n\t| \"openai-codex-responses\"\n\t| \"anthropic-messages\"\n\t| \"google-generative-ai\"\n\t| \"google-gemini-cli\"\n\t| \"google-vertex\";\n\nexport interface ApiOptionsMap {\n\t\"anthropic-messages\": AnthropicOptions;\n\t\"openai-completions\": OpenAICompletionsOptions;\n\t\"openai-responses\": OpenAIResponsesOptions;\n\t\"openai-codex-responses\": OpenAICodexResponsesOptions;\n\t\"google-generative-ai\": GoogleOptions;\n\t\"google-gemini-cli\": GoogleGeminiCliOptions;\n\t\"google-vertex\": GoogleVertexOptions;\n}\n\n// Compile-time exhaustiveness check - this will fail if ApiOptionsMap doesn't have all KnownApi keys\ntype _CheckExhaustive = ApiOptionsMap extends Record<Api, StreamOptions>\n\t? Record<Api, StreamOptions> extends ApiOptionsMap\n\t\t? true\n\t\t: [\"ApiOptionsMap is missing some KnownApi values\", Exclude<Api, keyof ApiOptionsMap>]\n\t: [\"ApiOptionsMap doesn't extend Record<KnownApi, StreamOptions>\"];\nconst _exhaustive: _CheckExhaustive = true;\n\n// Helper type to get options for a specific API\nexport type OptionsForApi<TApi extends Api> = ApiOptionsMap[TApi];\n\nexport type KnownProvider =\n\t| \"anthropic\"\n\t| \"google\"\n\t| \"google-gemini-cli\"\n\t| \"google-antigravity\"\n\t| \"google-vertex\"\n\t| \"openai\"\n\t| \"openai-codex\"\n\t| \"github-copilot\"\n\t| \"xai\"\n\t| \"groq\"\n\t| \"cerebras\"\n\t| \"openrouter\"\n\t| \"zai\"\n\t| \"mistral\";\nexport type Provider = KnownProvider | string;\n\nexport type ThinkingLevel = \"minimal\" | \"low\" | \"medium\" | \"high\" | \"xhigh\";\n\n/** Token budgets for each thinking level (token-based providers only) */\nexport interface ThinkingBudgets {\n\tminimal?: number;\n\tlow?: number;\n\tmedium?: number;\n\thigh?: number;\n}\n\n// Base options all providers share\nexport interface StreamOptions {\n\ttemperature?: number;\n\tmaxTokens?: number;\n\tsignal?: AbortSignal;\n\tapiKey?: string;\n\t/**\n\t * Optional session identifier for providers that support session-based caching.\n\t * Providers can use this to enable prompt caching, request routing, or other\n\t * session-aware features. Ignored by providers that don't support it.\n\t */\n\tsessionId?: string;\n}\n\n// Unified options with reasoning passed to streamSimple() and completeSimple()\nexport interface SimpleStreamOptions extends StreamOptions {\n\treasoning?: ThinkingLevel;\n\t/** Custom token budgets for thinking levels (token-based providers only) */\n\tthinkingBudgets?: ThinkingBudgets;\n}\n\n// Generic StreamFunction with typed options\nexport type StreamFunction<TApi extends Api> = (\n\tmodel: Model<TApi>,\n\tcontext: Context,\n\toptions: OptionsForApi<TApi>,\n) => AssistantMessageEventStream;\n\nexport interface TextContent {\n\ttype: \"text\";\n\ttext: string;\n\ttextSignature?: string; // e.g., for OpenAI responses, the message ID\n}\n\nexport interface ThinkingContent {\n\ttype: \"thinking\";\n\tthinking: string;\n\tthinkingSignature?: string; // e.g., for OpenAI responses, the reasoning item ID\n}\n\nexport interface ImageContent {\n\ttype: \"image\";\n\tdata: string; // base64 encoded image data\n\tmimeType: string; // e.g., \"image/jpeg\", \"image/png\"\n}\n\nexport interface ToolCall {\n\ttype: \"toolCall\";\n\tid: string;\n\tname: string;\n\targuments: Record<string, any>;\n\tthoughtSignature?: string; // Google-specific: opaque signature for reusing thought context\n}\n\nexport interface Usage {\n\tinput: number;\n\toutput: number;\n\tcacheRead: number;\n\tcacheWrite: number;\n\ttotalTokens: number;\n\tcost: {\n\t\tinput: number;\n\t\toutput: number;\n\t\tcacheRead: number;\n\t\tcacheWrite: number;\n\t\ttotal: number;\n\t};\n}\n\nexport type StopReason = \"stop\" | \"length\" | \"toolUse\" | \"error\" | \"aborted\";\n\nexport interface UserMessage {\n\trole: \"user\";\n\tcontent: string | (TextContent | ImageContent)[];\n\ttimestamp: number; // Unix timestamp in milliseconds\n}\n\nexport interface AssistantMessage {\n\trole: \"assistant\";\n\tcontent: (TextContent | ThinkingContent | ToolCall)[];\n\tapi: Api;\n\tprovider: Provider;\n\tmodel: string;\n\tusage: Usage;\n\tstopReason: StopReason;\n\terrorMessage?: string;\n\ttimestamp: number; // Unix timestamp in milliseconds\n}\n\nexport interface ToolResultMessage<TDetails = any> {\n\trole: \"toolResult\";\n\ttoolCallId: string;\n\ttoolName: string;\n\tcontent: (TextContent | ImageContent)[]; // Supports text and images\n\tdetails?: TDetails;\n\tisError: boolean;\n\ttimestamp: number; // Unix timestamp in milliseconds\n}\n\nexport type Message = UserMessage | AssistantMessage | ToolResultMessage;\n\nimport type { TSchema } from \"@sinclair/typebox\";\n\nexport interface Tool<TParameters extends TSchema = TSchema> {\n\tname: string;\n\tdescription: string;\n\tparameters: TParameters;\n}\n\nexport interface Context {\n\tsystemPrompt?: string;\n\tmessages: Message[];\n\ttools?: Tool[];\n}\n\nexport type AssistantMessageEvent =\n\t| { type: \"start\"; partial: AssistantMessage }\n\t| { type: \"text_start\"; contentIndex: number; partial: AssistantMessage }\n\t| { type: \"text_delta\"; contentIndex: number; delta: string; partial: AssistantMessage }\n\t| { type: \"text_end\"; contentIndex: number; content: string; partial: AssistantMessage }\n\t| { type: \"thinking_start\"; contentIndex: number; partial: AssistantMessage }\n\t| { type: \"thinking_delta\"; contentIndex: number; delta: string; partial: AssistantMessage }\n\t| { type: \"thinking_end\"; contentIndex: number; content: string; partial: AssistantMessage }\n\t| { type: \"toolcall_start\"; contentIndex: number; partial: AssistantMessage }\n\t| { type: \"toolcall_delta\"; contentIndex: number; delta: string; partial: AssistantMessage }\n\t| { type: \"toolcall_end\"; contentIndex: number; toolCall: ToolCall; partial: AssistantMessage }\n\t| { type: \"done\"; reason: Extract<StopReason, \"stop\" | \"length\" | \"toolUse\">; message: AssistantMessage }\n\t| { type: \"error\"; reason: Extract<StopReason, \"aborted\" | \"error\">; error: AssistantMessage };\n\n/**\n * Compatibility settings for openai-completions API.\n * Use this to override URL-based auto-detection for custom providers.