@fleetagent/pi-ai 0.1.4 → 0.1.7

package/dist/utils/oauth/github-copilot.js.map

@@ -1 +1 @@
-{"version":3,"file":"github-copilot.js","sourceRoot":"","sources":["../../../src/utils/oauth/github-copilot.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,OAAO,EAAE,SAAS,EAAE,MAAM,iBAAiB,CAAC;AAE5C,OAAO,EAAE,uBAAuB,EAAE,MAAM,kBAAkB,CAAC;AAO3D,MAAM,MAAM,GAAG,CAAC,CAAS,EAAE,EAAE,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;AACtC,MAAM,SAAS,GAAG,MAAM,CAAC,8BAA8B,CAAC,CAAC;AAEzD,MAAM,eAAe,GAAG;IACvB,YAAY,EAAE,0BAA0B;IACxC,gBAAgB,EAAE,gBAAgB;IAClC,uBAAuB,EAAE,qBAAqB;IAC9C,wBAAwB,EAAE,aAAa;CAC9B,CAAC;AAqBX,MAAM,UAAU,eAAe,CAAC,KAAa,EAAiB;IAC7D,MAAM,OAAO,GAAG,KAAK,CAAC,IAAI,EAAE,CAAC;IAC7B,IAAI,CAAC,OAAO;QAAE,OAAO,IAAI,CAAC;IAC1B,IAAI,CAAC;QACJ,MAAM,GAAG,GAAG,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,IAAI,GAAG,CAAC,WAAW,OAAO,EAAE,CAAC,CAAC;QACvF,OAAO,GAAG,CAAC,QAAQ,CAAC;IACrB,CAAC;IAAC,MAAM,CAAC;QACR,OAAO,IAAI,CAAC;IACb,CAAC;AAAA,CACD;AAED,SAAS,OAAO,CAAC,MAAc,EAI7B;IACD,OAAO;QACN,aAAa,EAAE,WAAW,MAAM,oBAAoB;QACpD,cAAc,EAAE,WAAW,MAAM,2BAA2B;QAC5D,eAAe,EAAE,eAAe,MAAM,4BAA4B;KAClE,CAAC;AAAA,CACF;AAED;;;;GAIG;AACH,SAAS,mBAAmB,CAAC,KAAa,EAAiB;IAC1D,MAAM,KAAK,GAAG,KAAK,CAAC,KAAK,CAAC,kBAAkB,CAAC,CAAC;IAC9C,IAAI,CAAC,KAAK;QAAE,OAAO,IAAI,CAAC;IACxB,MAAM,SAAS,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC;IAC3B,+BAA+B;IAC/B,MAAM,OAAO,GAAG,SAAS,CAAC,OAAO,CAAC,UAAU,EAAE,MAAM,CAAC,CAAC;IACtD,OAAO,WAAW,OAAO,EAAE,CAAC;AAAA,CAC5B;AAED,MAAM,UAAU,uBAAuB,CAAC,KAAc,EAAE,gBAAyB,EAAU;IAC1F,yDAAyD;IACzD,IAAI,KAAK,EAAE,CAAC;QACX,MAAM,YAAY,GAAG,mBAAmB,CAAC,KAAK,CAAC,CAAC;QAChD,IAAI,YAAY;YAAE,OAAO,YAAY,CAAC;IACvC,CAAC;IACD,oDAAoD;IACpD,IAAI,gBAAgB;QAAE,OAAO,uBAAuB,gBAAgB,EAAE,CAAC;IACvE,OAAO,0CAA0C,CAAC;AAAA,CAClD;AAED,KAAK,UAAU,SAAS,CAAC,GAAW,EAAE,IAAiB,EAAoB;IAC1E,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,GAAG,EAAE,IAAI,CAAC,CAAC;IACxC,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;QAClB,MAAM,IAAI,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;QACnC,MAAM,IAAI,KAAK,CAAC,GAAG,QAAQ,CAAC,MAAM,IAAI,QAAQ,CAAC,UAAU,KAAK,IAAI,EAAE,CAAC,CAAC;IACvE,CAAC;IACD,OAAO,QAAQ,CAAC,IAAI,EAAE,CAAC;AAAA,CACvB;AAED,KAAK,UAAU,eAAe,CAAC,MAAc,EAA+B;IAC3E,MAAM,IAAI,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC;IAC7B,MAAM,IAAI,GAAG,MAAM,SAAS,CAAC,IAAI,CAAC,aAAa,EAAE;QAChD,MAAM,EAAE,MAAM;QACd,OAAO,EAAE;YACR,MAAM,EAAE,kBAAkB;YAC1B,cAAc,EAAE,mCAAmC;YACnD,YAAY,EAAE,0BAA0B;SACxC;QACD,IAAI,EAAE,IAAI,eAAe,CAAC;YACzB,SAAS,EAAE,SAAS;YACpB,KAAK,EAAE,WAAW;SAClB,CAAC;KACF,CAAC,CAAC;IAEH,IAAI,CAAC,IAAI,IAAI,OAAO,IAAI,KAAK,QAAQ,EAAE,CAAC;QACvC,MAAM,IAAI,KAAK,CAAC,8BAA8B,CAAC,CAAC;IACjD,CAAC;IAED,MAAM,UAAU,GAAI,IAAgC,CAAC,WAAW,CAAC;IACjE,MAAM,QAAQ,GAAI,IAAgC,CAAC,SAAS,CAAC;IAC7D,MAAM,eAAe,GAAI,IAAgC,CAAC,gBAAgB,CAAC;IAC3E,MAAM,QAAQ,GAAI,IAAgC,CAAC,QAAQ,CAAC;IAC5D,MAAM,SAAS,GAAI,IAAgC,CAAC,UAAU,CAAC;IAE/D,IACC,OAAO,UAAU,KAAK,QAAQ;QAC9B,OAAO,QAAQ,KAAK,QAAQ;QAC5B,OAAO,eAAe,KAAK,QAAQ;QACnC,CAAC,QAAQ,KAAK,SAAS,IAAI,OAAO,QAAQ,KAAK,QAAQ,CAAC;QACxD,OAAO,SAAS,KAAK,QAAQ,EAC5B,CAAC;QACF,MAAM,IAAI,KAAK,CAAC,qCAAqC,CAAC,CAAC;IACxD,CAAC;IAED,OAAO;QACN,WAAW,EAAE,UAAU;QACvB,SAAS,EAAE,QAAQ;QACnB,gBAAgB,EAAE,eAAe;QACjC,QAAQ;QACR,UAAU,EAAE,SAAS;KACrB,CAAC;AAAA,CACF;AAED,KAAK,UAAU,wBAAwB,CAAC,MAAc,EAAE,MAA0B,EAAE,MAAoB,EAAE;IACzG,MAAM,IAAI,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC;IAC7B,OAAO,uBAAuB,CAAC;QAC9B,eAAe,EAAE,MAAM,CAAC,QAAQ;QAChC,gBAAgB,EAAE,MAAM,CAAC,UAAU;QACnC,MAAM;QACN,IAAI,EAAE,KAAK,IAAI,EAAE,CAAC;YACjB,MAAM,GAAG,GAAG,MAAM,SAAS,CAAC,IAAI,CAAC,cAAc,EAAE;gBAChD,MAAM,EAAE,MAAM;gBACd,OAAO,EAAE;oBACR,MAAM,EAAE,kBAAkB;oBAC1B,cAAc,EAAE,mCAAmC;oBACnD,YAAY,EAAE,0BAA0B;iBACxC;gBACD,IAAI,EAAE,IAAI,eAAe,CAAC;oBACzB,SAAS,EAAE,SAAS;oBACpB,WAAW,EAAE,MAAM,CAAC,WAAW;oBAC/B,UAAU,EAAE,8CAA8C;iBAC1D,CAAC;aACF,CAAC,CAAC;YAEH,IAAI,GAAG,IAAI,OAAO,GAAG,KAAK,QAAQ,IAAI,OAAQ,GAAkC,CAAC,YAAY,KAAK,QAAQ,EAAE,CAAC;gBAC5G,OAAO,EAAE,MAAM,EAAE,UAAU,EAAE,WAAW,EAAG,GAAkC,CAAC,YAAY,EAAE,CAAC;YAC9F,CAAC;YAED,IAAI,GAAG,IAAI,OAAO,GAAG,KAAK,QAAQ,IAAI,OAAQ,GAAgC,CAAC,KAAK,KAAK,QAAQ,EAAE,CAAC;gBACnG,MAAM,EAAE,KAAK,EAAE,iBAAiB,EAAE,WAAW,EAAE,GAAG,GAA+B,CAAC;gBAClF,IAAI,KAAK,KAAK,uBAAuB,EAAE,CAAC;oBACvC,OAAO,EAAE,MAAM,EAAE,SAAS,EAAE,CAAC;gBAC9B,CAAC;gBAED,IAAI,KAAK,KAAK,WAAW,EAAE,CAAC;oBAC3B,OAAO,EAAE,MAAM,EAAE,WAAW,EAAE,CAAC;gBAChC,CAAC;gBAED,MAAM,iBAAiB,GAAG,WAAW,CAAC,CAAC,CAAC,KAAK,WAAW,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;gBAChE,OAAO,EAAE,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,uBAAuB,KAAK,GAAG,iBAAiB,EAAE,EAAE,CAAC;YAC1F,CAAC;YAED,OAAO,EAAE,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,+BAA+B,EAAE,CAAC;QAAA,CACtE;KACD,CAAC,CAAC;AAAA,CACH;AAED;;GAEG;AACH,MAAM,CAAC,KAAK,UAAU,yBAAyB,CAC9C,YAAoB,EACpB,gBAAyB,EACG;IAC5B,MAAM,MAAM,GAAG,gBAAgB,IAAI,YAAY,CAAC;IAChD,MAAM,IAAI,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC;IAE7B,MAAM,GAAG,GAAG,MAAM,SAAS,CAAC,IAAI,CAAC,eAAe,EAAE;QACjD,OAAO,EAAE;YACR,MAAM,EAAE,kBAAkB;YAC1B,aAAa,EAAE,UAAU,YAAY,EAAE;YACvC,GAAG,eAAe;SAClB;KACD,CAAC,CAAC;IAEH,IAAI,CAAC,GAAG,IAAI,OAAO,GAAG,KAAK,QAAQ,EAAE,CAAC;QACrC,MAAM,IAAI,KAAK,CAAC,gCAAgC,CAAC,CAAC;IACnD,CAAC;IAED,MAAM,KAAK,GAAI,GAA+B,CAAC,KAAK,CAAC;IACrD,MAAM,SAAS,GAAI,GAA+B,CAAC,UAAU,CAAC;IAE9D,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,OAAO,SAAS,KAAK,QAAQ,EAAE,CAAC;QAChE,MAAM,IAAI,KAAK,CAAC,uCAAuC,CAAC,CAAC;IAC1D,CAAC;IAED,OAAO;QACN,OAAO,EAAE,YAAY;QACrB,MAAM,EAAE,KAAK;QACb,OAAO,EAAE,SAAS,GAAG,IAAI,GAAG,CAAC,GAAG,EAAE,GAAG,IAAI;QACzC,aAAa,EAAE,gBAAgB;KAC/B,CAAC;AAAA,CACF;AAED;;;GAGG;AACH,KAAK,UAAU,wBAAwB,CAAC,KAAa,EAAE,OAAe,EAAE,gBAAyB,EAAoB;IACpH,MAAM,OAAO,GAAG,uBAAuB,CAAC,KAAK,EAAE,gBAAgB,CAAC,CAAC;IACjE,MAAM,GAAG,GAAG,GAAG,OAAO,WAAW,OAAO,SAAS,CAAC;IAElD,IAAI,CAAC;QACJ,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,GAAG,EAAE;YACjC,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACR,cAAc,EAAE,kBAAkB;gBAClC,aAAa,EAAE,UAAU,KAAK,EAAE;gBAChC,GAAG,eAAe;gBAClB,eAAe,EAAE,aAAa;gBAC9B,oBAAoB,EAAE,aAAa;aACnC;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,EAAE,KAAK,EAAE,SAAS,EAAE,CAAC;SAC1C,CAAC,CAAC;QACH,OAAO,QAAQ,CAAC,EAAE,CAAC;IACpB,CAAC;IAAC,MAAM,CAAC;QACR,OAAO,KAAK,CAAC;IACd,CAAC;AAAA,CACD;AAED;;;GAGG;AACH,KAAK,UAAU,4BAA4B,CAC1C,KAAa,EACb,gBAAyB,EACzB,UAAsD,EACtC;IAChB,MAAM,MAAM,GAAG,SAAS,CAAC,gBAAgB,CAAC,CAAC;IAC3C,MAAM,OAAO,CAAC,GAAG,CAChB,MAAM,CAAC,GAAG,CAAC,KAAK,EAAE,KAAK,EAAE,EAAE,CAAC;QAC3B,MAAM,OAAO,GAAG,MAAM,wBAAwB,CAAC,KAAK,EAAE,KAAK,CAAC,EAAE,EAAE,gBAAgB,CAAC,CAAC;QAClF,UAAU,EAAE,CAAC,KAAK,CAAC,EAAE,EAAE,OAAO,CAAC,CAAC;IAAA,CAChC,CAAC,CACF,CAAC;AAAA,CACF;AAED;;;;;;;GAOG;AACH,MAAM,CAAC,KAAK,UAAU,kBAAkB,CAAC,OAKxC,EAA6B;IAC7B,MAAM,KAAK,GAAG,MAAM,OAAO,CAAC,QAAQ,CAAC;QACpC,OAAO,EAAE,qDAAqD;QAC9D,WAAW,EAAE,iBAAiB;QAC9B,UAAU,EAAE,IAAI;KAChB,CAAC,CAAC;IAEH,IAAI,OAAO,CAAC,MAAM,EAAE,OAAO,EAAE,CAAC;QAC7B,MAAM,IAAI,KAAK,CAAC,iBAAiB,CAAC,CAAC;IACpC,CAAC;IAED,MAAM,OAAO,GAAG,KAAK,CAAC,IAAI,EAAE,CAAC;IAC7B,MAAM,gBAAgB,GAAG,eAAe,CAAC,KAAK,CAAC,CAAC;IAChD,IAAI,OAAO,IAAI,CAAC,gBAAgB,EAAE,CAAC;QAClC,MAAM,IAAI,KAAK,CAAC,sCAAsC,CAAC,CAAC;IACzD,CAAC;IACD,MAAM,MAAM,GAAG,gBAAgB,IAAI,YAAY,CAAC;IAEhD,MAAM,MAAM,GAAG,MAAM,eAAe,CAAC,MAAM,CAAC,CAAC;IAC7C,OAAO,CAAC,YAAY,CAAC;QACpB,QAAQ,EAAE,MAAM,CAAC,SAAS;QAC1B,eAAe,EAAE,MAAM,CAAC,gBAAgB;QACxC,eAAe,EAAE,MAAM,CAAC,QAAQ;QAChC,gBAAgB,EAAE,MAAM,CAAC,UAAU;KACnC,CAAC,CAAC;IAEH,MAAM,iBAAiB,GAAG,MAAM,wBAAwB,CAAC,MAAM,EAAE,MAAM,EAAE,OAAO,CAAC,MAAM,CAAC,CAAC;IACzF,MAAM,WAAW,GAAG,MAAM,yBAAyB,CAAC,iBAAiB,EAAE,gBAAgB,IAAI,SAAS,CAAC,CAAC;IAEtG,2CAA2C;IAC3C,OAAO,CAAC,UAAU,EAAE,CAAC,oBAAoB,CAAC,CAAC;IAC3C,MAAM,4BAA4B,CAAC,WAAW,CAAC,MAAM,EAAE,gBAAgB,IAAI,SAAS,CAAC,CAAC;IACtF,OAAO,WAAW,CAAC;AAAA,CACnB;AAED,MAAM,CAAC,MAAM,0BAA0B,GAA2B;IACjE,EAAE,EAAE,gBAAgB;IACpB,IAAI,EAAE,gBAAgB;IAEtB,KAAK,CAAC,KAAK,CAAC,SAA8B,EAA6B;QACtE,IAAI,CAAC,SAAS,CAAC,YAAY,EAAE,CAAC;YAC7B,MAAM,IAAI,KAAK,CAAC,sDAAsD,CAAC,CAAC;QACzE,CAAC;QAED,OAAO,kBAAkB,CAAC;YACzB,YAAY,EAAE,SAAS,CAAC,YAAY;YACpC,QAAQ,EAAE,SAAS,CAAC,QAAQ;YAC5B,UAAU,EAAE,SAAS,CAAC,UAAU;YAChC,MAAM,EAAE,SAAS,CAAC,MAAM;SACxB,CAAC,CAAC;IAAA,CACH;IAED,KAAK,CAAC,YAAY,CAAC,WAA6B,EAA6B;QAC5E,MAAM,KAAK,GAAG,WAAiC,CAAC;QAChD,OAAO,yBAAyB,CAAC,KAAK,CAAC,OAAO,EAAE,KAAK,CAAC,aAAa,CAAC,CAAC;IAAA,CACrE;IAED,SAAS,CAAC,WAA6B,EAAU;QAChD,OAAO,WAAW,CAAC,MAAM,CAAC;IAAA,CAC1B;IAED,YAAY,CAAC,MAAoB,EAAE,WAA6B,EAAgB;QAC/E,MAAM,KAAK,GAAG,WAAiC,CAAC;QAChD,MAAM,MAAM,GAAG,KAAK,CAAC,aAAa,CAAC,CAAC,CAAC,CAAC,eAAe,CAAC,KAAK,CAAC,aAAa,CAAC,IAAI,SAAS,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;QACrG,MAAM,OAAO,GAAG,uBAAuB,CAAC,KAAK,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;QAC9D,OAAO,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,QAAQ,KAAK,gBAAgB,CAAC,CAAC,CAAC,EAAE,GAAG,CAAC,EAAE,OAAO,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;IAAA,CACpF;CACD,CAAC","sourcesContent":["/**\n * GitHub Copilot OAuth flow\n */\n\nimport { getModels } from \"../../models.ts\";\nimport type { Api, Model } from \"../../types.ts\";\nimport { pollOAuthDeviceCodeFlow } from \"./device-code.ts\";\nimport type { OAuthCredentials, OAuthDeviceCodeInfo, OAuthLoginCallbacks, OAuthProviderInterface } from \"./types.ts\";\n\ntype CopilotCredentials = OAuthCredentials & {\n\tenterpriseUrl?: string;\n};\n\nconst decode = (s: string) => atob(s);\nconst CLIENT_ID = decode(\"SXYxLmI1MDdhMDhjODdlY2ZlOTg=\");\n\nconst COPILOT_HEADERS = {\n\t\"User-Agent\": \"GitHubCopilotChat/0.35.0\",\n\t\"Editor-Version\": \"vscode/1.107.0\",\n\t\"Editor-Plugin-Version\": \"copilot-chat/0.35.0\",\n\t\"Copilot-Integration-Id\": \"vscode-chat\",\n} as const;\n\ntype DeviceCodeResponse = {\n\tdevice_code: string;\n\tuser_code: string;\n\tverification_uri: string;\n\tinterval?: number;\n\texpires_in: number;\n};\n\ntype DeviceTokenSuccessResponse = {\n\taccess_token: string;\n\ttoken_type?: string;\n\tscope?: string;\n};\n\ntype DeviceTokenErrorResponse = {\n\terror: string;\n\terror_description?: string;\n};\n\nexport function normalizeDomain(input: string): string | null {\n\tconst trimmed = input.trim();\n\tif (!trimmed) return null;\n\ttry {\n\t\tconst url = trimmed.includes(\"://\") ? new URL(trimmed) : new URL(`https://${trimmed}`);\n\t\treturn url.hostname;\n\t} catch {\n\t\treturn null;\n\t}\n}\n\nfunction getUrls(domain: string): {\n\tdeviceCodeUrl: string;\n\taccessTokenUrl: string;\n\tcopilotTokenUrl: string;\n} {\n\treturn {\n\t\tdeviceCodeUrl: `https://${domain}/login/device/code`,\n\t\taccessTokenUrl: `https://${domain}/login/oauth/access_token`,\n\t\tcopilotTokenUrl: `https://api.${domain}/copilot_internal/v2/token`,\n\t};\n}\n\n/**\n * Parse the proxy-ep from a Copilot token and convert to API base URL.\n * Token format: tid=...;exp=...;proxy-ep=proxy.individual.githubcopilot.com;...\n * Returns API URL like https://api.individual.githubcopilot.com\n */\nfunction getBaseUrlFromToken(token: string): string | null {\n\tconst match = token.match(/proxy-ep=([^;]+)/);\n\tif (!match) return null;\n\tconst proxyHost = match[1];\n\t// Convert proxy.xxx to api.xxx\n\tconst apiHost = proxyHost.replace(/^proxy\\./, \"api.\");\n\treturn `https://${apiHost}`;\n}\n\nexport function getGitHubCopilotBaseUrl(token?: string, enterpriseDomain?: string): string {\n\t// If we have a token, extract the base URL from proxy-ep\n\tif (token) {\n\t\tconst urlFromToken = getBaseUrlFromToken(token);\n\t\tif (urlFromToken) return urlFromToken;\n\t}\n\t// Fallback for enterprise or if token parsing fails\n\tif (enterpriseDomain) return `https://copilot-api.${enterpriseDomain}`;\n\treturn \"https://api.individual.githubcopilot.com\";\n}\n\nasync function fetchJson(url: string, init: RequestInit): Promise<unknown> {\n\tconst response = await fetch(url, init);\n\tif (!response.ok) {\n\t\tconst text = await response.text();\n\t\tthrow new Error(`${response.status} ${response.statusText}: ${text}`);\n\t}\n\treturn response.json();\n}\n\nasync function startDeviceFlow(domain: string): Promise<DeviceCodeResponse> {\n\tconst urls = getUrls(domain);\n\tconst data = await fetchJson(urls.deviceCodeUrl, {\n\t\tmethod: \"POST\",\n\t\theaders: {\n\t\t\tAccept: \"application/json\",\n\t\t\t\"Content-Type\": \"application/x-www-form-urlencoded\",\n\t\t\t\"User-Agent\": \"GitHubCopilotChat/0.35.0\",\n\t\t},\n\t\tbody: new URLSearchParams({\n\t\t\tclient_id: CLIENT_ID,\n\t\t\tscope: \"read:user\",\n\t\t}),\n\t});\n\n\tif (!data || typeof data !== \"object\") {\n\t\tthrow new Error(\"Invalid device code response\");\n\t}\n\n\tconst deviceCode = (data as Record<string, unknown>).device_code;\n\tconst userCode = (data as Record<string, unknown>).user_code;\n\tconst verificationUri = (data as Record<string, unknown>).verification_uri;\n\tconst interval = (data as Record<string, unknown>).interval;\n\tconst expiresIn = (data as Record<string, unknown>).expires_in;\n\n\tif (\n\t\ttypeof deviceCode !== \"string\" ||\n\t\ttypeof userCode !== \"string\" ||\n\t\ttypeof verificationUri !== \"string\" ||\n\t\t(interval !== undefined && typeof interval !== \"number\") ||\n\t\ttypeof expiresIn !== \"number\"\n\t) {\n\t\tthrow new Error(\"Invalid device code response fields\");\n\t}\n\n\treturn {\n\t\tdevice_code: deviceCode,\n\t\tuser_code: userCode,\n\t\tverification_uri: verificationUri,\n\t\tinterval,\n\t\texpires_in: expiresIn,\n\t};\n}\n\nasync function pollForGitHubAccessToken(domain: string, device: DeviceCodeResponse, signal?: AbortSignal) {\n\tconst urls = getUrls(domain);\n\treturn pollOAuthDeviceCodeFlow({\n\t\tintervalSeconds: device.interval,\n\t\texpiresInSeconds: device.expires_in,\n\t\tsignal,\n\t\tpoll: async () => {\n\t\t\tconst raw = await fetchJson(urls.accessTokenUrl, {\n\t\t\t\tmethod: \"POST\",\n\t\t\t\theaders: {\n\t\t\t\t\tAccept: \"application/json\",\n\t\t\t\t\t\"Content-Type\": \"application/x-www-form-urlencoded\",\n\t\t\t\t\t\"User-Agent\": \"GitHubCopilotChat/0.35.0\",\n\t\t\t\t},\n\t\t\t\tbody: new URLSearchParams({\n\t\t\t\t\tclient_id: CLIENT_ID,\n\t\t\t\t\tdevice_code: device.device_code,\n\t\t\t\t\tgrant_type: \"urn:ietf:params:oauth:grant-type:device_code\",\n\t\t\t\t}),\n\t\t\t});\n\n\t\t\tif (raw && typeof raw === \"object\" && typeof (raw as DeviceTokenSuccessResponse).access_token === \"string\") {\n\t\t\t\treturn { status: \"complete\", accessToken: (raw as DeviceTokenSuccessResponse).access_token };\n\t\t\t}\n\n\t\t\tif (raw && typeof raw === \"object\" && typeof (raw as DeviceTokenErrorResponse).error === \"string\") {\n\t\t\t\tconst { error, error_description: description } = raw as DeviceTokenErrorResponse;\n\t\t\t\tif (error === \"authorization_pending\") {\n\t\t\t\t\treturn { status: \"pending\" };\n\t\t\t\t}\n\n\t\t\t\tif (error === \"slow_down\") {\n\t\t\t\t\treturn { status: \"slow_down\" };\n\t\t\t\t}\n\n\t\t\t\tconst descriptionSuffix = description ? `: ${description}` : \"\";\n\t\t\t\treturn { status: \"failed\", message: `Device flow failed: ${error}${descriptionSuffix}` };\n\t\t\t}\n\n\t\t\treturn { status: \"failed\", message: \"Invalid device token response\" };\n\t\t},\n\t});\n}\n\n/**\n * Refresh GitHub Copilot token\n */\nexport async function refreshGitHubCopilotToken(\n\trefreshToken: string,\n\tenterpriseDomain?: string,\n): Promise<OAuthCredentials> {\n\tconst domain = enterpriseDomain || \"github.com\";\n\tconst urls = getUrls(domain);\n\n\tconst raw = await fetchJson(urls.copilotTokenUrl, {\n\t\theaders: {\n\t\t\tAccept: \"application/json\",\n\t\t\tAuthorization: `Bearer ${refreshToken}`,\n\t\t\t...COPILOT_HEADERS,\n\t\t},\n\t});\n\n\tif (!raw || typeof raw !== \"object\") {\n\t\tthrow new Error(\"Invalid Copilot token response\");\n\t}\n\n\tconst token = (raw as Record<string, unknown>).token;\n\tconst expiresAt = (raw as Record<string, unknown>).expires_at;\n\n\tif (typeof token !== \"string\" || typeof expiresAt !== \"number\") {\n\t\tthrow new Error(\"Invalid Copilot token response fields\");\n\t}\n\n\treturn {\n\t\trefresh: refreshToken,\n\t\taccess: token,\n\t\texpires: expiresAt * 1000 - 5 * 60 * 1000,\n\t\tenterpriseUrl: enterpriseDomain,\n\t};\n}\n\n/**\n * Enable a model for the user's GitHub Copilot account.\n * This is required for some models (like Claude, Grok) before they can be used.