@entelligentsia/forgecli 0.7.6 → 0.7.9

package/node_modules/@earendil-works/pi-ai/dist/utils/node-http-proxy.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"node-http-proxy.d.ts","sourceRoot":"","sources":["../../src/utils/node-http-proxy.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,KAAK,IAAI,SAAS,EAAE,MAAM,WAAW,CAAC;AACpD,OAAO,KAAK,EAAE,KAAK,IAAI,UAAU,EAAE,MAAM,YAAY,CAAC;AAatD,MAAM,WAAW,mBAAmB;IACnC,SAAS,EAAE,SAAS,CAAC;IACrB,UAAU,EAAE,UAAU,CAAC;CACvB;AAED,eAAO,MAAM,kCAAkC,4GAC2D,CAAC;AAsE3G,wBAAgB,4BAA4B,CAAC,SAAS,EAAE,MAAM,GAAG,GAAG,GAAG,GAAG,GAAG,SAAS,CAoBrF;AAED,wBAAgB,8BAA8B,CAAC,SAAS,EAAE,MAAM,GAAG,GAAG,GAAG,mBAAmB,GAAG,SAAS,CAUvG","sourcesContent":["import type { Agent as HttpAgent } from \"node:http\";\nimport type { Agent as HttpsAgent } from \"node:https\";\nimport { HttpProxyAgent } from \"http-proxy-agent\";\nimport { HttpsProxyAgent } from \"https-proxy-agent\";\n\nconst DEFAULT_PROXY_PORTS: Record<string, number> = {\n\tftp: 21,\n\tgopher: 70,\n\thttp: 80,\n\thttps: 443,\n\tws: 80,\n\twss: 443,\n};\n\nexport interface NodeHttpProxyAgents {\n\thttpAgent: HttpAgent;\n\thttpsAgent: HttpsAgent;\n}\n\nexport const UNSUPPORTED_PROXY_PROTOCOL_MESSAGE =\n\t\"Unsupported proxy protocol. SOCKS and PAC proxy URLs are not supported; use an HTTP or HTTPS proxy URL.\";\n\nfunction getProxyEnv(key: string): string {\n\treturn process.env[key.toLowerCase()] || process.env[key.toUpperCase()] || \"\";\n}\n\nfunction parseProxyTargetUrl(targetUrl: string | URL): URL | undefined {\n\tif (targetUrl instanceof URL) {\n\t\treturn targetUrl;\n\t}\n\n\ttry {\n\t\treturn new URL(targetUrl);\n\t} catch {\n\t\treturn undefined;\n\t}\n}\n\nfunction shouldProxyHostname(hostname: string, port: number): boolean {\n\tconst noProxy = getProxyEnv(\"no_proxy\").toLowerCase();\n\tif (!noProxy) {\n\t\treturn true;\n\t}\n\tif (noProxy === \"*\") {\n\t\treturn false;\n\t}\n\n\treturn noProxy.split(/[,\\s]/).every((proxy) => {\n\t\tif (!proxy) {\n\t\t\treturn true;\n\t\t}\n\n\t\tconst parsedProxy = proxy.match(/^(.+):(\\d+)$/);\n\t\tlet proxyHostname = parsedProxy ? parsedProxy[1] : proxy;\n\t\tconst proxyPort = parsedProxy ? Number.parseInt(parsedProxy[2]!, 10) : 0;\n\t\tif (proxyPort && proxyPort !== port) {\n\t\t\treturn true;\n\t\t}\n\n\t\tif (!/^[.*]/.test(proxyHostname)) {\n\t\t\treturn hostname !== proxyHostname;\n\t\t}\n\n\t\tif (proxyHostname.startsWith(\"*\")) {\n\t\t\tproxyHostname = proxyHostname.slice(1);\n\t\t}\n\t\treturn !hostname.endsWith(proxyHostname);\n\t});\n}\n\nfunction getProxyForUrl(targetUrl: string | URL): string {\n\tconst parsedUrl = parseProxyTargetUrl(targetUrl);\n\tif (!parsedUrl?.protocol || !parsedUrl.host) {\n\t\treturn \"\";\n\t}\n\n\tconst protocol = parsedUrl.protocol.split(\":\", 1)[0]!;\n\tconst hostname = parsedUrl.host.replace(/:\\d*$/, \"\");\n\tconst port = Number.parseInt(parsedUrl.port, 10) || DEFAULT_PROXY_PORTS[protocol] || 0;\n\tif (!shouldProxyHostname(hostname, port)) {\n\t\treturn \"\";\n\t}\n\n\tlet proxy = getProxyEnv(`${protocol}_proxy`) || getProxyEnv(\"all_proxy\");\n\tif (proxy && !proxy.includes(\"://\")) {\n\t\tproxy = `${protocol}://${proxy}`;\n\t}\n\treturn proxy;\n}\n\nexport function resolveHttpProxyUrlForTarget(targetUrl: string | URL): URL | undefined {\n\tconst proxy = getProxyForUrl(targetUrl);\n\tif (!proxy) {\n\t\treturn undefined;\n\t}\n\n\tlet proxyUrl: URL;\n\ttry {\n\t\tproxyUrl = new URL(proxy);\n\t} catch (error) {\n\t\tthrow new Error(\n\t\t\t`Invalid proxy URL ${JSON.stringify(proxy)}: ${error instanceof Error ? error.message : String(error)}`,\n\t\t);\n\t}\n\n\tif (proxyUrl.protocol !== \"http:\" && proxyUrl.protocol !== \"https:\") {\n\t\tthrow new Error(`${UNSUPPORTED_PROXY_PROTOCOL_MESSAGE} Got ${proxyUrl.protocol}`);\n\t}\n\n\treturn proxyUrl;\n}\n\nexport function createHttpProxyAgentsForTarget(targetUrl: string | URL): NodeHttpProxyAgents | undefined {\n\tconst proxyUrl = resolveHttpProxyUrlForTarget(targetUrl);\n\tif (!proxyUrl) {\n\t\treturn undefined;\n\t}\n\n\treturn {\n\t\thttpAgent: new HttpProxyAgent(proxyUrl),\n\t\thttpsAgent: new HttpsProxyAgent(proxyUrl) as unknown as HttpsAgent,\n\t};\n}\n"]}

package/node_modules/@earendil-works/pi-ai/dist/utils/node-http-proxy.js

@@ -0,0 +1,97 @@
+import { HttpProxyAgent } from "http-proxy-agent";
+import { HttpsProxyAgent } from "https-proxy-agent";
+const DEFAULT_PROXY_PORTS = {
+    ftp: 21,
+    gopher: 70,
+    http: 80,
+    https: 443,
+    ws: 80,
+    wss: 443,
+};
+export const UNSUPPORTED_PROXY_PROTOCOL_MESSAGE = "Unsupported proxy protocol. SOCKS and PAC proxy URLs are not supported; use an HTTP or HTTPS proxy URL.";
+function getProxyEnv(key) {
+    return process.env[key.toLowerCase()] || process.env[key.toUpperCase()] || "";
+}
+function parseProxyTargetUrl(targetUrl) {
+    if (targetUrl instanceof URL) {
+        return targetUrl;
+    }
+    try {
+        return new URL(targetUrl);
+    }
+    catch {
+        return undefined;
+    }
+}
+function shouldProxyHostname(hostname, port) {
+    const noProxy = getProxyEnv("no_proxy").toLowerCase();
+    if (!noProxy) {
+        return true;
+    }
+    if (noProxy === "*") {
+        return false;
+    }
+    return noProxy.split(/[,\s]/).every((proxy) => {
+        if (!proxy) {
+            return true;
+        }
+        const parsedProxy = proxy.match(/^(.+):(\d+)$/);
+        let proxyHostname = parsedProxy ? parsedProxy[1] : proxy;
+        const proxyPort = parsedProxy ? Number.parseInt(parsedProxy[2], 10) : 0;
+        if (proxyPort && proxyPort !== port) {
+            return true;
+        }