\n */\nexport interface OpenAICompat {\n\t/** Whether the provider supports the `store` field. Default: auto-detected from URL. */\n\tsupportsStore?: boolean;\n\t/** Whether the provider supports the `developer` role (vs `system`). Default: auto-detected from URL. */\n\tsupportsDeveloperRole?: boolean;\n\t/** Whether the provider supports `reasoning_effort`. Default: auto-detected from URL. */\n\tsupportsReasoningEffort?: boolean;\n\t/** Which field to use for max tokens. Default: auto-detected from URL. */\n\tmaxTokensField?: \"max_completion_tokens\" | \"max_tokens\";\n\t/** Whether tool results require the `name` field. Default: auto-detected from URL. */\n\trequiresToolResultName?: boolean;\n\t/** Whether a user message after tool results requires an assistant message in between. Default: auto-detected from URL. */\n\trequiresAssistantAfterToolResult?: boolean;\n\t/** Whether thinking blocks must be converted to text blocks with <thinking> delimiters. Default: auto-detected from URL. */\n\trequiresThinkingAsText?: boolean;\n\t/** Whether tool call IDs must be normalized to Mistral format (exactly 9 alphanumeric chars). Default: auto-detected from URL. */\n\trequiresMistralToolIds?: boolean;\n}\n\n// Model interface for the unified model system\nexport interface Model<TApi extends Api> {\n\tid: string;\n\tname: string;\n\tapi: TApi;\n\tprovider: Provider;\n\tbaseUrl: string;\n\treasoning: boolean;\n\tinput: (\"text\" | \"image\")[];\n\tcost: {\n\t\tinput: number; // $/million tokens\n\t\toutput: number; // $/million tokens\n\t\tcacheRead: number; // $/million tokens\n\t\tcacheWrite: number; // $/million tokens\n\t};\n\tcontextWindow: number;\n\tmaxTokens: number;\n\theaders?: Record<string, string>;\n\t/** Compatibility overrides for openai-completions API. If not set, auto-detected from baseUrl. */\n\tcompat?: TApi extends \"openai-completions\" ? OpenAICompat : never;\n}\n"]}

package/dist/types.js.map

@@ -1 +1 @@
-{"version":3,"file":"types.js","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAoCA,MAAM,WAAW,GAAqB,IAAI,CAAC","sourcesContent":["import type { AnthropicOptions } from \"./providers/anthropic.js\";\nimport type { GoogleOptions } from \"./providers/google.js\";\nimport type { GoogleGeminiCliOptions } from \"./providers/google-gemini-cli.js\";\nimport type { GoogleVertexOptions } from \"./providers/google-vertex.js\";\nimport type { OpenAICodexResponsesOptions } from \"./providers/openai-codex-responses.js\";\nimport type { OpenAICompletionsOptions } from \"./providers/openai-completions.js\";\nimport type { OpenAIResponsesOptions } from \"./providers/openai-responses.js\";\nimport type { AssistantMessageEventStream } from \"./utils/event-stream.js\";\n\nexport type { AssistantMessageEventStream } from \"./utils/event-stream.js\";\n\nexport type Api =\n\t| \"openai-completions\"\n\t| \"openai-responses\"\n\t| \"openai-codex-responses\"\n\t| \"anthropic-messages\"\n\t| \"google-generative-ai\"\n\t| \"google-gemini-cli\"\n\t| \"google-vertex\";\n\nexport interface ApiOptionsMap {\n\t\"anthropic-messages\": AnthropicOptions;\n\t\"openai-completions\": OpenAICompletionsOptions;\n\t\"openai-responses\": OpenAIResponsesOptions;\n\t\"openai-codex-responses\": OpenAICodexResponsesOptions;\n\t\"google-generative-ai\": GoogleOptions;\n\t\"google-gemini-cli\": GoogleGeminiCliOptions;\n\t\"google-vertex\": GoogleVertexOptions;\n}\n\n// Compile-time exhaustiveness check - this will fail if ApiOptionsMap doesn't have all KnownApi keys\ntype _CheckExhaustive = ApiOptionsMap extends Record<Api, StreamOptions>\n\t? Record<Api, StreamOptions> extends ApiOptionsMap\n\t\t? true\n\t\t: [\"ApiOptionsMap is missing some KnownApi values\", Exclude<Api, keyof ApiOptionsMap>]\n\t: [\"ApiOptionsMap doesn't extend Record<KnownApi, StreamOptions>\"];\nconst _exhaustive: _CheckExhaustive = true;\n\n// Helper type to get options for a specific API\nexport type OptionsForApi<TApi extends Api> = ApiOptionsMap[TApi];\n\nexport type KnownProvider =\n\t| \"anthropic\"\n\t| \"google\"\n\t| \"google-gemini-cli\"\n\t| \"google-antigravity\"\n\t| \"google-vertex\"\n\t| \"openai\"\n\t| \"openai-codex\"\n\t| \"github-copilot\"\n\t| \"xai\"\n\t| \"groq\"\n\t| \"cerebras\"\n\t| \"openrouter\"\n\t| \"zai\"\n\t| \"mistral\";\nexport type Provider = KnownProvider | string;\n\nexport type ThinkingLevel = \"minimal\" | \"low\" | \"medium\" | \"high\" | \"xhigh\";\n\n// Base options all providers share\nexport interface StreamOptions {\n\ttemperature?: number;\n\tmaxTokens?: number;\n\tsignal?: AbortSignal;\n\tapiKey?: string;\n\t/**\n\t * Optional session identifier for providers that support session-based caching.\n\t * Providers can use this to enable prompt caching, request routing, or other\n\t * session-aware features. Ignored by providers that don't support it.\n\t */\n\tsessionId?: string;\n}\n\n// Unified options with reasoning passed to streamSimple() and completeSimple()\nexport interface SimpleStreamOptions extends StreamOptions {\n\treasoning?: ThinkingLevel;\n}\n\n// Generic StreamFunction with typed options\nexport type StreamFunction<TApi extends Api> = (\n\tmodel: Model<TApi>,\n\tcontext: Context,\n\toptions: OptionsForApi<TApi>,\n) => AssistantMessageEventStream;\n\nexport interface TextContent {\n\ttype: \"text\";\n\ttext: string;\n\ttextSignature?: string; // e.g., for OpenAI responses, the message ID\n}\n\nexport interface ThinkingContent {\n\ttype: \"thinking\";\n\tthinking: string;\n\tthinkingSignature?: string; // e.g., for OpenAI responses, the reasoning item ID\n}\n\nexport interface ImageContent {\n\ttype: \"image\";\n\tdata: string; // base64 encoded image data\n\tmimeType: string; // e.g., \"image/jpeg\", \"image/png\"\n}\n\nexport interface ToolCall {\n\ttype: \"toolCall\";\n\tid: string;\n\tname: string;\n\targuments: Record<string, any>;\n\tthoughtSignature?: string; // Google-specific: opaque signature for reusing thought context\n}\n\nexport interface Usage {\n\tinput: number;\n\toutput: number;\n\tcacheRead: number;\n\tcacheWrite: number;\n\ttotalTokens: number;\n\tcost: {\n\t\tinput: number;\n\t\toutput: number;\n\t\tcacheRead: number;\n\t\tcacheWrite: number;\n\t\ttotal: number;\n\t};\n}\n\nexport type StopReason = \"stop\" | \"length\" | \"toolUse\" | \"error\" | \"aborted\";\n\nexport interface UserMessage {\n\trole: \"user\";\n\tcontent: string | (TextContent | ImageContent)[];\n\ttimestamp: number; // Unix timestamp in milliseconds\n}\n\nexport interface AssistantMessage {\n\trole: \"assistant\";\n\tcontent: (TextContent | ThinkingContent | ToolCall)[];\n\tapi: Api;\n\tprovider: Provider;\n\tmodel: string;\n\tusage: Usage;\n\tstopReason: StopReason;\n\terrorMessage?: string;\n\ttimestamp: number; // Unix timestamp in milliseconds\n}\n\nexport