\n */\nasync function enableGitHubCopilotModel(token: string, modelId: string, enterpriseDomain?: string): Promise<boolean> {\n\tconst baseUrl = getGitHubCopilotBaseUrl(token, enterpriseDomain);\n\tconst url = `${baseUrl}/models/${modelId}/policy`;\n\n\ttry {\n\t\tconst response = await fetch(url, {\n\t\t\tmethod: \"POST\",\n\t\t\theaders: {\n\t\t\t\t\"Content-Type\": \"application/json\",\n\t\t\t\tAuthorization: `Bearer ${token}`,\n\t\t\t\t...COPILOT_HEADERS,\n\t\t\t\t\"openai-intent\": \"chat-policy\",\n\t\t\t\t\"x-interaction-type\": \"chat-policy\",\n\t\t\t},\n\t\t\tbody: JSON.stringify({ state: \"enabled\" }),\n\t\t});\n\t\treturn response.ok;\n\t} catch {\n\t\treturn false;\n\t}\n}\n\n/**\n * Enable all known GitHub Copilot models that may require policy acceptance.\n * Called after successful login to ensure all models are available.\n */\nasync function enableAllGitHubCopilotModels(\n\ttoken: string,\n\tenterpriseDomain?: string,\n\tonProgress?: (model: string, success: boolean) => void,\n): Promise<void> {\n\tconst models = getModels(\"github-copilot\");\n\tawait Promise.all(\n\t\tmodels.map(async (model) => {\n\t\t\tconst success = await enableGitHubCopilotModel(token, model.id, enterpriseDomain);\n\t\t\tonProgress?.(model.id, success);\n\t\t}),\n\t);\n}\n\n/**\n * Login with GitHub Copilot OAuth (device code flow)\n *\n * @param options.onDeviceCode - Callback with URL and user code\n * @param options.onPrompt - Callback to prompt user for input\n * @param options.onProgress - Optional progress callback\n * @param options.signal - Optional AbortSignal for cancellation\n */\nexport async function loginGitHubCopilot(options: {\n\tonDeviceCode: (info: OAuthDeviceCodeInfo) => void;\n\tonPrompt: (prompt: { message: string; placeholder?: string; allowEmpty?: boolean }) => Promise<string>;\n\tonProgress?: (message: string) => void;\n\tsignal?: AbortSignal;\n}): Promise<OAuthCredentials> {\n\tconst input = await options.onPrompt({\n\t\tmessage: \"GitHub Enterprise URL/domain (blank for github.com)\",\n\t\tplaceholder: \"company.ghe.com\",\n\t\tallowEmpty: true,\n\t});\n\n\tif (options.signal?.aborted) {\n\t\tthrow new Error(\"Login cancelled\");\n\t}\n\n\tconst trimmed = input.trim();\n\tconst enterpriseDomain = normalizeDomain(input);\n\tif (trimmed && !enterpriseDomain) {\n\t\tthrow new Error(\"Invalid GitHub Enterprise URL/domain\");\n\t}\n\tconst domain = enterpriseDomain || \"github.com\";\n\n\tconst device = await startDeviceFlow(domain);\n\toptions.onDeviceCode({\n\t\tuserCode: device.user_code,\n\t\tverificationUri: device.verification_uri,\n\t\tintervalSeconds: device.interval,\n\t\texpiresInSeconds: device.expires_in,\n\t});\n\n\tconst githubAccessToken = await pollForGitHubAccessToken(domain, device, options.signal);\n\tconst credentials = await refreshGitHubCopilotToken(githubAccessToken, enterpriseDomain ?? undefined);\n\n\t// Enable all models after successful login\n\toptions.onProgress?.(\"Enabling models...\");\n\tawait enableAllGitHubCopilotModels(credentials.access, enterpriseDomain ?? undefined);\n\treturn credentials;\n}\n\nexport const githubCopilotOAuthProvider: OAuthProviderInterface = {\n\tid: \"github-copilot\",\n\tname: \"GitHub Copilot\",\n\n\tasync login(callbacks: OAuthLoginCallbacks): Promise<OAuthCredentials> {\n\t\tif (!callbacks.onDeviceCode) {\n\t\t\tthrow new Error(\"GitHub Copilot OAuth requires a device code callback\");\n\t\t}\n\n\t\treturn loginGitHubCopilot({\n\t\t\tonDeviceCode: callbacks.onDeviceCode,\n\t\t\tonPrompt: callbacks.onPrompt,\n\t\t\tonProgress: callbacks.onProgress,\n\t\t\tsignal: callbacks.signal,\n\t\t});\n\t},\n\n\tasync refreshToken(credentials: OAuthCredentials): Promise<OAuthCredentials> {\n\t\tconst creds = credentials as CopilotCredentials;\n\t\treturn refreshGitHubCopilotToken(creds.refresh, creds.enterpriseUrl);\n\t},\n\n\tgetApiKey(credentials: OAuthCredentials): string {\n\t\treturn credentials.access;\n\t},\n\n\tmodifyModels(models: Model<Api>[], credentials: OAuthCredentials): Model<Api>[] {\n\t\tconst creds = credentials as CopilotCredentials;\n\t\tconst domain = creds.enterpriseUrl ? (normalizeDomain(creds.enterpriseUrl) ?? undefined) : undefined;\n\t\tconst baseUrl = getGitHubCopilotBaseUrl(creds.access, domain);\n\t\treturn models.map((m) => (m.provider === \"github-copilot\" ? { ...m, baseUrl } : m));\n\t},\n};\n"]}
+{"version":3,"file":"github-copilot.js","sourceRoot":"","sources":["../../../src/utils/oauth/github-copilot.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,OAAO,EAAE,SAAS,EAAE,MAAM,iBAAiB,CAAC;AAE5C,OAAO,EAAE,uBAAuB,EAAE,MAAM,kBAAkB,CAAC;AAO3D,MAAM,MAAM,GAAG,CAAC,CAAS,EAAE,EAAE,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;AACtC,MAAM,SAAS,GAAG,MAAM,CAAC,8BAA8B,CAAC,CAAC;AAEzD,MAAM,eAAe,GAAG;IACvB,YAAY,EAAE,0BAA0B;IACxC,gBAAgB,EAAE,gBAAgB;IAClC,uBAAuB,EAAE,qBAAqB;IAC9C,wBAAwB,EAAE,aAAa;CAC9B,CAAC;AAsBX,MAAM,UAAU,eAAe,CAAC,KAAa,EAAiB;IAC7D,MAAM,OAAO,GAAG,KAAK,CAAC,IAAI,EAAE,CAAC;IAC7B,IAAI,CAAC,OAAO;QAAE,OAAO,IAAI,CAAC;IAC1B,IAAI,CAAC;QACJ,MAAM,GAAG,GAAG,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,IAAI,GAAG,CAAC,WAAW,OAAO,EAAE,CAAC,CAAC;QACvF,OAAO,GAAG,CAAC,QAAQ,CAAC;IACrB,CAAC;IAAC,MAAM,CAAC;QACR,OAAO,IAAI,CAAC;IACb,CAAC;AAAA,CACD;AAED,SAAS,OAAO,CAAC,MAAc,EAI7B;IACD,OAAO;QACN,aAAa,EAAE,WAAW,MAAM,oBAAoB;QACpD,cAAc,EAAE,WAAW,MAAM,2BAA2B;QAC5D,eAAe,EAAE,eAAe,MAAM,4BAA4B;KAClE,CAAC;AAAA,CACF;AAED;;;;GAIG;AACH,SAAS,mBAAmB,CAAC,KAAa,EAAiB;IAC1D,MAAM,KAAK,GAAG,KAAK,CAAC,KAAK,CAAC,kBAAkB,CAAC,CAAC;IAC9C,IAAI,CAAC,KAAK;QAAE,OAAO,IAAI,CAAC;IACxB,MAAM,SAAS,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC;IAC3B,+BAA+B;IAC/B,MAAM,OAAO,GAAG,SAAS,CAAC,OAAO,CAAC,UAAU,EAAE,MAAM,CAAC,CAAC;IACtD,OAAO,WAAW,OAAO,EAAE,CAAC;AAAA,CAC5B;AAED,MAAM,UAAU,uBAAuB,CAAC,KAAc,EAAE,gBAAyB,EAAU;IAC1F,yDAAyD;IACzD,IAAI,KAAK,EAAE,CAAC;QACX,MAAM,YAAY,GAAG,mBAAmB,CAAC,KAAK,CAAC,CAAC;QAChD,IAAI,YAAY;YAAE,OAAO,YAAY,CAAC;IACvC,CAAC;IACD,oDAAoD;IACpD,IAAI,gBAAgB;QAAE,OAAO,uBAAuB,gBAAgB,EAAE,CAAC;IACvE,OAAO,0CAA0C,CAAC;AAAA,CAClD;AAED,KAAK,UAAU,SAAS,CAAC,GAAW,EAAE,IAAiB,EAAoB;IAC1E,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,GAAG,EAAE,IAAI,CAAC,CAAC;IACxC,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;QAClB,MAAM,IAAI,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;QACnC,MAAM,IAAI,KAAK,CAAC,GAAG,QAAQ,CAAC,MAAM,IAAI,QAAQ,CAAC,UAAU,KAAK,IAAI,EAAE,CAAC,CAAC;IACvE,CAAC;IACD,OAAO,QAAQ,CAAC,IAAI,EAAE,CAAC;AAAA,CACvB;AAED,KAAK,UAAU,eAAe,CAAC,MAAc,EAA+B;IAC3E,MAAM,IAAI,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC;IAC7B,MAAM,IAAI,GAAG,MAAM,SAAS,CAAC,IAAI,CAAC,aAAa,EAAE;QAChD,MAAM,EAAE,MAAM;QACd,OAAO,EAAE;YACR,MAAM,EAAE,kBAAkB;YAC1B,cAAc,EAAE,mCAAmC;YACnD,YAAY,EAAE,0BAA0B;SACxC;QACD,IAAI,EAAE,IAAI,eAAe,CAAC;YACzB,SAAS,EAAE,SAAS;YACpB,KAAK,EAAE,WAAW;SAClB,CAAC;KACF,CAAC,CAAC;IAEH,IAAI,CAAC,IAAI,IAAI,OAAO,IAAI,KAAK,QAAQ,EAAE,CAAC;QACvC,MAAM,IAAI,KAAK,CAAC,8BAA8B,CAAC,CAAC;IACjD,CAAC;IAED,MAAM,UAAU,GAAI,IAAgC,CAAC,WAAW,CAAC;IACjE,MAAM,QAAQ,GAAI,IAAgC,CAAC,SAAS,CAAC;IAC7D,MAAM,eAAe,GAAI,IAAgC,CAAC,gBAAgB,CAAC;IAC3E,MAAM,QAAQ,GAAI,IAAgC,CAAC,QAAQ,CAAC;IAC5D,MAAM,SAAS,GAAI,IAAgC,CAAC,UAAU,CAAC;IAE/D,IACC,OAAO,UAAU,KAAK,QAAQ;QAC9B,OAAO,QAAQ,KAAK,QAAQ;QAC5B,OAAO,eAAe,KAAK,QAAQ;QACnC,CAAC,QAAQ,KAAK,SAAS,IAAI,OAAO,QAAQ,KAAK,QAAQ,CAAC;QACxD,OAAO,SAAS,KAAK,QAAQ,EAC5B,CAAC;QACF,MAAM,IAAI,KAAK,CAAC,qCAAqC,CAAC,CAAC;IACxD,CAAC;IAED,OAAO;QACN,WAAW,EAAE,UAAU;QACvB,SAAS,EAAE,QAAQ;QACnB,gBAAgB,EAAE,eAAe;QACjC,QAAQ;QACR,UAAU,EAAE,SAAS;KACrB,CAAC;AAAA,CACF;AAED,KAAK,UAAU,wBAAwB,CAAC,MAAc,EAAE,MAA0B,EAAE,MAAoB,EAAE;IACzG,MAAM,IAAI,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC;IAC7B,OAAO,uBAAuB,CAAC;QAC9B,eAAe,EAAE,MAAM,CAAC,QAAQ;QAChC,gBAAgB,EAAE,MAAM,CAAC,UAAU;QACnC,mBAAmB,EAAE,IAAI;QACzB,MAAM;QACN,IAAI,EAAE,KAAK,IAAI,EAAE,CAAC;YACjB,MAAM,GAAG,GAAG,MAAM,SAAS,CAAC,IAAI,CAAC,cAAc,EAAE;gBAChD,MAAM,EAAE,MAAM;gBACd,OAAO,EAAE;oBACR,MAAM,EAAE,kBAAkB;oBAC1B,cAAc,EAAE,mCAAmC;oBACnD,YAAY,EAAE,0BAA0B;iBACxC;gBACD,IAAI,EAAE,IAAI,eAAe,CAAC;oBACzB,SAAS,EAAE,SAAS;oBACpB,WAAW,EAAE,MAAM,CAAC,WAAW;oBAC/B,UAAU,EAAE,8CAA8C;iBAC1D,CAAC;aACF,CAAC,CAAC;YAEH,IAAI,GAAG,IAAI,OAAO,GAAG,KAAK,QAAQ,IAAI,OAAQ,GAAkC,CAAC,YAAY,KAAK,QAAQ,EAAE,CAAC;gBAC5G,OAAO,EAAE,MAAM,EAAE,UAAU,EAAE,WAAW,EAAG,GAAkC,CAAC,YAAY,EAAE,CAAC;YAC9F,CAAC;YAED,IAAI,GAAG,IAAI,OAAO,GAAG,KAAK,QAAQ,IAAI,OAAQ,GAAgC,CAAC,KAAK,KAAK,QAAQ,EAAE,CAAC;gBACnG,MAAM,EAAE,KAAK,EAAE,iBAAiB,EAAE,WAAW,EAAE,QAAQ,EAAE,GAAG,GAA+B,CAAC;gBAC5F,IAAI,KAAK,KAAK,uBAAuB,EAAE,CAAC;oBACvC,OAAO,EAAE,MAAM,EAAE,SAAS,EAAE,CAAC;gBAC9B,CAAC;gBAED,IAAI,KAAK,KAAK,WAAW,EAAE,CAAC;oBAC3B,OAAO,EAAE,MAAM,EAAE,WAAW,EAAE,eAAe,EAAE,OAAO,QAAQ,KAAK,QAAQ,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAC,SAAS,EAAE,CAAC;gBACtG,CAAC;gBAED,MAAM,iBAAiB,GAAG,WAAW,CAAC,CAAC,CAAC,KAAK,WAAW,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;gBAChE,OAAO,EAAE,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,uBAAuB,KAAK,GAAG,iBAAiB,EAAE,EAAE,CAAC;YAC1F,CAAC;YAED,OAAO,EAAE,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,+BAA+B,EAAE,CAAC;QAAA,CACtE;KACD,CAAC,CAAC;AAAA,CACH;AAED;;GAEG;AACH,MAAM,CAAC,KAAK,UAAU,yBAAyB,CAC9C,YAAoB,EACpB,gBAAyB,EACG;IAC5B,MAAM,MAAM,GAAG,gBAAgB,IAAI,YAAY,CAAC;IAChD,MAAM,IAAI,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC;IAE7B,MAAM,GAAG,GAAG,MAAM,SAAS,CAAC,IAAI,CAAC,eAAe,EAAE;QACjD,OAAO,EAAE;YACR,MAAM,EAAE,kBAAkB;YAC1B,aAAa,EAAE,UAAU,YAAY,EAAE;YACvC,GAAG,eAAe;SAClB;KACD,CAAC,CAAC;IAEH,IAAI,CAAC,GAAG,IAAI,OAAO,GAAG,KAAK,QAAQ,EAAE,CAAC;QACrC,MAAM,IAAI,KAAK,CAAC,gCAAgC,CAAC,CAAC;IACnD,CAAC;IAED,MAAM,KAAK,GAAI,GAA+B,CAAC,KAAK,CAAC;IACrD,MAAM,SAAS,GAAI,GAA+B,CAAC,UAAU,CAAC;IAE9D,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,OAAO,SAAS,KAAK,QAAQ,EAAE,CAAC;QAChE,MAAM,IAAI,KAAK,CAAC,uCAAuC,CAAC,CAAC;IAC1D,CAAC;IAED,OAAO;QACN,OAAO,EAAE,YAAY;QACrB,MAAM,EAAE,KAAK;QACb,OAAO,EAAE,SAAS,GAAG,IAAI,GAAG,CAAC,GAAG,EAAE,GAAG,IAAI;QACzC,aAAa,EAAE,gBAAgB;KAC/B,CAAC;AAAA,CACF;AAED;;;GAGG;AACH,KAAK,UAAU,wBAAwB,CAAC,KAAa,EAAE,OAAe,EAAE,gBAAyB,EAAoB;IACpH,MAAM,OAAO,GAAG,uBAAuB,CAAC,KAAK,EAAE,gBAAgB,CAAC,CAAC;IACjE,MAAM,GAAG,GAAG,GAAG,OAAO,WAAW,OAAO,SAAS,CAAC;IAElD,IAAI,CAAC;QACJ,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,GAAG,EAAE;YACjC,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACR,cAAc,EAAE,kBAAkB;gBAClC,aAAa,EAAE,UAAU,KAAK,EAAE;gBAChC,GAAG,eAAe;gBAClB,eAAe,EAAE,aAAa;gBAC9B,oBAAoB,EAAE,aAAa;aACnC;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,EAAE,KAAK,EAAE,SAAS,EAAE,CAAC;SAC1C,CAAC,CAAC;QACH,OAAO,QAAQ,CAAC,EAAE,CAAC;IACpB,CAAC;IAAC,MAAM,CAAC;QACR,OAAO,KAAK,CAAC;IACd,CAAC;AAAA,CACD;AAED;;;GAGG;AACH,KAAK,UAAU,4BAA4B,CAC1C,KAAa,EACb,gBAAyB,EACzB,UAAsD,EACtC;IAChB,MAAM,MAAM,GAAG,SAAS,CAAC,gBAAgB,CAAC,CAAC;IAC3C,MAAM,OAAO,CAAC,GAAG,CAChB,MAAM,CAAC,GAAG,CAAC,KAAK,EAAE,KAAK,EAAE,EAAE,CAAC;QAC3B,MAAM,OAAO,GAAG,MAAM,wBAAwB,CAAC,KAAK,EAAE,KAAK,CAAC,EAAE,EAAE,gBAAgB,CAAC,CAAC;QAClF,UAAU,EAAE,CAAC,KAAK,CAAC,EAAE,EAAE,OAAO,CAAC,CAAC;IAAA,CAChC,CAAC,CACF,CAAC;AAAA,CACF;AAED;;;;;;;GAOG;AACH,MAAM,CAAC,KAAK,UAAU,kBAAkB,CAAC,OAKxC,EAA6B;IAC7B,MAAM,KAAK,GAAG,MAAM,OAAO,CAAC,QAAQ,CAAC;QACpC,OAAO,EAAE,qDAAqD;QAC9D,WAAW,EAAE,iBAAiB;QAC9B,UAAU,EAAE,IAAI;KAChB,CAAC,CAAC;IAEH,IAAI,OAAO,CAAC,MAAM,EAAE,OAAO,EAAE,CAAC;QAC7B,MAAM,IAAI,KAAK,CAAC,iBAAiB,CAAC,CAAC;IACpC,CAAC;IAED,MAAM,OAAO,GAAG,KAAK,CAAC,IAAI,EAAE,CAAC;IAC7B,MAAM,gBAAgB,GAAG,eAAe,CAAC,KAAK,CAAC,CAAC;IAChD,IAAI,OAAO,IAAI,CAAC,gBAAgB,EAAE,CAAC;QAClC,MAAM,IAAI,KAAK,CAAC,sCAAsC,CAAC,CAAC;IACzD,CAAC;IACD,MAAM,MAAM,GAAG,gBAAgB,IAAI,YAAY,CAAC;IAEhD,MAAM,MAAM,GAAG,MAAM,eAAe,CAAC,MAAM,CAAC,CAAC;IAC7C,OAAO,CAAC,YAAY,CAAC;QACpB,QAAQ,EAAE,MAAM,CAAC,SAAS;QAC1B,eAAe,EAAE,MAAM,CAAC,gBAAgB;QACxC,eAAe,EAAE,MAAM,CAAC,QAAQ;QAChC,gBAAgB,EAAE,MAAM,CAAC,UAAU;KACnC,CAAC,CAAC;IAEH,MAAM,iBAAiB,GAAG,MAAM,wBAAwB,CAAC,MAAM,EAAE,MAAM,EAAE,OAAO,CAAC,MAAM,CAAC,CAAC;IACzF,MAAM,WAAW,GAAG,MAAM,yBAAyB,CAAC,iBAAiB,EAAE,gBAAgB,IAAI,SAAS,CAAC,CAAC;IAEtG,2CAA2C;IAC3C,OAAO,CAAC,UAAU,EAAE,CAAC,oBAAoB,CAAC,CAAC;IAC3C,MAAM,4BAA4B,CAAC,WAAW,CAAC,MAAM,EAAE,gBAAgB,IAAI,SAAS,CAAC,CAAC;IACtF,OAAO,WAAW,CAAC;AAAA,CACnB;AAED,MAAM,CAAC,MAAM,0BAA0B,GAA2B;IACjE,EAAE,EAAE,gBAAgB;IACpB,IAAI,EAAE,gBAAgB;IAEtB,KAAK,CAAC,KAAK,CAAC,SAA8B,EAA6B;QACtE,IAAI,CAAC,SAAS,CAAC,YAAY,EAAE,CAAC;YAC7B,MAAM,IAAI,KAAK,CAAC,sDAAsD,CAAC,CAAC;QACzE,CAAC;QAED,OAAO,kBAAkB,CAAC;YACzB,YAAY,EAAE,SAAS,CAAC,YAAY;YACpC,QAAQ,EAAE,SAAS,CAAC,QAAQ;YAC5B,UAAU,EAAE,SAAS,CAAC,UAAU;YAChC,MAAM,EAAE,SAAS,CAAC,MAAM;SACxB,CAAC,CAAC;IAAA,CACH;IAED,KAAK,CAAC,YAAY,CAAC,WAA6B,EAA6B;QAC5E,MAAM,KAAK,GAAG,WAAiC,CAAC;QAChD,OAAO,yBAAyB,CAAC,KAAK,CAAC,OAAO,EAAE,KAAK,CAAC,aAAa,CAAC,CAAC;IAAA,CACrE;IAED,SAAS,CAAC,WAA6B,EAAU;QAChD,OAAO,WAAW,CAAC,MAAM,CAAC;IAAA,CAC1B;IAED,YAAY,CAAC,MAAoB,EAAE,WAA6B,EAAgB;QAC/E,MAAM,KAAK,GAAG,WAAiC,CAAC;QAChD,MAAM,MAAM,GAAG,KAAK,CAAC,aAAa,CAAC,CAAC,CAAC,CAAC,eAAe,CAAC,KAAK,CAAC,aAAa,CAAC,IAAI,SAAS,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;QACrG,MAAM,OAAO,GAAG,uBAAuB,CAAC,KAAK,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;QAC9D,OAAO,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,QAAQ,KAAK,gBAAgB,CAAC,CAAC,CAAC,EAAE,GAAG,CAAC,EAAE,OAAO,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;IAAA,CACpF;CACD,CAAC","sourcesContent":["/**\n * GitHub Copilot OAuth flow\n */\n\nimport { getModels } from \"../../models.ts\";\nimport type { Api, Model } from \"../../types.ts\";\nimport { pollOAuthDeviceCodeFlow } from \"./device-code.ts\";\nimport type { OAuthCredentials, OAuthDeviceCodeInfo, OAuthLoginCallbacks, OAuthProviderInterface } from \"./types.ts\";\n\ntype CopilotCredentials = OAuthCredentials & {\n\tenterpriseUrl?: string;\n};\n\nconst decode = (s: string) => atob(s);\nconst CLIENT_ID = decode(\"SXYxLmI1MDdhMDhjODdlY2ZlOTg=\");\n\nconst COPILOT_HEADERS = {\n\t\"User-Agent\": \"GitHubCopilotChat/0.35.0\",\n\t\"Editor-Version\": \"vscode/1.107.0\",\n\t\"Editor-Plugin-Version\": \"copilot-chat/0.35.0\",\n\t\"Copilot-Integration-Id\": \"vscode-chat\",\n} as const;\n\ntype DeviceCodeResponse = {\n\tdevice_code: string;\n\tuser_code: string;\n\tverification_uri: string;\n\tinterval?: number;\n\texpires_in: number;\n};\n\ntype DeviceTokenSuccessResponse = {\n\taccess_token: string;\n\ttoken_type?: string;\n\tscope?: string;\n};\n\ntype DeviceTokenErrorResponse = {\n\terror: string;\n\terror_description?: string;\n\tinterval?: number;\n};\n\nexport function normalizeDomain(input: string): string | null {\n\tconst trimmed = input.trim();\n\tif (!trimmed) return null;\n\ttry {\n\t\tconst url = trimmed.includes(\"://\") ? new URL(trimmed) : new URL(`https://${trimmed}`);\n\t\treturn url.hostname;\n\t} catch {\n\t\treturn null;\n\t}\n}\n\nfunction getUrls(domain: string): {\n\tdeviceCodeUrl: string;\n\taccessTokenUrl: string;\n\tcopilotTokenUrl: string;\n} {\n\treturn {\n\t\tdeviceCodeUrl: `https://${domain}/login/device/code`,\n\t\taccessTokenUrl: `https://${domain}/login/oauth/access_token`,\n\t\tcopilotTokenUrl: `https://api.${domain}/copilot_internal/v2/token`,\n\t};\n}\n\n/**\n * Parse the proxy-ep from a Copilot token and convert to API base URL.\n * Token format: tid=...;exp=...;proxy-ep=proxy.individual.githubcopilot.com;...\n * Returns API URL like https://api.individual.githubcopilot.com\n */\nfunction getBaseUrlFromToken(token: string): string | null {\n\tconst match = token.match(/proxy-ep=([^;]+)/);\n\tif (!match) return null;\n\tconst proxyHost = match[1];\n\t// Convert proxy.xxx to api.xxx\n\tconst apiHost = proxyHost.replace(/^proxy\\./, \"api.\");\n\treturn `https://${apiHost}`;\n}\n\nexport function getGitHubCopilotBaseUrl(token?: string, enterpriseDomain?: string): string {\n\t// If we have a token, extract the base URL from proxy-ep\n\tif (token) {\n\t\tconst urlFromToken = getBaseUrlFromToken(token);\n\t\tif (urlFromToken) return urlFromToken;\n\t}\n\t// Fallback for enterprise or if token parsing fails\n\tif (enterpriseDomain) return `https://copilot-api.${enterpriseDomain}`;\n\treturn \"https://api.individual.githubcopilot.com\";\n}\n\nasync function fetchJson(url: string, init: RequestInit): Promise<unknown> {\n\tconst response = await fetch(url, init);\n\tif (!response.ok) {\n\t\tconst text = await response.text();\n\t\tthrow new Error(`${response.status} ${response.statusText}: ${text}`);\n\t}\n\treturn response.json();\n}\n\nasync function startDeviceFlow(domain: string): Promise<DeviceCodeResponse> {\n\tconst urls = getUrls(domain);\n\tconst data = await fetchJson(urls.deviceCodeUrl, {\n\t\tmethod: \"POST\",\n\t\theaders: {\n\t\t\tAccept: \"application/json\",\n\t\t\t\"Content-Type\": \"application/x-www-form-urlencoded\",\n\t\t\t\"User-Agent\": \"GitHubCopilotChat/0.35.0\",\n\t\t},\n\t\tbody: new URLSearchParams({\n\t\t\tclient_id: CLIENT_ID,\n\t\t\tscope: \"read:user\",\n\t\t}),\n\t});\n\n\tif (!data || typeof data !