+        if (!/^[.*]/.test(proxyHostname)) {
+            return hostname !== proxyHostname;
+        }
+        if (proxyHostname.startsWith("*")) {
+            proxyHostname = proxyHostname.slice(1);
+        }
+        return !hostname.endsWith(proxyHostname);
+    });
+}
+function getProxyForUrl(targetUrl) {
+    const parsedUrl = parseProxyTargetUrl(targetUrl);
+    if (!parsedUrl?.protocol || !parsedUrl.host) {
+        return "";
+    }
+    const protocol = parsedUrl.protocol.split(":", 1)[0];
+    const hostname = parsedUrl.host.replace(/:\d*$/, "");
+    const port = Number.parseInt(parsedUrl.port, 10) || DEFAULT_PROXY_PORTS[protocol] || 0;
+    if (!shouldProxyHostname(hostname, port)) {
+        return "";
+    }
+    let proxy = getProxyEnv(`${protocol}_proxy`) || getProxyEnv("all_proxy");
+    if (proxy && !proxy.includes("://")) {
+        proxy = `${protocol}://${proxy}`;
+    }
+    return proxy;
+}
+export function resolveHttpProxyUrlForTarget(targetUrl) {
+    const proxy = getProxyForUrl(targetUrl);
+    if (!proxy) {
+        return undefined;
+    }
+    let proxyUrl;
+    try {
+        proxyUrl = new URL(proxy);
+    }
+    catch (error) {
+        throw new Error(`Invalid proxy URL ${JSON.stringify(proxy)}: ${error instanceof Error ? error.message : String(error)}`);
+    }
+    if (proxyUrl.protocol !== "http:" && proxyUrl.protocol !== "https:") {
+        throw new Error(`${UNSUPPORTED_PROXY_PROTOCOL_MESSAGE} Got ${proxyUrl.protocol}`);
+    }
+    return proxyUrl;
+}
+export function createHttpProxyAgentsForTarget(targetUrl) {
+    const proxyUrl = resolveHttpProxyUrlForTarget(targetUrl);
+    if (!proxyUrl) {
+        return undefined;
+    }
+    return {
+        httpAgent: new HttpProxyAgent(proxyUrl),
+        httpsAgent: new HttpsProxyAgent(proxyUrl),
+    };
+}
+//# sourceMappingURL=node-http-proxy.js.map

package/node_modules/@earendil-works/pi-ai/dist/utils/node-http-proxy.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"node-http-proxy.js","sourceRoot":"","sources":["../../src/utils/node-http-proxy.ts"],"names":[],"mappings":"AAEA,OAAO,EAAE,cAAc,EAAE,MAAM,kBAAkB,CAAC;AAClD,OAAO,EAAE,eAAe,EAAE,MAAM,mBAAmB,CAAC;AAEpD,MAAM,mBAAmB,GAA2B;IACnD,GAAG,EAAE,EAAE;IACP,MAAM,EAAE,EAAE;IACV,IAAI,EAAE,EAAE;IACR,KAAK,EAAE,GAAG;IACV,EAAE,EAAE,EAAE;IACN,GAAG,EAAE,GAAG;CACR,CAAC;AAOF,MAAM,CAAC,MAAM,kCAAkC,GAC9C,yGAAyG,CAAC;AAE3G,SAAS,WAAW,CAAC,GAAW,EAAU;IACzC,OAAO,OAAO,CAAC,GAAG,CAAC,GAAG,CAAC,WAAW,EAAE,CAAC,IAAI,OAAO,CAAC,GAAG,CAAC,GAAG,CAAC,WAAW,EAAE,CAAC,IAAI,EAAE,CAAC;AAAA,CAC9E;AAED,SAAS,mBAAmB,CAAC,SAAuB,EAAmB;IACtE,IAAI,SAAS,YAAY,GAAG,EAAE,CAAC;QAC9B,OAAO,SAAS,CAAC;IAClB,CAAC;IAED,IAAI,CAAC;QACJ,OAAO,IAAI,GAAG,CAAC,SAAS,CAAC,CAAC;IAC3B,CAAC;IAAC,MAAM,CAAC;QACR,OAAO,SAAS,CAAC;IAClB,CAAC;AAAA,CACD;AAED,SAAS,mBAAmB,CAAC,QAAgB,EAAE,IAAY,EAAW;