interface ToolResultMessage<TDetails = any> {\n\trole: \"toolResult\";\n\ttoolCallId: string;\n\ttoolName: string;\n\tcontent: (TextContent | ImageContent)[]; // Supports text and images\n\tdetails?: TDetails;\n\tisError: boolean;\n\ttimestamp: number; // Unix timestamp in milliseconds\n}\n\nexport type Message = UserMessage | AssistantMessage | ToolResultMessage;\n\nimport type { TSchema } from \"@sinclair/typebox\";\n\nexport interface Tool<TParameters extends TSchema = TSchema> {\n\tname: string;\n\tdescription: string;\n\tparameters: TParameters;\n}\n\nexport interface Context {\n\tsystemPrompt?: string;\n\tmessages: Message[];\n\ttools?: Tool[];\n}\n\nexport type AssistantMessageEvent =\n\t| { type: \"start\"; partial: AssistantMessage }\n\t| { type: \"text_start\"; contentIndex: number; partial: AssistantMessage }\n\t| { type: \"text_delta\"; contentIndex: number; delta: string; partial: AssistantMessage }\n\t| { type: \"text_end\"; contentIndex: number; content: string; partial: AssistantMessage }\n\t| { type: \"thinking_start\"; contentIndex: number; partial: AssistantMessage }\n\t| { type: \"thinking_delta\"; contentIndex: number; delta: string; partial: AssistantMessage }\n\t| { type: \"thinking_end\"; contentIndex: number; content: string; partial: AssistantMessage }\n\t| { type: \"toolcall_start\"; contentIndex: number; partial: AssistantMessage }\n\t| { type: \"toolcall_delta\"; contentIndex: number; delta: string; partial: AssistantMessage }\n\t| { type: \"toolcall_end\"; contentIndex: number; toolCall: ToolCall; partial: AssistantMessage }\n\t| { type: \"done\"; reason: Extract<StopReason, \"stop\" | \"length\" | \"toolUse\">; message: AssistantMessage }\n\t| { type: \"error\"; reason: Extract<StopReason, \"aborted\" | \"error\">; error: AssistantMessage };\n\n/**\n * Compatibility settings for openai-completions API.\n * Use this to override URL-based auto-detection for custom providers.\n */\nexport interface OpenAICompat {\n\t/** Whether the provider supports the `store` field. Default: auto-detected from URL. */\n\tsupportsStore?: boolean;\n\t/** Whether the provider supports the `developer` role (vs `system`). Default: auto-detected from URL. */\n\tsupportsDeveloperRole?: boolean;\n\t/** Whether the provider supports `reasoning_effort`. Default: auto-detected from URL. */\n\tsupportsReasoningEffort?: boolean;\n\t/** Which field to use for max tokens. Default: auto-detected from URL. */\n\tmaxTokensField?: \"max_completion_tokens\" | \"max_tokens\";\n\t/** Whether tool results require the `name` field. Default: auto-detected from URL. */\n\trequiresToolResultName?: boolean;\n\t/** Whether a user message after tool results requires an assistant message in between. Default: auto-detected from URL. */\n\trequiresAssistantAfterToolResult?: boolean;\n\t/** Whether thinking blocks must be converted to text blocks with <thinking> delimiters. Default: auto-detected from URL. */\n\trequiresThinkingAsText?: boolean;\n\t/** Whether tool call IDs must be normalized to Mistral format (exactly 9 alphanumeric chars). Default: auto-detected from URL. */\n\trequiresMistralToolIds?: boolean;\n}\n\n// Model interface for the unified model system\nexport interface Model<TApi extends Api> {\n\tid: string;\n\tname: string;\n\tapi: TApi;\n\tprovider: Provider;\n\tbaseUrl: string;\n\treasoning: boolean;\n\tinput: (\"text\" | \"image\")[];\n\tcost: {\n\t\tinput: number; // $/million tokens\n\t\toutput: number; // $/million tokens\n\t\tcacheRead: number; // $/million tokens\n\t\tcacheWrite: number; // $/million tokens\n\t};\n\tcontextWindow: number;\n\tmaxTokens: number;\n\theaders?: Record<string, string>;\n\t/** Compatibility overrides for openai-completions API. If not set, auto-detected from baseUrl. */\n\tcompat?: TApi extends \"openai-completions\" ? OpenAICompat : never;\n}\n"]}
+{"version":3,"file":"types.js","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAoCA,MAAM,WAAW,GAAqB,IAAI,CAAC","sourcesContent":["import type { AnthropicOptions } from \"./providers/anthropic.js\";\nimport type { GoogleOptions } from \"./providers/google.js\";\nimport type { GoogleGeminiCliOptions } from \"./providers/google-gemini-cli.js\";\nimport type { GoogleVertexOptions } from \"./providers/google-vertex.js\";\nimport type { OpenAICodexResponsesOptions } from \"./providers/openai-codex-responses.js\";\nimport type { OpenAICompletionsOptions } from \"./providers/openai-completions.js\";\nimport type { OpenAIResponsesOptions } from \"./providers/openai-responses.js\";\nimport type { AssistantMessageEventStream } from \"./utils/event-stream.js\";\n\nexport type { AssistantMessageEventStream } from \"./utils/event-stream.js\";\n\nexport type Api =\n\t| \"openai-completions\"\n\t| \"openai-responses\"\n\t| \"openai-codex-responses\"\n\t| \"anthropic-messages\"\n\t| \"google-generative-ai\"\n\t| \"google-gemini-cli\"\n\t| \"google-vertex\";\n\nexport interface ApiOptionsMap {\n\t\"anthropic-messages\": AnthropicOptions;\n\t\"openai-completions\": OpenAICompletionsOptions;\n\t\"openai-responses\": OpenAIResponsesOptions;\n\t\"openai-codex-responses\": OpenAICodexResponsesOptions;\n\t\"google-generative-ai\": GoogleOptions;\n\t\"google-gemini-cli\": GoogleGeminiCliOptions;\n\t\"google-vertex\": GoogleVertexOptions;\n}\n\n// Compile-time exhaustiveness check - this will fail if ApiOptionsMap doesn't have all KnownApi keys\ntype _CheckExhaustive = ApiOptionsMap extends Record<Api, StreamOptions>\n\t? Record<Api, StreamOptions> extends ApiOptionsMap\n\t\t? true\n\t\t: [\"ApiOptionsMap is missing some KnownApi values\", Exclude<Api, keyof ApiOptionsMap>]\n\t: [\"ApiOptionsMap doesn't extend Record<KnownApi, StreamOptions>\"];\nconst _exhaustive: _CheckExhaustive = true;\n\n// Helper type to get options for a specific API\nexport type OptionsForApi<TApi extends Api> = ApiOptionsMap[TApi];\n\nexport type KnownProvider =\n\t| \"anthropic\"\n\t| \"google\"\n\t| \"google-gemini-cli\"\n\t| \"google-antigravity\"\n\t| \"google-vertex\"\n\t| \"openai\"\n\t| \"openai-codex\"\n\t| \"github-copilot\"\n\t| \"xai\"\n\t| \"groq\"\n\t| \"cerebras\"\n\t| \"openrouter\"\n\t| \"zai\"\n\t| \"mistral\";\nexport type Provider = KnownProvider | string;\n\nexport type ThinkingLevel = \"minimal\" | \"low\" | \"medium\" | \"high\" | \"xhigh\";\n\n/** Token budgets for each thinking level (token-based providers only) */\nexport interface ThinkingBudgets {\n\tminimal?: number;\n\tlow?: number;\n\tmedium?: number;\n\thigh?: number;\n}\n\n// Base options all providers share\nexport interface StreamOptions {\n\ttemperature?: number;\n\tmaxTokens?: number;\n\tsignal?: AbortSignal;\n\tapiKey?: string;\n\t/**\n\t * Optional session identifier for providers that support session-based caching.