== \"object\") {\n\t\tthrow new Error(\"Invalid device code response\");\n\t}\n\n\tconst deviceCode = (data as Record<string, unknown>).device_code;\n\tconst userCode = (data as Record<string, unknown>).user_code;\n\tconst verificationUri = (data as Record<string, unknown>).verification_uri;\n\tconst interval = (data as Record<string, unknown>).interval;\n\tconst expiresIn = (data as Record<string, unknown>).expires_in;\n\n\tif (\n\t\ttypeof deviceCode !== \"string\" ||\n\t\ttypeof userCode !== \"string\" ||\n\t\ttypeof verificationUri !== \"string\" ||\n\t\t(interval !== undefined && typeof interval !== \"number\") ||\n\t\ttypeof expiresIn !== \"number\"\n\t) {\n\t\tthrow new Error(\"Invalid device code response fields\");\n\t}\n\n\treturn {\n\t\tdevice_code: deviceCode,\n\t\tuser_code: userCode,\n\t\tverification_uri: verificationUri,\n\t\tinterval,\n\t\texpires_in: expiresIn,\n\t};\n}\n\nasync function pollForGitHubAccessToken(domain: string, device: DeviceCodeResponse, signal?: AbortSignal) {\n\tconst urls = getUrls(domain);\n\treturn pollOAuthDeviceCodeFlow({\n\t\tintervalSeconds: device.interval,\n\t\texpiresInSeconds: device.expires_in,\n\t\twaitBeforeFirstPoll: true,\n\t\tsignal,\n\t\tpoll: async () => {\n\t\t\tconst raw = await fetchJson(urls.accessTokenUrl, {\n\t\t\t\tmethod: \"POST\",\n\t\t\t\theaders: {\n\t\t\t\t\tAccept: \"application/json\",\n\t\t\t\t\t\"Content-Type\": \"application/x-www-form-urlencoded\",\n\t\t\t\t\t\"User-Agent\": \"GitHubCopilotChat/0.35.0\",\n\t\t\t\t},\n\t\t\t\tbody: new URLSearchParams({\n\t\t\t\t\tclient_id: CLIENT_ID,\n\t\t\t\t\tdevice_code: device.device_code,\n\t\t\t\t\tgrant_type: \"urn:ietf:params:oauth:grant-type:device_code\",\n\t\t\t\t}),\n\t\t\t});\n\n\t\t\tif (raw && typeof raw === \"object\" && typeof (raw as DeviceTokenSuccessResponse).access_token === \"string\") {\n\t\t\t\treturn { status: \"complete\", accessToken: (raw as DeviceTokenSuccessResponse).access_token };\n\t\t\t}\n\n\t\t\tif (raw && typeof raw === \"object\" && typeof (raw as DeviceTokenErrorResponse).error === \"string\") {\n\t\t\t\tconst { error, error_description: description, interval } = raw as DeviceTokenErrorResponse;\n\t\t\t\tif (error === \"authorization_pending\") {\n\t\t\t\t\treturn { status: \"pending\" };\n\t\t\t\t}\n\n\t\t\t\tif (error === \"slow_down\") {\n\t\t\t\t\treturn { status: \"slow_down\", intervalSeconds: typeof interval === \"number\" ? interval : undefined };\n\t\t\t\t}\n\n\t\t\t\tconst descriptionSuffix = description ? `: ${description}` : \"\";\n\t\t\t\treturn { status: \"failed\", message: `Device flow failed: ${error}${descriptionSuffix}` };\n\t\t\t}\n\n\t\t\treturn { status: \"failed\", message: \"Invalid device token response\" };\n\t\t},\n\t});\n}\n\n/**\n * Refresh GitHub Copilot token\n */\nexport async function refreshGitHubCopilotToken(\n\trefreshToken: string,\n\tenterpriseDomain?: string,\n): Promise<OAuthCredentials> {\n\tconst domain = enterpriseDomain || \"github.com\";\n\tconst urls = getUrls(domain);\n\n\tconst raw = await fetchJson(urls.copilotTokenUrl, {\n\t\theaders: {\n\t\t\tAccept: \"application/json\",\n\t\t\tAuthorization: `Bearer ${refreshToken}`,\n\t\t\t...COPILOT_HEADERS,\n\t\t},\n\t});\n\n\tif (!raw || typeof raw !== \"object\") {\n\t\tthrow new Error(\"Invalid Copilot token response\");\n\t}\n\n\tconst token = (raw as Record<string, unknown>).token;\n\tconst expiresAt = (raw as Record<string, unknown>).expires_at;\n\n\tif (typeof token !== \"string\" || typeof expiresAt !== \"number\") {\n\t\tthrow new Error(\"Invalid Copilot token response fields\");\n\t}\n\n\treturn {\n\t\trefresh: refreshToken,\n\t\taccess: token,\n\t\texpires: expiresAt * 1000 - 5 * 60 * 1000,\n\t\tenterpriseUrl: enterpriseDomain,\n\t};\n}\n\n/**\n * Enable a model for the user's GitHub Copilot account.\n * This is required for some models (like Claude, Grok) before they can be used.\n */\nasync function enableGitHubCopilotModel(token: string, modelId: string, enterpriseDomain?: string): Promise<boolean> {\n\tconst baseUrl = getGitHubCopilotBaseUrl(token, enterpriseDomain);\n\tconst url = `${baseUrl}/models/${modelId}/policy`;\n\n\ttry {\n\t\tconst response = await fetch(url, {\n\t\t\tmethod: \"POST\",\n\t\t\theaders: {\n\t\t\t\t\"Content-Type\": \"application/json\",\n\t\t\t\tAuthorization: `Bearer ${token}`,\n\t\t\t\t...COPILOT_HEADERS,\n\t\t\t\t\"openai-intent\": \"chat-policy\",\n\t\t\t\t\"x-interaction-type\": \"chat-policy\",\n\t\t\t},\n\t\t\tbody: JSON.stringify({ state: \"enabled\" }),\n\t\t});\n\t\treturn response.ok;\n\t} catch {\n\t\treturn false;\n\t}\n}\n\n/**\n * Enable all known GitHub Copilot models that may require policy acceptance.\n * Called after successful login to ensure all models are available.\n */\nasync function enableAllGitHubCopilotModels(\n\ttoken: string,\n\tenterpriseDomain?: string,\n\tonProgress?: (model: string, success: boolean) => void,\n): Promise<void> {\n\tconst models = getModels(\"github-copilot\");\n\tawait Promise.all(\n\t\tmodels.map(async (model) => {\n\t\t\tconst success = await enableGitHubCopilotModel(token, model.id, enterpriseDomain);\n\t\t\tonProgress?.(model.id, success);\n\t\t}),\n\t);\n}\n\n/**\n * Login with GitHub Copilot OAuth (device code flow)\n *\n * @param options.onDeviceCode - Callback with URL and user code\n * @param options.onPrompt - Callback to prompt user for input\n * @param options.onProgress - Optional progress callback\n * @param options.signal - Optional AbortSignal for cancellation\n */\nexport async function loginGitHubCopilot(options: {\n\tonDeviceCode: (info: OAuthDeviceCodeInfo) => void;\n\tonPrompt: (prompt: { message: string; placeholder?: string; allowEmpty?: boolean }) => Promise<string>;\n\tonProgress?: (message: string) => void;\n\tsignal?: AbortSignal;\n}): Promise<OAuthCredentials> {\n\tconst input = await options.onPrompt({\n\t\tmessage: \"GitHub Enterprise URL/domain (blank for github.com)\",\n\t\tplaceholder: \"company.ghe.com\",\n\t\tallowEmpty: true,\n\t});\n\n\tif (options.signal?.aborted) {\n\t\tthrow new Error(\"Login cancelled\");\n\t}\n\n\tconst trimmed = input.trim();\n\tconst enterpriseDomain = normalizeDomain(input);\n\tif (trimmed && !enterpriseDomain) {\n\t\tthrow new Error(\"Invalid GitHub Enterprise URL/domain\");\n\t}\n\tconst domain = enterpriseDomain || \"github.com\";\n\n\tconst device = await startDeviceFlow(domain);\n\toptions.onDeviceCode({\n\t\tuserCode: device.user_code,\n\t\tverificationUri: device.verification_uri,\n\t\tintervalSeconds: device.interval,\n\t\texpiresInSeconds: device.expires_in,\n\t});\n\n\tconst githubAccessToken = await pollForGitHubAccessToken(domain, device, options.signal);\n\tconst credentials = await refreshGitHubCopilotToken(githubAccessToken, enterpriseDomain ?? undefined);\n\n\t// Enable all models after successful login\n\toptions.onProgress?.(\"Enabling models...\");\n\tawait enableAllGitHubCopilotModels(credentials.access, enterpriseDomain ?? undefined);\n\treturn credentials;\n}\n\nexport const githubCopilotOAuthProvider: OAuthProviderInterface = {\n\tid: \"github-copilot\",\n\tname: \"GitHub Copilot\",\n\n\tasync login(callbacks: OAuthLoginCallbacks): Promise<OAuthCredentials> {\n\t\tif (!callbacks.onDeviceCode) {\n\t\t\tthrow new Error(\"GitHub Copilot OAuth requires a device code callback\");\n\t\t}\n\n\t\treturn loginGitHubCopilot({\n\t\t\tonDeviceCode: callbacks.onDeviceCode,\n\t\t\tonPrompt: callbacks.onPrompt,\n\t\t\tonProgress: callbacks.onProgress,\n\t\t\tsignal: callbacks.signal,\n\t\t});\n\t},\n\n\tasync refreshToken(credentials: OAuthCredentials): Promise<OAuthCredentials> {\n\t\tconst creds = credentials as CopilotCredentials;\n\t\treturn refreshGitHubCopilotToken(creds.refresh, creds.enterpriseUrl);\n\t},\n\n\tgetApiKey(credentials: OAuthCredentials): string {\n\t\treturn credentials.access;\n\t},\n\n\tmodifyModels(models: Model<Api>[], credentials: OAuthCredentials): Model<Api>[] {\n\t\tconst creds = credentials as CopilotCredentials;\n\t\tconst domain = creds.enterpriseUrl ? (normalizeDomain(creds.enterpriseUrl) ?? undefined) : undefined;\n\t\tconst baseUrl = getGitHubCopilotBaseUrl(creds.access, domain);\n\t\treturn models.map((m) => (m.provider === \"github-copilot\" ? { ...m, baseUrl } : m));\n\t},\n};\n"]}