IACrE,MAAM,OAAO,GAAG,WAAW,CAAC,UAAU,CAAC,CAAC,WAAW,EAAE,CAAC;IACtD,IAAI,CAAC,OAAO,EAAE,CAAC;QACd,OAAO,IAAI,CAAC;IACb,CAAC;IACD,IAAI,OAAO,KAAK,GAAG,EAAE,CAAC;QACrB,OAAO,KAAK,CAAC;IACd,CAAC;IAED,OAAO,OAAO,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,KAAK,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC;QAC9C,IAAI,CAAC,KAAK,EAAE,CAAC;YACZ,OAAO,IAAI,CAAC;QACb,CAAC;QAED,MAAM,WAAW,GAAG,KAAK,CAAC,KAAK,CAAC,cAAc,CAAC,CAAC;QAChD,IAAI,aAAa,GAAG,WAAW,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC;QACzD,MAAM,SAAS,GAAG,WAAW,CAAC,CAAC,CAAC,MAAM,CAAC,QAAQ,CAAC,WAAW,CAAC,CAAC,CAAE,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;QACzE,IAAI,SAAS,IAAI,SAAS,KAAK,IAAI,EAAE,CAAC;YACrC,OAAO,IAAI,CAAC;QACb,CAAC;QAED,IAAI,CAAC,OAAO,CAAC,IAAI,CAAC,aAAa,CAAC,EAAE,CAAC;YAClC,OAAO,QAAQ,KAAK,aAAa,CAAC;QACnC,CAAC;QAED,IAAI,aAAa,CAAC,UAAU,CAAC,GAAG,CAAC,EAAE,CAAC;YACnC,aAAa,GAAG,aAAa,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;QACxC,CAAC;QACD,OAAO,CAAC,QAAQ,CAAC,QAAQ,CAAC,aAAa,CAAC,CAAC;IAAA,CACzC,CAAC,CAAC;AAAA,CACH;AAED,SAAS,cAAc,CAAC,SAAuB,EAAU;IACxD,MAAM,SAAS,GAAG,mBAAmB,CAAC,SAAS,CAAC,CAAC;IACjD,IAAI,CAAC,SAAS,EAAE,QAAQ,IAAI,CAAC,SAAS,CAAC,IAAI,EAAE,CAAC;QAC7C,OAAO,EAAE,CAAC;IACX,CAAC;IAED,MAAM,QAAQ,GAAG,SAAS,CAAC,QAAQ,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,CAAE,CAAC;IACtD,MAAM,QAAQ,GAAG,SAAS,CAAC,IAAI,CAAC,OAAO,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC;IACrD,MAAM,IAAI,GAAG,MAAM,CAAC,QAAQ,CAAC,SAAS,CAAC,IAAI,EAAE,EAAE,CAAC,IAAI,mBAAmB,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC;IACvF,IAAI,CAAC,mBAAmB,CAAC,QAAQ,EAAE,IAAI,CAAC,EAAE,CAAC;QAC1C,OAAO,EAAE,CAAC;IACX,CAAC;IAED,IAAI,KAAK,GAAG,WAAW,CAAC,GAAG,QAAQ,QAAQ,CAAC,IAAI,WAAW,CAAC,WAAW,CAAC,CAAC;IACzE,IAAI,KAAK,IAAI,CAAC,KAAK,CAAC,QAAQ,CAAC,KAAK,CAAC,EAAE,CAAC;QACrC,KAAK,GAAG,GAAG,QAAQ,MAAM,KAAK,EAAE,CAAC;IAClC,CAAC;IACD,OAAO,KAAK,CAAC;AAAA,CACb;AAED,MAAM,UAAU,4BAA4B,CAAC,SAAuB,EAAmB;IACtF,MAAM,KAAK,GAAG,cAAc,CAAC,SAAS,CAAC,CAAC;IACxC,IAAI,CAAC,KAAK,EAAE,CAAC;QACZ,OAAO,SAAS,CAAC;IAClB,CAAC;IAED,IAAI,QAAa,CAAC;IAClB,IAAI,CAAC;QACJ,QAAQ,GAAG,IAAI,GAAG,CAAC,KAAK,CAAC,CAAC;IAC3B,CAAC;IAAC,OAAO,KAAK,EAAE,CAAC;QAChB,MAAM,IAAI,KAAK,CACd,qBAAqB,IAAI,CAAC,SAAS,CAAC,KAAK,CAAC,KAAK,KAAK,YAAY,KAAK,CAAC,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,CACvG,CAAC;IACH,CAAC;IAED,IAAI,QAAQ,CAAC,QAAQ,KAAK,OAAO,IAAI,QAAQ,CAAC,QAAQ,KAAK,QAAQ,EAAE,CAAC;QACrE,MAAM,IAAI,KAAK,CAAC,GAAG,kCAAkC,QAAQ,QAAQ,CAAC,QAAQ,EAAE,CAAC,CAAC;IACnF,CAAC;IAED,OAAO,QAAQ,CAAC;AAAA,CAChB;AAED,MAAM,UAAU,8BAA8B,CAAC,SAAuB,EAAmC;IACxG,MAAM,QAAQ,GAAG,4BAA4B,CAAC,SAAS,CAAC,CAAC;IACzD,IAAI,CAAC,QAAQ,EAAE,CAAC;QACf,OAAO,SAAS,CAAC;IAClB,CAAC;IAED,OAAO;QACN,SAAS,EAAE,IAAI,cAAc,CAAC,QAAQ,CAAC;QACvC,UAAU,EAAE,IAAI,eAAe,CAAC,QAAQ,CAA0B;KAClE,CAAC;AAAA,CACF","sourcesContent":["import type { Agent as HttpAgent } from \"node:http\";\nimport type { Agent as HttpsAgent } from \"node:https\";\nimport { HttpProxyAgent } from \"http-proxy-agent\";\nimport { HttpsProxyAgent } from \"https-proxy-agent\";\n\nconst DEFAULT_PROXY_PORTS: Record<string, number> = {\n\tftp: 21,\n\tgopher: 70,\n\thttp: 80,\n\thttps: 443,\n\tws: 80,\n\twss: 443,\n};\n\nexport interface NodeHttpProxyAgents {\n\thttpAgent: HttpAgent;\n\thttpsAgent: HttpsAgent;\n}\n\nexport const UNSUPPORTED_PROXY_PROTOCOL_MESSAGE =\n\t\"Unsupported proxy protocol. SOCKS and PAC proxy URLs are not supported; use an HTTP or HTTPS proxy URL.\";\n\nfunction getProxyEnv(key: string): string {\n\treturn process.env[key.toLowerCase()] || process.env[key.toUpperCase()] || \"\";\n}\n\nfunction parseProxyTargetUrl(targetUrl: string | URL): URL | undefined {\n\tif (targetUrl instanceof URL) {\n\t\treturn targetUrl;\n\t}\n\n\ttry {\n\t\treturn new URL(targetUrl);\n\t} catch {\n\t\treturn undefined;\n\t}\n}\n\nfunction shouldProxyHostname(hostname: string, port: number): boolean {\n\tconst noProxy = getProxyEnv(\"no_proxy\").toLowerCase();\n\tif (!noProxy) {\n\t\treturn true;\n\t}\n\tif (noProxy === \"*\") {\n\t\treturn false;\n\t}\n\n\treturn noProxy.split(/[,\\s]/).every((proxy) => {\n\t\tif (!proxy) {\n\t\t\treturn true;\n\t\t}\n\n\t\tconst parsedProxy = proxy.match(/^(.+):(\\d+)$/);\n\t\tlet proxyHostname = parsedProxy ? parsedProxy[1] : proxy;\n\t\tconst proxyPort = parsedProxy ? Number.parseInt(parsedProxy[2]!, 10) : 0;\n\t\tif (proxyPort && proxyPort !== port) {\n\t\t\treturn true;\n\t\t}\n\n\t\tif (!/^[.*]/.test(proxyHostname)) {\n\t\t\treturn hostname !== proxyHostname;\n\t\t}\n\n\t\tif (proxyHostname.startsWith(\"*\")) {\n\t\t\tproxyHostname = proxyHostname.slice(1);\n\t\t}\n\t\treturn !hostname.endsWith(proxyHostname);\n\t});\n}\n\nfunction getProxyForUrl(targetUrl: string | URL): string {\n\tconst parsedUrl = parseProxyTargetUrl(targetUrl);\n\tif (!parsedUrl?.protocol || !parsedUrl.host) {\n\t\treturn \"\";\n\t}\n\n\tconst protocol = parsedUrl.protocol.split(\":\", 1)[0]!;\n\tconst hostname = parsedUrl.host.replace(/:\\d*$/, \"\");\n\tconst port = Number.parseInt(parsedUrl.port, 10) || DEFAULT_PROXY_PORTS[protocol] || 0;\n\tif (!shouldProxyHostname(hostname, port)) {\n\t\treturn \"\";\n\t}\n\n\tlet proxy = getProxyEnv(`${protocol}_proxy`) || getProxyEnv(\"all_proxy\");\n\tif (proxy && !proxy.includes(\"://\")) {\n\t\tproxy = `${protocol}://${proxy}`;\n\t}\n\treturn proxy;\n}\n\nexport function resolveHttpProxyUrlForTarget(targetUrl: string | URL): URL | undefined {\n\tconst proxy = getProxyForUrl(targetUrl);\n\tif (!proxy) {\n\t\treturn undefined;\n\t}\n\n\tlet proxyUrl: URL;\n\ttry {\n\t\tproxyUrl = new URL(proxy);\n\t} catch (error) {\n\t\tthrow new Error(\n\t\t\t`Invalid proxy URL ${JSON.stringify(proxy)}: ${error instanceof Error ? error.message : String(error)}`,\n\t\t);\n\t}\n\n\tif (proxyUrl.protocol !