\n\t * Providers can use this to enable prompt caching, request routing, or other\n\t * session-aware features. Ignored by providers that don't support it.\n\t */\n\tsessionId?: string;\n}\n\n// Unified options with reasoning passed to streamSimple() and completeSimple()\nexport interface SimpleStreamOptions extends StreamOptions {\n\treasoning?: ThinkingLevel;\n\t/** Custom token budgets for thinking levels (token-based providers only) */\n\tthinkingBudgets?: ThinkingBudgets;\n}\n\n// Generic StreamFunction with typed options\nexport type StreamFunction<TApi extends Api> = (\n\tmodel: Model<TApi>,\n\tcontext: Context,\n\toptions: OptionsForApi<TApi>,\n) => AssistantMessageEventStream;\n\nexport interface TextContent {\n\ttype: \"text\";\n\ttext: string;\n\ttextSignature?: string; // e.g., for OpenAI responses, the message ID\n}\n\nexport interface ThinkingContent {\n\ttype: \"thinking\";\n\tthinking: string;\n\tthinkingSignature?: string; // e.g., for OpenAI responses, the reasoning item ID\n}\n\nexport interface ImageContent {\n\ttype: \"image\";\n\tdata: string; // base64 encoded image data\n\tmimeType: string; // e.g., \"image/jpeg\", \"image/png\"\n}\n\nexport interface ToolCall {\n\ttype: \"toolCall\";\n\tid: string;\n\tname: string;\n\targuments: Record<string, any>;\n\tthoughtSignature?: string; // Google-specific: opaque signature for reusing thought context\n}\n\nexport interface Usage {\n\tinput: number;\n\toutput: number;\n\tcacheRead: number;\n\tcacheWrite: number;\n\ttotalTokens: number;\n\tcost: {\n\t\tinput: number;\n\t\toutput: number;\n\t\tcacheRead: number;\n\t\tcacheWrite: number;\n\t\ttotal: number;\n\t};\n}\n\nexport type StopReason = \"stop\" | \"length\" | \"toolUse\" | \"error\" | \"aborted\";\n\nexport interface UserMessage {\n\trole: \"user\";\n\tcontent: string | (TextContent | ImageContent)[];\n\ttimestamp: number; // Unix timestamp in milliseconds\n}\n\nexport interface AssistantMessage {\n\trole: \"assistant\";\n\tcontent: (TextContent | ThinkingContent | ToolCall)[];\n\tapi: Api;\n\tprovider: Provider;\n\tmodel: string;\n\tusage: Usage;\n\tstopReason: StopReason;\n\terrorMessage?: string;\n\ttimestamp: number; // Unix timestamp in milliseconds\n}\n\nexport interface ToolResultMessage<TDetails = any> {\n\trole: \"toolResult\";\n\ttoolCallId: string;\n\ttoolName: string;\n\tcontent: (TextContent | ImageContent)[]; // Supports text and images\n\tdetails?: TDetails;\n\tisError: boolean;\n\ttimestamp: number; // Unix timestamp in milliseconds\n}\n\nexport type Message = UserMessage | AssistantMessage | ToolResultMessage;\n\nimport type { TSchema } from \"@sinclair/typebox\";\n\nexport interface Tool<TParameters extends TSchema = TSchema> {\n\tname: string;\n\tdescription: string;\n\tparameters: TParameters;\n}\n\nexport interface Context {\n\tsystemPrompt?: string;\n\tmessages: Message[];\n\ttools?: Tool[];\n}\n\nexport type AssistantMessageEvent =\n\t| { type: \"start\"; partial: AssistantMessage }\n\t| { type: \"text_start\"; contentIndex: number; partial: AssistantMessage }\n\t| { type: \"text_delta\"; contentIndex: number; delta: string; partial: AssistantMessage }\n\t| { type: \"text_end\"; contentIndex: number; content: string; partial: AssistantMessage }\n\t| { type: \"thinking_start\"; contentIndex: number; partial: AssistantMessage }\n\t| { type: \"thinking_delta\"; contentIndex: number; delta: string; partial: AssistantMessage }\n\t| { type: \"thinking_end\"; contentIndex: number; content: string; partial: AssistantMessage }\n\t| { type: \"toolcall_start\"; contentIndex: number; partial: AssistantMessage }\n\t| { type: \"toolcall_delta\"; contentIndex: number; delta: string; partial: AssistantMessage }\n\t| { type: \"toolcall_end\"; contentIndex: number; toolCall: ToolCall; partial: AssistantMessage }\n\t| { type: \"done\"; reason: Extract<StopReason, \"stop\" | \"length\" | \"toolUse\">; message: AssistantMessage }\n\t| { type: \"error\"; reason: Extract<StopReason, \"aborted\" | \"error\">; error: AssistantMessage };\n\n/**\n * Compatibility settings for openai-completions API.\n * Use this to override URL-based auto-detection for custom providers.\n */\nexport interface OpenAICompat {\n\t/** Whether the provider supports the `store` field. Default: auto-detected from URL. */\n\tsupportsStore?: boolean;\n\t/** Whether the provider supports the `developer` role (vs `system`). Default: auto-detected from URL. */\n\tsupportsDeveloperRole?: boolean;\n\t/** Whether the provider supports `reasoning_effort`. Default: auto-detected from URL. */\n\tsupportsReasoningEffort?: boolean;\n\t/** Which field to use for max tokens. Default: auto-detected from URL. */\n\tmaxTokensField?: \"max_completion_tokens\" | \"max_tokens\";\n\t/** Whether tool results require the `name` field. Default: auto-detected from URL. */\n\trequiresToolResultName?: boolean;\n\t/** Whether a user message after tool results requires an assistant message in between. Default: auto-detected from URL. */\n\trequiresAssistantAfterToolResult?: boolean;\n\t/** Whether thinking blocks must be converted to text blocks with <thinking> delimiters. Default: auto-detected from URL. */\n\trequiresThinkingAsText?: boolean;\n\t/** Whether tool call IDs must be normalized to Mistral format (exactly 9 alphanumeric chars). Default: auto-detected from URL. */\n\trequiresMistralToolIds?: boolean;\n}\n\n// Model interface for the unified model system\nexport interface Model<TApi extends Api> {\n\tid: string;\n\tname: string;\n\tapi: TApi;\n\tprovider: Provider;\n\tbaseUrl: string;\n\treasoning: boolean;\n\tinput: (\"text\" | \"image\")[];\n\tcost: {\n\t\tinput: number; // $/million tokens\n\t\toutput: number; // $/million tokens\n\t\tcacheRead: number; // $/million tokens\n\t\tcacheWrite: number; // $/million tokens\n\t};\n\tcontextWindow: number;\n\tmaxTokens: number;\n\theaders?: Record<string, string>;\n\t/** Compatibility overrides for openai-completions API. If not set, auto-detected from baseUrl. */\n\tcompat?: TApi extends \"openai-completions\" ? OpenAICompat : never;\n}\n"]}