package/dist/utils/overflow.d.ts

@@ -24,6 +24,7 @@ import type { AssistantMessage } from "../types.ts";
  * - llama.cpp: "exceeds the available context size"
  * - LM Studio: "greater than the context length"
  * - Kimi For Coding: "exceeded model token limit: X (requested: Y)"
+ * - DS4: "Prompt has X tokens, but the configured context size is Y tokens"
  *
  * **Unreliable detection:**
  * - z.ai: Sometimes accepts overflow silently (detectable via usage.input > contextWindow),

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;AA0EpD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAiDG;AACH,wBAAgB,iBAAiB,CAAC,OAAO,EAAE,gBAAgB,EAAE,aAAa,CAAC,EAAE,MAAM,GAAG,OAAO,CA6B5F;AAED;;GAEG;AACH,wBAAgB,mBAAmB,IAAI,MAAM,EAAE,CAE9C","sourcesContent":["import type { AssistantMessage } from \"../types.ts\";\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 * - Anthropic: \"413 {\\\"error\\\":{\\\"type\\\":\\\"request_too_large\\\",\\\"message\\\":\\\"Request exceeds the maximum size\\\"}}\"\n * - OpenAI: \"Your input exceeds the context window of this model\"\n * - OpenAI/LiteLLM: \"Requested token count exceeds the model's maximum context length of 131072 tokens\"\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 * - OpenRouter/Poolside: \"Input length X exceeds the maximum allowed input length of Y tokens.\"\n * - Together AI: \"The input (X tokens) is longer than the model's context length (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 * - MiniMax: \"invalid params, context window exceeds limit\"\n * - Kimi For Coding: \"Your request exceeded model token limit: X (requested: Y)\"\n * - Cerebras: \"400/413 status code (no body)\"\n * - Mistral: \"Prompt contains X tokens ... too large for model with Y maximum context length\"\n * - z.ai: Does NOT error, accepts overflow silently - handled via usage.input > contextWindow\n * - Xiaomi MiMo: Truncates input to fill contextWindow exactly, then returns finish_reason \"length\"\n *   with output=0 (no room left to generate). Detected via stopReason \"length\" + zero output +\n *   input filling the context window.\n * - Ollama: Some deployments truncate silently, others return errors like \"prompt too long; exceeded max context length by X tokens\"\n */\nconst OVERFLOW_PATTERNS = [\n\t/prompt is too long/i, // Anthropic token overflow\n\t/request_too_large/i, // Anthropic request byte-size overflow (HTTP 413)\n\t/input is too long for requested model/i, // Amazon Bedrock\n\t/exceeds the context window/i, // OpenAI (Completions & Responses API)\n\t/exceeds (?:the )?(?:model'?s )?maximum context length of [\\d,]+ tokens?/i, // OpenAI-compatible proxies (LiteLLM)\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 (most backends)\n\t/exceeds (?:the )?maximum allowed input length of [\\d,]+ tokens?/i, // OpenRouter/Poolside\n\t/input \\(\\d+ tokens\\) is longer than the model'?s context length \\(\\d+ tokens\\)/i, // Together AI\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 window exceeds limit/i, // MiniMax\n\t/exceeded model token limit/i, // Kimi For Coding\n\t/too large for model with \\d+ maximum context length/i, // Mistral\n\t/model_context_window_exceeded/i, // z.ai non-standard finish_reason surfaced as error text\n\t/prompt too long; exceeded (?:max )?context length/i, // Ollama explicit overflow error\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\t/^4(?:00|13)\\s*(?:status code)?\\s*\\(no body\\)/i, // Cerebras: 400/413 with no body\n];\n\n/**\n * Patterns that indicate non-overflow errors (e.g. rate limiting, server errors).\n * Error messages matching any of these are excluded from overflow detection\n * even if they also match an OVERFLOW_PATTERN.\n *\n * Example: Bedrock formats throttling errors as \"ThrottlingException: Too many tokens,\n * please wait before trying again.\" which would match the /too many tokens/i overflow\n * pattern without this exclusion.\n */\nconst NON_OVERFLOW_PATTERNS = [\n\t/^(Throttling error|Service unavailable):/i, // AWS Bedrock non-overflow errors (human-readable prefixes from formatBedrockError)\n\t/rate limit/i, // Generic rate limiting\n\t/too many requests/i, // Generic HTTP 429 style\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\" or \"request_too_large\"\n * - OpenAI (Completions & Responses): \"exceeds the context window\" or \"exceeds the model's maximum context length of X tokens\"\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 status code (no body)\n * - Mistral: \"Prompt contains X tokens ... too large for model with Y maximum context length\"\n * - OpenRouter (most backends): \"maximum context length is X tokens\"\n * - OpenRouter/Poolside: \"Input length X exceeds the maximum allowed input length of Y tokens.\"\n * - Together AI: \"The input (X tokens) is longer than the model's context length (Y tokens).\"\n * - llama.cpp: \"exceeds the available context size\"\n * - LM Studio: \"greater than the context length\"\n * - Kimi For Coding: \"exceeded model token limit: X (requested: Y)\"\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 * - Xiaomi MiMo: Truncates input to fit contextWindow then returns stopReason \"length\" with\n *   output=0. Pass contextWindow param to detect via the \"filled context + zero output\" signal.\n * - Ollama: May truncate input silently for some setups, but may also return explicit\n *   overflow errors that match the patterns above. Silent truncation still cannot be\n *   detected here because we do not know the expected token count.\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// Skip messages matching known non-overflow patterns (e.g. throttling / rate-limit)\n\t\tconst isNonOverflow = NON_OVERFLOW_PATTERNS.some((p) => p.test(message.errorMessage!));\n\t\tif (!isNonOverflow && OVERFLOW_PATTERNS.some((p) => p.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\t// Case 3: Length-stop overflow (Xiaomi MiMo style) - server truncates oversized input\n\t// to fit the context window, leaving no room for output. Returns stopReason \"length\"\n\t// with output=0 and input+cacheRead filling the context window.\n\tif (contextWindow && message.stopReason === \"length\" && message.usage.output === 0) {\n\t\tconst inputTokens = message.usage.input + message.usage.cacheRead;\n\t\tif (inputTokens >= contextWindow * 0.99) {\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;AA4EpD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAkDG;AACH,wBAAgB,iBAAiB,CAAC,OAAO,EAAE,gBAAgB,EAAE,aAAa,CAAC,EAAE,MAAM,GAAG,OAAO,CA6B5F;AAED;;GAEG;AACH,wBAAgB,mBAAmB,IAAI,MAAM,EAAE,CAE9C","sourcesContent":["import type { AssistantMessage } from \"../types.ts\";\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 * - Anthropic: \"413 {\\\"error\\\":{\\\"type\\\":\\\"request_too_large\\\",\\\"message\\\":\\\"Request exceeds the maximum size\\\"}}\"\n * - OpenAI: \"Your input exceeds the context window of this model\"\n * - OpenAI/LiteLLM: \"Requested token count exceeds the model's maximum context length of 131072 tokens\"\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 * - OpenRouter/Poolside: \"Input length X exceeds the maximum allowed input length of Y tokens.\"\n * - Together AI: \"The input (X tokens) is longer than the model's context length (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 * - MiniMax: \"invalid params, context window exceeds limit\"\n * - Kimi For Coding: \"Your request exceeded model token limit: X (requested: Y)\"\n * - DS4: \"Prompt has X tokens, but the configured context size is Y tokens\"\n * - Cerebras: \"400/413 status code (no body)\"\n * - Mistral: \"Prompt contains X tokens ... too large for model with Y maximum context length\"\n * - z.ai: Does NOT error, accepts overflow silently - handled via usage.input > contextWindow\n * - Xiaomi MiMo: Truncates input to fill contextWindow exactly, then returns finish_reason \"length\"\n *   with output=0 (no room left to generate). Detected via stopReason \"length\" + zero output +\n *   input filling the context window.\n * - Ollama: Some deployments truncate silently, others return errors like \"prompt too long; exceeded max context length by X tokens\"\n */\nconst OVERFLOW_PATTERNS = [\n\t/prompt is too long/i, // Anthropic token overflow\n\t/request_too_large/i, // Anthropic request byte-size overflow (HTTP 413)\n\t/input is too long for requested model/i, // Amazon Bedrock\n\t/exceeds the context window/i, // OpenAI (Completions & Responses API)\n\t/exceeds (?:the )?(?:model'?s )?maximum context length of [\\d,]+ tokens?/i, // OpenAI-compatible proxies (LiteLLM)\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 (most backends)\n\t/exceeds (?:the )?maximum allowed input length of [\\d,]+ tokens?/i, // OpenRouter/Poolside\n\t/input \\(\\d+ tokens\\) is longer than the model'?s context length \\(\\d+ tokens\\)/i, // Together AI\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 window exceeds limit/i, // MiniMax\n\t/exceeded model token limit/i, // Kimi For Coding\n\t/too large for model with \\d+ maximum context length/i, // Mistral\n\t/prompt has [\\d,]+ tokens?, but the configured context size is [\\d,]+ tokens?/i, // DS4 server\n\t/model_context_window_exceeded/i, // z.ai non-standard finish_reason surfaced as error text\n\t/prompt too long; exceeded (?:max )?context length/i, // Ollama explicit overflow error\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\t/^4(?:00|13)\\s*(?:status code)?\\s*\\(no body\\)/i, // Cerebras: 400/413 with no body\n];\n\n/**\n * Patterns that indicate non-overflow errors (e.g. rate limiting, server errors).\n * Error messages matching any of these are excluded from overflow detection\n * even if they also match an OVERFLOW_PATTERN.\n *\n * Example: Bedrock formats throttling errors as \"ThrottlingException: Too many tokens,\n * please wait before trying again.\" which would match the /too many tokens/i overflow\n * pattern without this exclusion.\n */\nconst NON_OVERFLOW_PATTERNS = [\n\t/^(Throttling error|Service unavailable):/i, // AWS Bedrock non-overflow errors (human-readable prefixes from formatBedrockError)\n\t/rate limit/i, // Generic rate limiting\n\t/too many requests/i, // Generic HTTP 429 style\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\" or \"request_too_large\"\n * - OpenAI (Completions & Responses): \"exceeds the context window\" or \"exceeds the model's maximum context length of X tokens\"\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 status code (no body)\n * - Mistral: \"Prompt contains X tokens ... too large for model with Y maximum context length\"\n * - OpenRouter (most backends): \"maximum context length is X tokens\"\n * - OpenRouter/Poolside: \"Input length X exceeds the maximum allowed input length of Y tokens.\"\n * - Together AI: \"The input (X tokens) is longer than the model's context length (Y tokens).\"\n * - llama.cpp: \"exceeds the available context size\"\n * - LM Studio: \"greater than the context length\"\n * - Kimi For Coding: \"exceeded model token limit: X (requested: Y)\"\n * - DS4: \"Prompt has X tokens, but the configured context size is Y tokens\"\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 * - Xiaomi MiMo: Truncates input to fit contextWindow then returns stopReason \"length\" with\n *   output=0. Pass contextWindow param to detect via the \"filled context + zero output\" signal.\n * - Ollama: May truncate input silently for some setups, but may also return explicit\n *   overflow errors that match the patterns above. Silent truncation still cannot be\n *   detected here because we do not know the expected token count.\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// Skip messages matching known non-overflow patterns (e.g. throttling / rate-limit)\n\t\tconst isNonOverflow = NON_OVERFLOW_PATTERNS.some((p) => p.test(message.errorMessage!));\n\t\tif (!isNonOverflow && OVERFLOW_PATTERNS.some((p) => p.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\t// Case 3: Length-stop overflow (Xiaomi MiMo style) - server truncates oversized input\n\t// to fit the context window, leaving no room for output. Returns stopReason \"length\"\n\t// with output=0 and input+cacheRead filling the context window.\n\tif (contextWindow && message.stopReason === \"length\" && message.usage.output === 0) {\n\t\tconst inputTokens = message.usage.input + message.usage.cacheRead;\n\t\tif (inputTokens >= contextWindow * 0.99) {\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