== \"http:\" && proxyUrl.protocol !== \"https:\") {\n\t\tthrow new Error(`${UNSUPPORTED_PROXY_PROTOCOL_MESSAGE} Got ${proxyUrl.protocol}`);\n\t}\n\n\treturn proxyUrl;\n}\n\nexport function createHttpProxyAgentsForTarget(targetUrl: string | URL): NodeHttpProxyAgents | undefined {\n\tconst proxyUrl = resolveHttpProxyUrlForTarget(targetUrl);\n\tif (!proxyUrl) {\n\t\treturn undefined;\n\t}\n\n\treturn {\n\t\thttpAgent: new HttpProxyAgent(proxyUrl),\n\t\thttpsAgent: new HttpsProxyAgent(proxyUrl) as unknown as HttpsAgent,\n\t};\n}\n"]}

package/node_modules/@earendil-works/pi-ai/dist/utils/overflow.d.ts

@@ -0,0 +1,56 @@
+import type { AssistantMessage } from "../types.js";
+/**
+ * Check if an assistant message represents a context overflow error.
+ *
+ * This handles two cases:
+ * 1. Error-based overflow: Most providers return stopReason "error" with a
+ *    specific error message pattern.
+ * 2. Silent overflow: Some providers accept overflow requests and return
+ *    successfully. For these, we check if usage.input exceeds the context window.
+ *
+ * ## Reliability by Provider
+ *
+ * **Reliable detection (returns error with detectable message):**
+ * - Anthropic: "prompt is too long: X tokens > Y maximum" or "request_too_large"
+ * - OpenAI (Completions & Responses): "exceeds the context window" or "exceeds the model's maximum context length of X tokens"
+ * - Google Gemini: "input token count exceeds the maximum"
+ * - xAI (Grok): "maximum prompt length is X but request contains Y"
+ * - Groq: "reduce the length of the messages"
+ * - Cerebras: 400/413 status code (no body)
+ * - Mistral: "Prompt contains X tokens ... too large for model with Y maximum context length"
+ * - OpenRouter (all backends): "maximum context length is X tokens"
+ * - Together AI: "The input (X tokens) is longer than the model's context length (Y tokens)."
+ * - llama.cpp: "exceeds the available context size"
+ * - LM Studio: "greater than the context length"
+ * - Kimi For Coding: "exceeded model token limit: X (requested: Y)"
+ *
+ * **Unreliable detection:**
+ * - z.ai: Sometimes accepts overflow silently (detectable via usage.input > contextWindow),
+ *   sometimes returns rate limit errors. Pass contextWindow param to detect silent overflow.
+ * - Xiaomi MiMo: Truncates input to fit contextWindow then returns stopReason "length" with
+ *   output=0. Pass contextWindow param to detect via the "filled context + zero output" signal.
+ * - Ollama: May truncate input silently for some setups, but may also return explicit
+ *   overflow errors that match the patterns above. Silent truncation still cannot be
+ *   detected here because we do not know the expected token count.