package/dist/utils/overflow.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"overflow.d.ts","sourceRoot":"","sources":["../../src/utils/overflow.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,aAAa,CAAC;AAuCpD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2CG;AACH,wBAAgB,iBAAiB,CAAC,OAAO,EAAE,gBAAgB,EAAE,aAAa,CAAC,EAAE,MAAM,GAAG,OAAO,CAwB5F;AAED;;GAEG;AACH,wBAAgB,mBAAmB,IAAI,MAAM,EAAE,CAE9C","sourcesContent":["import type { AssistantMessage } from \"../types.js\";\n\n/**\n * Regex patterns to detect context overflow errors from different providers.\n *\n * These patterns match error messages returned when the input exceeds\n * the model's context window.\n *\n * Provider-specific patterns (with example error messages):\n *\n * - Anthropic: \"prompt is too long: 213462 tokens > 200000 maximum\"\n * - OpenAI: \"Your input exceeds the context window of this model\"\n * - Google: \"The input token count (1196265) exceeds the maximum number of tokens allowed (1048575)\"\n * - xAI: \"This model's maximum prompt length is 131072 but the request contains 537812 tokens\"\n * - Groq: \"Please reduce the length of the messages or completion\"\n * - OpenRouter: \"This endpoint's maximum context length is X tokens. However, you requested about Y tokens\"\n * - llama.cpp: \"the request exceeds the available context size, try increasing it\"\n * - LM Studio: \"tokens to keep from the initial prompt is greater than the context length\"\n * - GitHub Copilot: \"prompt token count of X exceeds the limit of Y\"\n * - Cerebras: Returns \"400 status code (no body)\" - handled separately below\n * - Mistral: Returns \"400 status code (no body)\" - handled separately below\n * - z.ai: Does NOT error, accepts overflow silently - handled via usage.input > contextWindow\n * - Ollama: Silently truncates input - not detectable via error message\n */\nconst OVERFLOW_PATTERNS = [\n\t/prompt is too long/i, // Anthropic\n\t/exceeds the context window/i, // OpenAI (Completions & Responses API)\n\t/input token count.*exceeds the maximum/i, // Google (Gemini)\n\t/maximum prompt length is \\d+/i, // xAI (Grok)\n\t/reduce the length of the messages/i, // Groq\n\t/maximum context length is \\d+ tokens/i, // OpenRouter (all backends)\n\t/exceeds the limit of \\d+/i, // GitHub Copilot\n\t/exceeds the available context size/i, // llama.cpp server\n\t/greater than the context length/i, // LM Studio\n\t/context length exceeded/i, // Generic fallback\n\t/too many tokens/i, // Generic fallback\n\t/token limit exceeded/i, // Generic fallback\n];\n\n/**\n * Check if an assistant message represents a context overflow error.\n *\n * This handles two cases:\n * 1. Error-based overflow: Most providers return stopReason \"error\" with a\n *    specific error message pattern.\n * 2. Silent overflow: Some providers accept overflow requests and return\n *    successfully. For these, we check if usage.input exceeds the context window.\n *\n * ## Reliability by Provider\n *\n * **Reliable detection (returns error with detectable message):**\n * - Anthropic: \"prompt is too long: X tokens > Y maximum\"\n * - OpenAI (Completions & Responses): \"exceeds the context window\"\n * - Google Gemini: \"input token count exceeds the maximum\"\n * - xAI (Grok): \"maximum prompt length is X but request contains Y\"\n * - Groq: \"reduce the length of the messages\"\n * - Cerebras: 400/413/429 status code (no body)\n * - Mistral: 400/413/429 status code (no body)\n * - OpenRouter (all backends): \"maximum context length is X tokens\"\n * - llama.cpp: \"exceeds the available context size\"\n * - LM Studio: \"greater than the context length\"\n *\n * **Unreliable detection:**\n * - z.ai: Sometimes accepts overflow silently (detectable via usage.input > contextWindow),\n *   sometimes returns rate limit errors. Pass contextWindow param to detect silent overflow.\n * - Ollama: Silently truncates input without error. Cannot be detected via this function.\n *   The response will have usage.input < expected, but we don't know the expected value.\n *\n * ## Custom Providers\n *\n * If you've added custom models via settings.json, this function may not detect\n * overflow errors from those providers. To add support:\n *\n * 1. Send a request that exceeds the model's context window\n * 2. Check the errorMessage in the response\n * 3. Create a regex pattern that matches the error\n * 4. The pattern should be added to OVERFLOW_PATTERNS in this file, or\n *    check the errorMessage yourself before calling this function\n *\n * @param message - The assistant message to check\n * @param contextWindow - Optional context window size for detecting silent overflow (z.ai)\n * @returns true if the message indicates a context overflow\n */\nexport function isContextOverflow(message: AssistantMessage, contextWindow?: number): boolean {\n\t// Case 1: Check error message patterns\n\tif (message.stopReason === \"error\" && message.errorMessage) {\n\t\t// Check known patterns\n\t\tif (OVERFLOW_PATTERNS.some((p) => p.test(message.errorMessage!))) {\n\t\t\treturn true;\n\t\t}\n\n\t\t// Cerebras and Mistral return 400/413/429 with no body - check for status code pattern\n\t\t// 429 can indicate token-based rate limiting which correlates with context overflow\n\t\tif (/^4(00|13|29)\\s*(status code)?\\s*\\(no body\\)/i.test(message.errorMessage)) {\n\t\t\treturn true;\n\t\t}\n\t}\n\n\t// Case 2: Silent overflow (z.ai style) - successful but usage exceeds context\n\tif (contextWindow && message.stopReason === \"stop\") {\n\t\tconst inputTokens = message.usage.input + message.usage.cacheRead;\n\t\tif (inputTokens > contextWindow) {\n\t\t\treturn true;\n\t\t}\n\t}\n\n\treturn false;\n}\n\n/**\n * Get the overflow patterns for testing purposes.\n */\nexport function getOverflowPatterns(): RegExp[] {\n\treturn [...OVERFLOW_PATTERNS];\n}\n"]}