@@ -21,6 +21,7 @@
  * - GitHub Copilot: "prompt token count of X exceeds the limit of Y"
  * - MiniMax: "invalid params, context window exceeds limit"
  * - Kimi For Coding: "Your request exceeded model token limit: X (requested: Y)"
+ * - DS4: "Prompt has X tokens, but the configured context size is Y tokens"
  * - Cerebras: "400/413 status code (no body)"
  * - Mistral: "Prompt contains X tokens ... too large for model with Y maximum context length"
  * - z.ai: Does NOT error, accepts overflow silently - handled via usage.input > contextWindow
@@ -47,6 +48,7 @@ const OVERFLOW_PATTERNS = [
     /context window exceeds limit/i, // MiniMax
     /exceeded model token limit/i, // Kimi For Coding
     /too large for model with \d+ maximum context length/i, // Mistral
+    /prompt has [\d,]+ tokens?, but the configured context size is [\d,]+ tokens?/i, // DS4 server
     /model_context_window_exceeded/i, // z.ai non-standard finish_reason surfaced as error text
     /prompt too long; exceeded (?:max )?context length/i, // Ollama explicit overflow error
     /context[_ ]length[_ ]exceeded/i, // Generic fallback
@@ -93,6 +95,7 @@ const NON_OVERFLOW_PATTERNS = [
  * - llama.cpp: "exceeds the available context size"
  * - LM Studio: "greater than the context length"
  * - Kimi For Coding: "exceeded model token limit: X (requested: Y)"
+ * - DS4: "Prompt has X tokens, but the configured context size is Y tokens"
  *
  * **Unreliable detection:**
  * - z.ai: Sometimes accepts overflow silently (detectable via usage.input > contextWindow),

package/dist/utils/overflow.js.map

@@ -1 +1 @@
-{"version":3,"file":"overflow.js","sourceRoot":"","sources":["../../src/utils/overflow.ts"],"names":[],"mappings":"AAEA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8BG;AACH,MAAM,iBAAiB,GAAG;IACzB,qBAAqB,EAAE,2BAA2B;IAClD,oBAAoB,EAAE,kDAAkD;IACxE,wCAAwC,EAAE,iBAAiB;IAC3D,6BAA6B,EAAE,uCAAuC;IACtE,0EAA0E,EAAE,sCAAsC;IAClH,yCAAyC,EAAE,kBAAkB;IAC7D,+BAA+B,EAAE,aAAa;IAC9C,oCAAoC,EAAE,OAAO;IAC7C,uCAAuC,EAAE,6BAA6B;IACtE,kEAAkE,EAAE,sBAAsB;IAC1F,iFAAiF,EAAE,cAAc;IACjG,2BAA2B,EAAE,iBAAiB;IAC9C,qCAAqC,EAAE,mBAAmB;IAC1D,kCAAkC,EAAE,YAAY;IAChD,+BAA+B,EAAE,UAAU;IAC3C,6BAA6B,EAAE,kBAAkB;IACjD,sDAAsD,EAAE,UAAU;IAClE,gCAAgC,EAAE,yDAAyD;IAC3F,oDAAoD,EAAE,iCAAiC;IACvF,gCAAgC,EAAE,mBAAmB;IACrD,kBAAkB,EAAE,mBAAmB;IACvC,uBAAuB,EAAE,mBAAmB;IAC5C,+CAA+C,EAAE,iCAAiC;CAClF,CAAC;AAEF;;;;;;;;GAQG;AACH,MAAM,qBAAqB,GAAG;IAC7B,2CAA2C,EAAE,oFAAoF;IACjI,aAAa,EAAE,wBAAwB;IACvC,oBAAoB,EAAE,yBAAyB;CAC/C,CAAC;AAEF;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAiDG;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,oFAAoF;QACpF,MAAM,aAAa,GAAG,qBAAqB,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,OAAO,CAAC,YAAa,CAAC,CAAC,CAAC;QACvF,IAAI,CAAC,aAAa,IAAI,iBAAiB,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,OAAO,CAAC,YAAa,CAAC,CAAC,EAAE,CAAC;YACpF,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,sFAAsF;IACtF,qFAAqF;IACrF,gEAAgE;IAChE,IAAI,aAAa,IAAI,OAAO,CAAC,UAAU,KAAK,QAAQ,IAAI,OAAO,CAAC,KAAK,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QACpF,MAAM,WAAW,GAAG,OAAO,CAAC,KAAK,CAAC,KAAK,GAAG,OAAO,CAAC,KAAK,CAAC,SAAS,CAAC;QAClE,IAAI,WAAW,IAAI,aAAa,GAAG,IAAI,EAAE,CAAC;YACzC,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.ts\";\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 * - Anthropic: \"413 {\\\"error\\\":{\\\"type\\\":\\\"request_too_large\\\",\\\"message\\\":\\\"Request exceeds the maximum size\\\"}}\"\n * - OpenAI: \"Your input exceeds the context window of this model\"\n * - OpenAI/LiteLLM: \"Requested token count exceeds the model's maximum context length of 131072 tokens\"\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 * - OpenRouter/Poolside: \"Input length X exceeds the maximum allowed input length of Y tokens.\"\n * - Together AI: \"The input (X tokens) is longer than the model's context length (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 * - MiniMax: \"invalid params, context window exceeds limit\"\n * - Kimi For Coding: \"Your request exceeded model token limit: X (requested: Y)\"\n * - Cerebras: \"400/413 status code (no body)\"\n * - Mistral: \"Prompt contains X tokens ... too large for model with Y maximum context length\"\n * - z.ai: Does NOT error, accepts overflow silently - handled via usage.input > contextWindow\n * - Xiaomi MiMo: Truncates input to fill contextWindow exactly, then returns finish_reason \"length\"\n *   with output=0 (no room left to generate). Detected via stopReason \"length\" + zero output +\n *   input filling the context window.\n * - Ollama: Some deployments truncate silently, others return errors like \"prompt too long; exceeded max context length by X tokens\"\n */\nconst OVERFLOW_PATTERNS = [\n\t/prompt is too long/i, // Anthropic token overflow\n\t/request_too_large/i, // Anthropic request byte-size overflow (HTTP 413)\n\t/input is too long for requested model/i, // Amazon Bedrock\n\t/exceeds the context window/i, // OpenAI (Completions & Responses API)\n\t/exceeds (?:the )?(?:model'?s )?maximum context length of [\\d,]+ tokens?/i, // OpenAI-compatible proxies (LiteLLM)\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 (most backends)\n\t/exceeds (?:the )?maximum allowed input length of [\\d,]+ tokens?/i, // OpenRouter/Poolside\n\t/input \\(\\d+ tokens\\) is longer than the model'?s context length \\(\\d+ tokens\\)/i, // Together AI\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 window exceeds limit/i, // MiniMax\n\t/exceeded model token limit/i, // Kimi For Coding\n\t/too large for model with \\d+ maximum context length/i, // Mistral\n\t/model_context_window_exceeded/i, // z.ai non-standard finish_reason surfaced as error text\n\t/prompt too long; exceeded (?:max )?context length/i, // Ollama explicit overflow error\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\t/^4(?:00|13)\\s*(?:status code)?\\s*\\(no body\\)/i, // Cerebras: 400/413 with no body\n];\n\n/**\n * Patterns that indicate non-overflow errors (e.g. rate limiting, server errors).\n * Error messages matching any of these are excluded from overflow detection\n * even if they also match an OVERFLOW_PATTERN.\n *\n * Example: Bedrock formats throttling errors as \"ThrottlingException: Too many tokens,\n * please wait before trying again.\" which would match the /too many tokens/i overflow\n * pattern without this exclusion.\n */\nconst NON_OVERFLOW_PATTERNS = [\n\t/^(Throttling error|Service unavailable):/i, // AWS Bedrock non-overflow errors (human-readable prefixes from formatBedrockError)\n\t/rate limit/i, // Generic rate limiting\n\t/too many requests/i, // Generic HTTP 429 style\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\" or \"request_too_large\"\n * - OpenAI (Completions & Responses): \"exceeds the context window\" or \"exceeds the model's maximum context length of X tokens\"\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 status code (no body)\n * - Mistral: \"Prompt contains X tokens ... too large for model with Y maximum context length\"\n * - OpenRouter (most backends): \"maximum context length is X tokens\"\n * - OpenRouter/Poolside: \"Input length X exceeds the maximum allowed input length of Y tokens.\"\n * - Together AI: \"The input (X tokens) is longer than the model's context length (Y tokens).\"\n * - llama.cpp: \"exceeds the available context size\"\n * - LM Studio: \"greater than the context length\"\n * - Kimi For Coding: \"exceeded model token limit: X (requested: Y)\"\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 * - Xiaomi MiMo: Truncates input to fit contextWindow then returns stopReason \"length\" with\n *   output=0. Pass contextWindow param to detect via the \"filled context + zero output\" signal.\n * - Ollama: May truncate input silently for some setups, but may also return explicit\n *   overflow errors that match the patterns above. Silent truncation still cannot be\n *   detected here because we do not know the expected token count.\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// Skip messages matching known non-overflow patterns (e.g. throttling / rate-limit)\n\t\tconst isNonOverflow = NON_OVERFLOW_PATTERNS.some((p) => p.test(message.errorMessage!));\n\t\tif (!isNonOverflow && OVERFLOW_PATTERNS.some((p) => p.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\t// Case 3: Length-stop overflow (Xiaomi MiMo style) - server truncates oversized input\n\t// to fit the context window, leaving no room for output. Returns stopReason \"length\"\n\t// with output=0 and input+cacheRead filling the context window.\n\tif (contextWindow && message.stopReason === \"length\" && message.usage.output === 0) {\n\t\tconst inputTokens = message.usage.input + message.usage.cacheRead;\n\t\tif (inputTokens >= contextWindow * 0.99) {\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;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA+BG;AACH,MAAM,iBAAiB,GAAG;IACzB,qBAAqB,EAAE,2BAA2B;IAClD,oBAAoB,EAAE,kDAAkD;IACxE,wCAAwC,EAAE,iBAAiB;IAC3D,6BAA6B,EAAE,uCAAuC;IACtE,0EAA0E,EAAE,sCAAsC;IAClH,yCAAyC,EAAE,kBAAkB;IAC7D,+BAA+B,EAAE,aAAa;IAC9C,oCAAoC,EAAE,OAAO;IAC7C,uCAAuC,EAAE,6BAA6B;IACtE,kEAAkE,EAAE,sBAAsB;IAC1F,iFAAiF,EAAE,cAAc;IACjG,2BAA2B,EAAE,iBAAiB;IAC9C,qCAAqC,EAAE,mBAAmB;IAC1D,kCAAkC,EAAE,YAAY;IAChD,+BAA+B,EAAE,UAAU;IAC3C,6BAA6B,EAAE,kBAAkB;IACjD,sDAAsD,EAAE,UAAU;IAClE,+EAA+E,EAAE,aAAa;IAC9F,gCAAgC,EAAE,yDAAyD;IAC3F,oDAAoD,EAAE,iCAAiC;IACvF,gCAAgC,EAAE,mBAAmB;IACrD,kBAAkB,EAAE,mBAAmB;IACvC,uBAAuB,EAAE,mBAAmB;IAC5C,+CAA+C,EAAE,iCAAiC;CAClF,CAAC;AAEF;;;;;;;;GAQG;AACH,MAAM,qBAAqB,GAAG;IAC7B,2CAA2C,EAAE,oFAAoF;IACjI,aAAa,EAAE,wBAAwB;IACvC,oBAAoB,EAAE,yBAAyB;CAC/C,CAAC;AAEF;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAkDG;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,oFAAoF;QACpF,MAAM,aAAa,GAAG,qBAAqB,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,OAAO,CAAC,YAAa,CAAC,CAAC,CAAC;QACvF,IAAI,CAAC,aAAa,IAAI,iBAAiB,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,OAAO,CAAC,YAAa,CAAC,CAAC,EAAE,CAAC;YACpF,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,sFAAsF;IACtF,qFAAqF;IACrF,gEAAgE;IAChE,IAAI,aAAa,IAAI,OAAO,CAAC,UAAU,KAAK,QAAQ,IAAI,OAAO,CAAC,KAAK,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QACpF,MAAM,WAAW,GAAG,OAAO,CAAC,KAAK,CAAC,KAAK,GAAG,OAAO,CAAC,KAAK,CAAC,SAAS,CAAC;QAClE,IAAI,WAAW,IAAI,aAAa,GAAG,IAAI,EAAE,CAAC;YACzC,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.ts\";\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 * - Anthropic: \"413 {\\\"error\\\":{\\\"type\\\":\\\"request_too_large\\\",\\\"message\\\":\\\"Request exceeds the maximum size\\\"}}\"\n * - OpenAI: \"Your input exceeds the context window of this model\"\n * - OpenAI/LiteLLM: \"Requested token count exceeds the model's maximum context length of 131072 tokens\"\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 * - OpenRouter/Poolside: \"Input length X exceeds the maximum allowed input length of Y tokens.\"\n * - Together AI: \"The input (X tokens) is longer than the model's context length (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 * - MiniMax: \"invalid params, context window exceeds limit\"\n * - Kimi For Coding: \"Your request exceeded model token limit: X (requested: Y)\"\n * - DS4: \"Prompt has X tokens, but the configured context size is Y tokens\"\n * - Cerebras: \"400/413 status code (no body)\"\n * - Mistral: \"Prompt contains X tokens ... too large for model with Y maximum context length\"\n * - z.ai: Does NOT error, accepts overflow silently - handled via usage.input > contextWindow\n * - Xiaomi MiMo: Truncates input to fill contextWindow exactly, then returns finish_reason \"length\"\n *   with output=0 (no room left to generate). Detected via stopReason \"length\" + zero output +\n *   input filling the context window.\n * - Ollama: Some deployments truncate silently, others return errors like \"prompt too long; exceeded max context length by X tokens\"\n */\nconst OVERFLOW_PATTERNS = [\n\t/prompt is too long/i, // Anthropic token overflow\n\t/request_too_large/i, // Anthropic request byte-size overflow (HTTP 413)\n\t/input is too long for requested model/i, // Amazon Bedrock\n\t/exceeds the context window/i, // OpenAI (Completions & Responses API)\n\t/exceeds (?:the )?(?:model'?s )?maximum context length of [\\d,]+ tokens?/i, // OpenAI-compatible proxies (LiteLLM)\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 (most backends)\n\t/exceeds (?:the )?maximum allowed input length of [\\d,]+ tokens?/i, // OpenRouter/Poolside\n\t/input \\(\\d+ tokens\\) is longer than the model'?s context length \\(\\d+ tokens\\)/i, // Together AI\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 window exceeds limit/i, // MiniMax\n\t/exceeded model token limit/i, // Kimi For Coding\n\t/too large for model with \\d+ maximum context length/i, // Mistral\n\t/prompt has [\\d,]+ tokens?, but the configured context size is [\\d,]+ tokens?/i, // DS4 server\n\t/model_context_window_exceeded/i, // z.ai non-standard finish_reason surfaced as error text\n\t/prompt too long; exceeded (?:max )?context length/i, // Ollama explicit overflow error\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\t/^4(?:00|13)\\s*(?:status code)?\\s*\\(no body\\)/i, // Cerebras: 400/413 with no body\n];\n\n/**\n * Patterns that indicate non-overflow errors (e.g. rate limiting, server errors).\n * Error messages matching any of these are excluded from overflow detection\n * even if they also match an OVERFLOW_PATTERN.\n *\n * Example: Bedrock formats throttling errors as \"ThrottlingException: Too many tokens,\n * please wait before trying again.\" which would match the /too many tokens/i overflow\n * pattern without this exclusion.\n */\nconst NON_OVERFLOW_PATTERNS = [\n\t/^(Throttling error|Service unavailable):/i, // AWS Bedrock non-overflow errors (human-readable prefixes from formatBedrockError)\n\t/rate limit/i, // Generic rate limiting\n\t/too many requests/i, // Generic HTTP 429 style\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\" or \"request_too_large\"\n * - OpenAI (Completions & Responses): \"exceeds the context window\" or \"exceeds the model's maximum context length of X tokens\"\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 status code (no body)\n * - Mistral: \"Prompt contains X tokens ... too large for model with Y maximum context length\"\n * - OpenRouter (most backends): \"maximum context length is X tokens\"\n * - OpenRouter/Poolside: \"Input length X exceeds the maximum allowed input length of Y tokens.\"\n * - Together AI: \"The input (X tokens) is longer than the model's context length (Y tokens).\"\n * - llama.cpp: \"exceeds the available context size\"\n * - LM Studio: \"greater than the context length\"\n * - Kimi For Coding: \"exceeded model token limit: X (requested: Y)\"\n * - DS4: \"Prompt has X tokens, but the configured context size is Y tokens\"\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 * - Xiaomi MiMo: Truncates input to fit contextWindow then returns stopReason \"length\" with\n *   output=0. Pass contextWindow param to detect via the \"filled context + zero output\" signal.\n * - Ollama: May truncate input silently for some setups, but may also return explicit\n *   overflow errors that match the patterns above. Silent truncation still cannot be\n *   detected here because we do not know the expected token count.\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// Skip messages matching known non-overflow patterns (e.g. throttling / rate-limit)\n\t\tconst isNonOverflow = NON_OVERFLOW_PATTERNS.some((p) => p.test(message.errorMessage!));\n\t\tif (!isNonOverflow && OVERFLOW_PATTERNS.some((p) => p.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\t// Case 3: Length-stop overflow (Xiaomi MiMo style) - server truncates oversized input\n\t// to fit the context window, leaving no room for output. Returns stopReason \"length\"\n\t// with output=0 and input+cacheRead filling the context window.\n\tif (contextWindow && message.stopReason === \"length\" && message.usage.output === 0) {\n\t\tconst inputTokens = message.usage.input + message.usage.cacheRead;\n\t\tif (inputTokens >= contextWindow * 0.99) {\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/retry.d.ts