+{"version":3,"file":"overflow.d.ts","sourceRoot":"","sources":["../../src/utils/overflow.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,aAAa,CAAC;AAuCpD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2CG;AACH,wBAAgB,iBAAiB,CAAC,OAAO,EAAE,gBAAgB,EAAE,aAAa,CAAC,EAAE,MAAM,GAAG,OAAO,CAwB5F;AAED;;GAEG;AACH,wBAAgB,mBAAmB,IAAI,MAAM,EAAE,CAE9C","sourcesContent":["import type { AssistantMessage } from \"../types.js\";\n\n/**\n * Regex patterns to detect context overflow errors from different providers.\n *\n * These patterns match error messages returned when the input exceeds\n * the model's context window.\n *\n * Provider-specific patterns (with example error messages):\n *\n * - Anthropic: \"prompt is too long: 213462 tokens > 200000 maximum\"\n * - OpenAI: \"Your input exceeds the context window of this model\"\n * - Google: \"The input token count (1196265) exceeds the maximum number of tokens allowed (1048575)\"\n * - xAI: \"This model's maximum prompt length is 131072 but the request contains 537812 tokens\"\n * - Groq: \"Please reduce the length of the messages or completion\"\n * - OpenRouter: \"This endpoint's maximum context length is X tokens. However, you requested about Y tokens\"\n * - llama.cpp: \"the request exceeds the available context size, try increasing it\"\n * - LM Studio: \"tokens to keep from the initial prompt is greater than the context length\"\n * - GitHub Copilot: \"prompt token count of X exceeds the limit of Y\"\n * - Cerebras: Returns \"400 status code (no body)\" - handled separately below\n * - Mistral: Returns \"400 status code (no body)\" - handled separately below\n * - z.ai: Does NOT error, accepts overflow silently - handled via usage.input > contextWindow\n * - Ollama: Silently truncates input - not detectable via error message\n */\nconst OVERFLOW_PATTERNS = [\n\t/prompt is too long/i, // Anthropic\n\t/exceeds the context window/i, // OpenAI (Completions & Responses API)\n\t/input token count.*exceeds the maximum/i, // Google (Gemini)\n\t/maximum prompt length is \\d+/i, // xAI (Grok)\n\t/reduce the length of the messages/i, // Groq\n\t/maximum context length is \\d+ tokens/i, // OpenRouter (all backends)\n\t/exceeds the limit of \\d+/i, // GitHub Copilot\n\t/exceeds the available context size/i, // llama.cpp server\n\t/greater than the context length/i, // LM Studio\n\t/context[_ ]length[_ ]exceeded/i, // Generic fallback\n\t/too many tokens/i, // Generic fallback\n\t/token limit exceeded/i, // Generic fallback\n];\n\n/**\n * Check if an assistant message represents a context overflow error.\n *\n * This handles two cases:\n * 1. Error-based overflow: Most providers return stopReason \"error\" with a\n *    specific error message pattern.\n * 2. Silent overflow: Some providers accept overflow requests and return\n *    successfully. For these, we check if usage.input exceeds the context window.\n *\n * ## Reliability by Provider\n *\n * **Reliable detection (returns error with detectable message):**\n * - Anthropic: \"prompt is too long: X tokens > Y maximum\"\n * - OpenAI (Completions & Responses): \"exceeds the context window\"\n * - Google Gemini: \"input token count exceeds the maximum\"\n * - xAI (Grok): \"maximum prompt length is X but request contains Y\"\n * - Groq: \"reduce the length of the messages\"\n * - Cerebras: 400/413/429 status code (no body)\n * - Mistral: 400/413/429 status code (no body)\n * - OpenRouter (all backends): \"maximum context length is X tokens\"\n * - llama.cpp: \"exceeds the available context size\"\n * - LM Studio: \"greater than the context length\"\n *\n * **Unreliable detection:**\n * - z.ai: Sometimes accepts overflow silently (detectable via usage.input > contextWindow),\n *   sometimes returns rate limit errors. Pass contextWindow param to detect silent overflow.\n * - Ollama: Silently truncates input without error. Cannot be detected via this function.\n *   The response will have usage.input < expected, but we don't know the expected value.\n *\n * ## Custom Providers\n *\n * If you've added custom models via settings.json, this function may not detect\n * overflow errors from those providers. To add support:\n *\n * 1. Send a request that exceeds the model's context window\n * 2. Check the errorMessage in the response\n * 3. Create a regex pattern that matches the error\n * 4. The pattern should be added to OVERFLOW_PATTERNS in this file, or\n *    check the errorMessage yourself before calling this function\n *\n * @param message - The assistant message to check\n * @param contextWindow - Optional context window size for detecting silent overflow (z.ai)\n * @returns true if the message indicates a context overflow\n */\nexport function isContextOverflow(message: AssistantMessage, contextWindow?: number): boolean {\n\t// Case 1: Check error message patterns\n\tif (message.stopReason === \"error\" && message.errorMessage) {\n\t\t// Check known patterns\n\t\tif (OVERFLOW_PATTERNS.some((p) => p.test(message.errorMessage!))) {\n\t\t\treturn true;\n\t\t}\n\n\t\t// Cerebras and Mistral return 400/413/429 with no body - check for status code pattern\n\t\t// 429 can indicate token-based rate limiting which correlates with context overflow\n\t\tif (/^4(00|13|29)\\s*(status code)?\\s*\\(no body\\)/i.test(message.errorMessage)) {\n\t\t\treturn true;\n\t\t}\n\t}\n\n\t// Case 2: Silent overflow (z.ai style) - successful but usage exceeds context\n\tif (contextWindow && message.stopReason === \"stop\") {\n\t\tconst inputTokens = message.usage.input + message.usage.cacheRead;\n\t\tif (inputTokens > contextWindow) {\n\t\t\treturn true;\n\t\t}\n\t}\n\n\treturn false;\n}\n\n/**\n * Get the overflow patterns for testing purposes.\n */\nexport function getOverflowPatterns(): RegExp[] {\n\treturn [...OVERFLOW_PATTERNS];\n}\n"]}

package/dist/utils/overflow.js

@@ -30,7 +30,7 @@ const OVERFLOW_PATTERNS = [
     /exceeds the limit of \d+/i, // GitHub Copilot
     /exceeds the available context size/i, // llama.cpp server
     /greater than the context length/i, // LM Studio
-    /context length exceeded/i, // Generic fallback
+    /context[_ ]length[_ ]exceeded/i, // Generic fallback
     /too many tokens/i, // Generic fallback
     /token limit exceeded/i, // Generic fallback
 ];

package/dist/utils/overflow.js.map

@@ -1 +1 @@
-{"version":3,"file":"overflow.js","sourceRoot":"","sources":["../../src/utils/overflow.ts"],"names":[],"mappings":"AAEA;;;;;;;;;;;;;;;;;;;;;GAqBG;AACH,MAAM,iBAAiB,GAAG;IACzB,qBAAqB,EAAE,YAAY;IACnC,6BAA6B,EAAE,uCAAuC;IACtE,yCAAyC,EAAE,kBAAkB;IAC7D,+BAA+B,EAAE,aAAa;IAC9C,oCAAoC,EAAE,OAAO;IAC7C,uCAAuC,EAAE,4BAA4B;IACrE,2BAA2B,EAAE,iBAAiB;IAC9C,qCAAqC,EAAE,mBAAmB;IAC1D,kCAAkC,EAAE,YAAY;IAChD,0BAA0B,EAAE,mBAAmB;IAC/C,kBAAkB,EAAE,mBAAmB;IACvC,uBAAuB,EAAE,mBAAmB;CAC5C,CAAC;AAEF;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2CG;AACH,MAAM,UAAU,iBAAiB,CAAC,OAAyB,EAAE,aAAsB,EAAW;IAC7F,uCAAuC;IACvC,IAAI,OAAO,CAAC,UAAU,KAAK,OAAO,IAAI,OAAO,CAAC,YAAY,EAAE,CAAC;QAC5D,uBAAuB;QACvB,IAAI,iBAAiB,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,OAAO,CAAC,YAAa,CAAC,CAAC,EAAE,CAAC;YAClE,OAAO,IAAI,CAAC;QACb,CAAC;QAED,uFAAuF;QACvF,oFAAoF;QACpF,IAAI,8CAA8C,CAAC,IAAI,CAAC,OAAO,CAAC,YAAY,CAAC,EAAE,CAAC;YAC/E,OAAO,IAAI,CAAC;QACb,CAAC;IACF,CAAC;IAED,8EAA8E;IAC9E,IAAI,aAAa,IAAI,OAAO,CAAC,UAAU,KAAK,MAAM,EAAE,CAAC;QACpD,MAAM,WAAW,GAAG,OAAO,CAAC,KAAK,CAAC,KAAK,GAAG,OAAO,CAAC,KAAK,CAAC,SAAS,CAAC;QAClE,IAAI,WAAW,GAAG,aAAa,EAAE,CAAC;YACjC,OAAO,IAAI,CAAC;QACb,CAAC;IACF,CAAC;IAED,OAAO,KAAK,CAAC;AAAA,CACb;AAED;;GAEG;AACH,MAAM,UAAU,mBAAmB,GAAa;IAC/C,OAAO,CAAC,GAAG,iBAAiB,CAAC,CAAC;AAAA,CAC9B","sourcesContent":["import type { AssistantMessage } from \"../types.js\";\n\n/**\n * Regex patterns to detect context overflow errors from different providers.\n *\n * These patterns match error messages returned when the input exceeds\n * the model's context window.\n *\n * Provider-specific patterns (with example error messages):\n *\n * - Anthropic: \"prompt is too long: 213462 tokens > 200000 maximum\"\n * - OpenAI: \"Your input exceeds the context window of this model\"\n * - Google: \"The input token count (1196265) exceeds the maximum number of tokens allowed (1048575)\"\n * - xAI: \"This model's maximum prompt length is 131072 but the request contains 537812 tokens\"\n * - Groq: \"Please reduce the length of the messages or completion\"\n * - OpenRouter: \"This endpoint's maximum context length is X tokens. However, you requested about Y tokens\"\n * - llama.cpp: \"the request exceeds the available context size, try increasing it\"\n * - LM Studio: \"tokens to keep from the initial prompt is greater than the context length\"\n * - GitHub Copilot: \"prompt token count of X exceeds the limit of Y\"\n * - Cerebras: Returns \"400 status code (no body)\" - handled separately below\n * - Mistral: Returns \"400 status code (no body)\" - handled separately below\n * - z.ai: Does NOT error, accepts overflow silently - handled via usage.input > contextWindow\n * - Ollama: Silently truncates input - not detectable via error message\n */\nconst OVERFLOW_PATTERNS = [\n\t/prompt is too long/i, // Anthropic\n\t/exceeds the context window/i, // OpenAI (Completions & Responses API)\n\t/input token count.*exceeds the maximum/i, // Google (Gemini)\n\t/maximum prompt length is \\d+/i, // xAI (Grok)\n\t/reduce the length of the messages/i, // Groq\n\t/maximum context length is \\d+ tokens/i, // OpenRouter (all backends)\n\t/exceeds the limit of \\d+/i, // GitHub Copilot\n\t/exceeds the available context size/i, // llama.cpp server\n\t/greater than the context length/i, // LM Studio\n\t/context length exceeded/i, // Generic fallback\n\t/too many tokens/i, // Generic fallback\n\t/token limit exceeded/i, // Generic fallback\n];\n\n/**\n * Check if an assistant message represents a context overflow error.\n *\n * This handles two cases:\n * 1. Error-based overflow: Most providers return stopReason \"error\" with a\n *    specific error message pattern.\n * 2. Silent overflow: Some providers accept overflow requests and return\n *    successfully. For these, we check if usage.input exceeds the context window.\n *\n * ## Reliability by Provider\n *\n * **Reliable detection (returns error with detectable message):**\n * - Anthropic: \"prompt is too long: X tokens > Y maximum\"\n * - OpenAI (Completions & Responses): \"exceeds the context window\"\n * - Google Gemini: \"input token count exceeds the maximum\"\n * - xAI (Grok): \"maximum prompt length is X but request contains Y\"\n * - Groq: \"reduce the length of the messages\"\n * - Cerebras: 400/413/429 status code (no body)\n * - Mistral: 400/413/429 status code (no body)\n * - OpenRouter (all backends): \"maximum context length is X tokens\"\n * - llama.cpp: \"exceeds the available context size\"\n * - LM Studio: \"greater than the context length\"\n *\n * **Unreliable detection:**\n * - z.ai: Sometimes accepts overflow silently (detectable via usage.input > contextWindow),\n *   sometimes returns rate limit errors. Pass contextWindow param to detect silent overflow.\n * - Ollama: Silently truncates input without error. Cannot be detected via this function.\n *   The response will have usage.input < expected, but we don't know the expected value.\n *\n * ## Custom Providers\n *\n * If you've added custom models via settings.json, this function may not detect\n * overflow errors from those providers. To add support:\n *\n * 1. Send a request that exceeds the model's context window\n * 2. Check the errorMessage in the response\n * 3. Create a regex pattern that matches the error\n * 4. The pattern should be added to OVERFLOW_PATTERNS in this file, or\n *    check the errorMessage yourself before calling this function\n *\n * @param message - The assistant message to check\n * @param contextWindow - Optional context window size for detecting silent overflow (z.ai)\n * @returns true if the message indicates a context overflow\n */\nexport function isContextOverflow(message: AssistantMessage, contextWindow?: number): boolean {\n\t// Case 1: Check error message patterns\n\tif (message.stopReason === \"error\" && message.errorMessage) {\n\t\t// Check known patterns\n\t\tif (OVERFLOW_PATTERNS.some((p) => p.test(message.errorMessage!))) {\n\t\t\treturn true;\n\t\t}\n\n\t\t// Cerebras and Mistral return 400/413/429 with no body - check for status code pattern\n\t\t// 429 can indicate token-based rate limiting which correlates with context overflow\n\t\tif (/^4(00|13|29)\\s*(status code)?\\s*\\(no body\\)/i.test(message.errorMessage)) {\n\t\t\treturn true;\n\t\t}\n\t}\n\n\t// Case 2: Silent overflow (z.ai style) - successful but usage exceeds context\n\tif (contextWindow && message.stopReason === \"stop\") {\n\t\tconst inputTokens = message.usage.input + message.usage.cacheRead;\n\t\tif (inputTokens > contextWindow) {\n\t\t\treturn true;\n\t\t}\n\t}\n\n\treturn false;\n}\n\n/**\n * Get the overflow patterns for testing purposes.\n */\nexport function getOverflowPatterns(): RegExp[] {\n\treturn [...OVERFLOW_PATTERNS];\n}\n"]}