@@ -0,0 +1,3 @@
+import type { AssistantMessage } from "../types.ts";
+export declare function isRetryableAssistantError(message: AssistantMessage): boolean;
+//# sourceMappingURL=retry.d.ts.map

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

@@ -0,0 +1 @@
+{"version":3,"file":"retry.d.ts","sourceRoot":"","sources":["../../src/utils/retry.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,aAAa,CAAC;AAsDpD,wBAAgB,yBAAyB,CAAC,OAAO,EAAE,gBAAgB,GAAG,OAAO,CAK5E","sourcesContent":["import type { AssistantMessage } from \"../types.ts\";\n\nfunction buildProviderErrorPattern(patterns: readonly string[]): RegExp {\n\treturn new RegExp(patterns.join(\"|\"), \"i\");\n}\n\nconst NON_RETRYABLE_PROVIDER_LIMIT_ERROR_PATTERN = buildProviderErrorPattern([\n\t\"GoUsageLimitError\",\n\t\"FreeUsageLimitError\",\n\t\"Monthly usage limit reached\",\n\t\"available balance\",\n\t\"insufficient_quota\",\n\t\"out of budget\",\n\t\"quota exceeded\",\n\t\"billing\",\n]);\n\nconst RETRYABLE_PROVIDER_ERROR_PATTERN = buildProviderErrorPattern([\n\t\"overloaded\",\n\t\"rate.?limit\",\n\t\"too many requests\",\n\t\"429\",\n\t\"500\",\n\t\"502\",\n\t\"503\",\n\t\"504\",\n\t\"524\",\n\t\"service.?unavailable\",\n\t\"server.?error\",\n\t\"internal.?error\",\n\t\"provider.?returned.?error\",\n\t\"network.?error\",\n\t\"connection.?error\",\n\t\"connection.?refused\",\n\t\"connection.?lost\",\n\t\"other side closed\",\n\t\"fetch failed\",\n\t\"upstream.?connect\",\n\t\"reset before headers\",\n\t\"socket hang up\",\n\t\"timed? out\",\n\t\"timeout\",\n\t\"terminated\",\n\t\"websocket.?closed\",\n\t\"websocket.?error\",\n\t\"ended without\",\n\t\"stream ended before message_stop\",\n\t\"http2 request did not get a response\",\n\t\"retry delay\",\n\t\"you can retry your request\",\n\t\"try your request again\",\n\t\"please retry your request\",\n]);\n\nexport function isRetryableAssistantError(message: AssistantMessage): boolean {\n\tif (message.stopReason !== \"error\" || !message.errorMessage) return false;\n\tconst errorMessage = message.errorMessage;\n\tif (NON_RETRYABLE_PROVIDER_LIMIT_ERROR_PATTERN.test(errorMessage)) return false;\n\treturn RETRYABLE_PROVIDER_ERROR_PATTERN.test(errorMessage);\n}\n"]}

package/dist/utils/retry.js

@@ -0,0 +1,58 @@
+function buildProviderErrorPattern(patterns) {
+    return new RegExp(patterns.join("|"), "i");
+}
+const NON_RETRYABLE_PROVIDER_LIMIT_ERROR_PATTERN = buildProviderErrorPattern([
+    "GoUsageLimitError",
+    "FreeUsageLimitError",
+    "Monthly usage limit reached",
+    "available balance",
+    "insufficient_quota",
+    "out of budget",
+    "quota exceeded",
+    "billing",
+]);
+const RETRYABLE_PROVIDER_ERROR_PATTERN = buildProviderErrorPattern([
+    "overloaded",
+    "rate.?limit",
+    "too many requests",
+    "429",
+    "500",
+    "502",
+    "503",
+    "504",
+    "524",
+    "service.?unavailable",
+    "server.?error",
+    "internal.?error",
+    "provider.?returned.?error",
+    "network.?error",
+    "connection.?error",
+    "connection.?refused",
+    "connection.?lost",
+    "other side closed",
+    "fetch failed",
+    "upstream.?connect",
+    "reset before headers",
+    "socket hang up",
+    "timed? out",
+    "timeout",
+    "terminated",
+    "websocket.?closed",
+    "websocket.?error",
+    "ended without",
+    "stream ended before message_stop",
+    "http2 request did not get a response",
+    "retry delay",
+    "you can retry your request",
+    "try your request again",
+    "please retry your request",
+]);
+export function isRetryableAssistantError(message) {
+    if (message.stopReason !== "error" || !message.errorMessage)
+        return false;
+    const errorMessage = message.errorMessage;
+    if (NON_RETRYABLE_PROVIDER_LIMIT_ERROR_PATTERN.test(errorMessage))
+        return false;
+    return RETRYABLE_PROVIDER_ERROR_PATTERN.test(errorMessage);
+}
+//# sourceMappingURL=retry.js.map

package/dist/utils/retry.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"retry.js","sourceRoot":"","sources":["../../src/utils/retry.ts"],"names":[],"mappings":"AAEA,SAAS,yBAAyB,CAAC,QAA2B,EAAU;IACvE,OAAO,IAAI,MAAM,CAAC,QAAQ,CAAC,IAAI,CAAC,GAAG,CAAC,EAAE,GAAG,CAAC,CAAC;AAAA,CAC3C;AAED,MAAM,0CAA0C,GAAG,yBAAyB,CAAC;IAC5E,mBAAmB;IACnB,qBAAqB;IACrB,6BAA6B;IAC7B,mBAAmB;IACnB,oBAAoB;IACpB,eAAe;IACf,gBAAgB;IAChB,SAAS;CACT,CAAC,CAAC;AAEH,MAAM,gCAAgC,GAAG,yBAAyB,CAAC;IAClE,YAAY;IACZ,aAAa;IACb,mBAAmB;IACnB,KAAK;IACL,KAAK;IACL,KAAK;IACL,KAAK;IACL,KAAK;IACL,KAAK;IACL,sBAAsB;IACtB,eAAe;IACf,iBAAiB;IACjB,2BAA2B;IAC3B,gBAAgB;IAChB,mBAAmB;IACnB,qBAAqB;IACrB,kBAAkB;IAClB,mBAAmB;IACnB,cAAc;IACd,mBAAmB;IACnB,sBAAsB;IACtB,gBAAgB;IAChB,YAAY;IACZ,SAAS;IACT,YAAY;IACZ,mBAAmB;IACnB,kBAAkB;IAClB,eAAe;IACf,kCAAkC;IAClC,sCAAsC;IACtC,aAAa;IACb,4BAA4B;IAC5B,wBAAwB;IACxB,2BAA2B;CAC3B,CAAC,CAAC;AAEH,MAAM,UAAU,yBAAyB,CAAC,OAAyB,EAAW;IAC7E,IAAI,OAAO,CAAC,UAAU,KAAK,OAAO,IAAI,CAAC,OAAO,CAAC,YAAY;QAAE,OAAO,KAAK,CAAC;IAC1E,MAAM,YAAY,GAAG,OAAO,CAAC,YAAY,CAAC;IAC1C,IAAI,0CAA0C,CAAC,IAAI,CAAC,YAAY,CAAC;QAAE,OAAO,KAAK,CAAC;IAChF,OAAO,gCAAgC,CAAC,IAAI,CAAC,YAAY,CAAC,CAAC;AAAA,CAC3D","sourcesContent":["import type { AssistantMessage } from \"../types.ts\";\n\nfunction buildProviderErrorPattern(patterns: readonly string[]): RegExp {\n\treturn new RegExp(patterns.join(\"|\"), \"i\");\n}\n\nconst NON_RETRYABLE_PROVIDER_LIMIT_ERROR_PATTERN = buildProviderErrorPattern([\n\t\"GoUsageLimitError\",\n\t\"FreeUsageLimitError\",\n\t\"Monthly usage limit reached\",\n\t\"available balance\",\n\t\"insufficient_quota\",\n\t\"out of budget\",\n\t\"quota exceeded\",\n\t\"billing\",\n]);\n\nconst RETRYABLE_PROVIDER_ERROR_PATTERN = buildProviderErrorPattern([\n\t\"overloaded\",\n\t\"rate.?limit\",\n\t\"too many requests\",\n\t\"429\",\n\t\"500\",\n\t\"502\",\n\t\"503\",\n\t\"504\",\n\t\"524\",\n\t\"service.?unavailable\",\n\t\"server.?error\",\n\t\"internal.?error\",\n\t\"provider.?returned.?error\",\n\t\"network.?error\",\n\t\"connection.?error\",\n\t\"connection.?refused\",\n\t\"connection.?lost\",\n\t\"other side closed\",\n\t\"fetch failed\",\n\t\"upstream.?connect\",\n\t\"reset before headers\",\n\t\"socket hang up\",\n\t\"timed? out\",\n\t\"timeout\",\n\t\"terminated\",\n\t\"websocket.?closed\",\n\t\"websocket.?error\",\n\t\"ended without\",\n\t\"stream ended before message_stop\",\n\t\"http2 request did not get a response\",\n\t\"retry delay\",\n\t\"you can retry your request\",\n\t\"try your request again\",\n\t\"please retry your request\",\n]);\n\nexport function isRetryableAssistantError(message: AssistantMessage): boolean {\n\tif (message.stopReason !== \"error\" || !message.errorMessage) return false;\n\tconst errorMessage = message.errorMessage;\n\tif (NON_RETRYABLE_PROVIDER_LIMIT_ERROR_PATTERN.test(errorMessage)) return false;\n\treturn RETRYABLE_PROVIDER_ERROR_PATTERN.test(errorMessage);\n}\n"]}

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@fleetagent/pi-ai",
-	"version": "0.1.4",
+	"version": "0.1.7",
 	"description": "Unified LLM API with automatic model discovery and provider configuration",
 	"type": "module",
 	"main": "./dist/index.js",
@@ -71,7 +71,8 @@
 		"@aws-sdk/client-bedrock-runtime": "3.1048.0",
 		"@smithy/node-http-handler": "4.7.3",
 		"@google/genai": "1.52.0",
-		"@mistralai/mistralai": "2.2.1",
+		"@mistralai/mistralai": "2.2.6",
+		"@opentelemetry/api": "1.9.0",
 		"http-proxy-agent": "7.0.2",
 		"https-proxy-agent": "7.0.6",
 		"openai": "6.26.0",
@@ -101,6 +102,6 @@
 	"devDependencies": {
 		"@types/node": "24.12.4",
 		"canvas": "3.2.3",
-		"vitest": "3.2.4"
+		"vitest": "3.2.7"
 	}
 }