+{"version":3,"file":"overflow.js","sourceRoot":"","sources":["../../src/utils/overflow.ts"],"names":[],"mappings":"AAEA;;;;;;;;;;;;;;;;;;;;;GAqBG;AACH,MAAM,iBAAiB,GAAG;IACzB,qBAAqB,EAAE,YAAY;IACnC,6BAA6B,EAAE,uCAAuC;IACtE,yCAAyC,EAAE,kBAAkB;IAC7D,+BAA+B,EAAE,aAAa;IAC9C,oCAAoC,EAAE,OAAO;IAC7C,uCAAuC,EAAE,4BAA4B;IACrE,2BAA2B,EAAE,iBAAiB;IAC9C,qCAAqC,EAAE,mBAAmB;IAC1D,kCAAkC,EAAE,YAAY;IAChD,gCAAgC,EAAE,mBAAmB;IACrD,kBAAkB,EAAE,mBAAmB;IACvC,uBAAuB,EAAE,mBAAmB;CAC5C,CAAC;AAEF;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2CG;AACH,MAAM,UAAU,iBAAiB,CAAC,OAAyB,EAAE,aAAsB,EAAW;IAC7F,uCAAuC;IACvC,IAAI,OAAO,CAAC,UAAU,KAAK,OAAO,IAAI,OAAO,CAAC,YAAY,EAAE,CAAC;QAC5D,uBAAuB;QACvB,IAAI,iBAAiB,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,OAAO,CAAC,YAAa,CAAC,CAAC,EAAE,CAAC;YAClE,OAAO,IAAI,CAAC;QACb,CAAC;QAED,uFAAuF;QACvF,oFAAoF;QACpF,IAAI,8CAA8C,CAAC,IAAI,CAAC,OAAO,CAAC,YAAY,CAAC,EAAE,CAAC;YAC/E,OAAO,IAAI,CAAC;QACb,CAAC;IACF,CAAC;IAED,8EAA8E;IAC9E,IAAI,aAAa,IAAI,OAAO,CAAC,UAAU,KAAK,MAAM,EAAE,CAAC;QACpD,MAAM,WAAW,GAAG,OAAO,CAAC,KAAK,CAAC,KAAK,GAAG,OAAO,CAAC,KAAK,CAAC,SAAS,CAAC;QAClE,IAAI,WAAW,GAAG,aAAa,EAAE,CAAC;YACjC,OAAO,IAAI,CAAC;QACb,CAAC;IACF,CAAC;IAED,OAAO,KAAK,CAAC;AAAA,CACb;AAED;;GAEG;AACH,MAAM,UAAU,mBAAmB,GAAa;IAC/C,OAAO,CAAC,GAAG,iBAAiB,CAAC,CAAC;AAAA,CAC9B","sourcesContent":["import type { AssistantMessage } from \"../types.js\";\n\n/**\n * Regex patterns to detect context overflow errors from different providers.\n *\n * These patterns match error messages returned when the input exceeds\n * the model's context window.\n *\n * Provider-specific patterns (with example error messages):\n *\n * - Anthropic: \"prompt is too long: 213462 tokens > 200000 maximum\"\n * - OpenAI: \"Your input exceeds the context window of this model\"\n * - Google: \"The input token count (1196265) exceeds the maximum number of tokens allowed (1048575)\"\n * - xAI: \"This model's maximum prompt length is 131072 but the request contains 537812 tokens\"\n * - Groq: \"Please reduce the length of the messages or completion\"\n * - OpenRouter: \"This endpoint's maximum context length is X tokens. However, you requested about Y tokens\"\n * - llama.cpp: \"the request exceeds the available context size, try increasing it\"\n * - LM Studio: \"tokens to keep from the initial prompt is greater than the context length\"\n * - GitHub Copilot: \"prompt token count of X exceeds the limit of Y\"\n * - Cerebras: Returns \"400 status code (no body)\" - handled separately below\n * - Mistral: Returns \"400 status code (no body)\" - handled separately below\n * - z.ai: Does NOT error, accepts overflow silently - handled via usage.input > contextWindow\n * - Ollama: Silently truncates input - not detectable via error message\n */\nconst OVERFLOW_PATTERNS = [\n\t/prompt is too long/i, // Anthropic\n\t/exceeds the context window/i, // OpenAI (Completions & Responses API)\n\t/input token count.*exceeds the maximum/i, // Google (Gemini)\n\t/maximum prompt length is \\d+/i, // xAI (Grok)\n\t/reduce the length of the messages/i, // Groq\n\t/maximum context length is \\d+ tokens/i, // OpenRouter (all backends)\n\t/exceeds the limit of \\d+/i, // GitHub Copilot\n\t/exceeds the available context size/i, // llama.cpp server\n\t/greater than the context length/i, // LM Studio\n\t/context[_ ]length[_ ]exceeded/i, // Generic fallback\n\t/too many tokens/i, // Generic fallback\n\t/token limit exceeded/i, // Generic fallback\n];\n\n/**\n * Check if an assistant message represents a context overflow error.\n *\n * This handles two cases:\n * 1. Error-based overflow: Most providers return stopReason \"error\" with a\n *    specific error message pattern.\n * 2. Silent overflow: Some providers accept overflow requests and return\n *    successfully. For these, we check if usage.input exceeds the context window.\n *\n * ## Reliability by Provider\n *\n * **Reliable detection (returns error with detectable message):**\n * - Anthropic: \"prompt is too long: X tokens > Y maximum\"\n * - OpenAI (Completions & Responses): \"exceeds the context window\"\n * - Google Gemini: \"input token count exceeds the maximum\"\n * - xAI (Grok): \"maximum prompt length is X but request contains Y\"\n * - Groq: \"reduce the length of the messages\"\n * - Cerebras: 400/413/429 status code (no body)\n * - Mistral: 400/413/429 status code (no body)\n * - OpenRouter (all backends): \"maximum context length is X tokens\"\n * - llama.cpp: \"exceeds the available context size\"\n * - LM Studio: \"greater than the context length\"\n *\n * **Unreliable detection:**\n * - z.ai: Sometimes accepts overflow silently (detectable via usage.input > contextWindow),\n *   sometimes returns rate limit errors. Pass contextWindow param to detect silent overflow.\n * - Ollama: Silently truncates input without error. Cannot be detected via this function.\n *   The response will have usage.input < expected, but we don't know the expected value.\n *\n * ## Custom Providers\n *\n * If you've added custom models via settings.json, this function may not detect\n * overflow errors from those providers. To add support:\n *\n * 1. Send a request that exceeds the model's context window\n * 2. Check the errorMessage in the response\n * 3. Create a regex pattern that matches the error\n * 4. The pattern should be added to OVERFLOW_PATTERNS in this file, or\n *    check the errorMessage yourself before calling this function\n *\n * @param message - The assistant message to check\n * @param contextWindow - Optional context window size for detecting silent overflow (z.ai)\n * @returns true if the message indicates a context overflow\n */\nexport function isContextOverflow(message: AssistantMessage, contextWindow?: number): boolean {\n\t// Case 1: Check error message patterns\n\tif (message.stopReason === \"error\" && message.errorMessage) {\n\t\t// Check known patterns\n\t\tif (OVERFLOW_PATTERNS.some((p) => p.test(message.errorMessage!))) {\n\t\t\treturn true;\n\t\t}\n\n\t\t// Cerebras and Mistral return 400/413/429 with no body - check for status code pattern\n\t\t// 429 can indicate token-based rate limiting which correlates with context overflow\n\t\tif (/^4(00|13|29)\\s*(status code)?\\s*\\(no body\\)/i.test(message.errorMessage)) {\n\t\t\treturn true;\n\t\t}\n\t}\n\n\t// Case 2: Silent overflow (z.ai style) - successful but usage exceeds context\n\tif (contextWindow && message.stopReason === \"stop\") {\n\t\tconst inputTokens = message.usage.input + message.usage.cacheRead;\n\t\tif (inputTokens > contextWindow) {\n\t\t\treturn true;\n\t\t}\n\t}\n\n\treturn false;\n}\n\n/**\n * Get the overflow patterns for testing purposes.\n */\nexport function getOverflowPatterns(): RegExp[] {\n\treturn [...OVERFLOW_PATTERNS];\n}\n"]}

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@mariozechner/pi-ai",
-	"version": "0.37.7",
+	"version": "0.38.0",
 	"description": "Unified LLM API with automatic model discovery and provider configuration",
 	"type": "module",
 	"main": "./dist/index.js",