+ *
+ * ## Custom Providers
+ *
+ * If you've added custom models via settings.json, this function may not detect
+ * overflow errors from those providers. To add support:
+ *
+ * 1. Send a request that exceeds the model's context window
+ * 2. Check the errorMessage in the response
+ * 3. Create a regex pattern that matches the error
+ * 4. The pattern should be added to OVERFLOW_PATTERNS in this file, or
+ *    check the errorMessage yourself before calling this function
+ *
+ * @param message - The assistant message to check
+ * @param contextWindow - Optional context window size for detecting silent overflow (z.ai)
+ * @returns true if the message indicates a context overflow
+ */
+export declare function isContextOverflow(message: AssistantMessage, contextWindow?: number): boolean;
+/**
+ * Get the overflow patterns for testing purposes.
+ */
+export declare function getOverflowPatterns(): RegExp[];
+//# sourceMappingURL=overflow.d.ts.map

package/node_modules/@earendil-works/pi-ai/dist/utils/overflow.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"overflow.d.ts","sourceRoot":"","sources":["../../src/utils/overflow.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,aAAa,CAAC;AAwEpD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAgDG;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.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 * - 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 * - 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 (all backends)\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 (all backends): \"maximum context length is X 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"]}

package/node_modules/@earendil-works/pi-ai/dist/utils/overflow.js

@@ -0,0 +1,151 @@
+/**
+ * Regex patterns to detect context overflow errors from different providers.
+ *
+ * These patterns match error messages returned when the input exceeds
+ * the model's context window.
+ *
+ * Provider-specific patterns (with example error messages):
+ *
+ * - Anthropic: "prompt is too long: 213462 tokens > 200000 maximum"
+ * - Anthropic: "413 {\"error\":{\"type\":\"request_too_large\",\"message\":\"Request exceeds the maximum size\"}}"
+ * - OpenAI: "Your input exceeds the context window of this model"
+ * - OpenAI/LiteLLM: "Requested token count exceeds the model's maximum context length of 131072 tokens"
+ * - Google: "The input token count (1196265) exceeds the maximum number of tokens allowed (1048575)"
+ * - xAI: "This model's maximum prompt length is 131072 but the request contains 537812 tokens"
+ * - Groq: "Please reduce the length of the messages or completion"
+ * - OpenRouter: "This endpoint's maximum context length is X tokens. However, you requested about Y tokens"
+ * - Together AI: "The input (X tokens) is longer than the model's context length (Y tokens)."
+ * - llama.cpp: "the request exceeds the available context size, try increasing it"
+ * - LM Studio: "tokens to keep from the initial prompt is greater than the context length"
+ * - 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)"
+ * - 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
+ * - Xiaomi MiMo: Truncates input to fill contextWindow exactly, then returns finish_reason "length"
+ *   with output=0 (no room left to generate). Detected via stopReason "length" + zero output +
+ *   input filling the context window.
+ * - Ollama: Some deployments truncate silently, others return errors like "prompt too long; exceeded max context length by X tokens"
+ */
+const OVERFLOW_PATTERNS = [
+    /prompt is too long/i, // Anthropic token overflow
+    /request_too_large/i, // Anthropic request byte-size overflow (HTTP 413)
+    /input is too long for requested model/i, // Amazon Bedrock
+    /exceeds the context window/i, // OpenAI (Completions & Responses API)
+    /exceeds (?:the )?(?:model'?s )?maximum context length of [\d,]+ tokens?/i, // OpenAI-compatible proxies (LiteLLM)
+    /input token count.*exceeds the maximum/i, // Google (Gemini)
+    /maximum prompt length is \d+/i, // xAI (Grok)
+    /reduce the length of the messages/i, // Groq
+    /maximum context length is \d+ tokens/i, // OpenRouter (all backends)
+    /input \(\d+ tokens\) is longer than the model'?s context length \(\d+ tokens\)/i, // Together AI
+    /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 window exceeds limit/i, // MiniMax
+    /exceeded model token limit/i, // Kimi For Coding
+    /too large for model with \d+ maximum context length/i, // Mistral
+    /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
+    /too many tokens/i, // Generic fallback
+    /token limit exceeded/i, // Generic fallback
+    /^4(?:00|13)\s*(?:status code)?\s*\(no body\)/i, // Cerebras: 400/413 with no body
+];
+/**
+ * Patterns that indicate non-overflow errors (e.g. rate limiting, server errors).
+ * Error messages matching any of these are excluded from overflow detection
+ * even if they also match an OVERFLOW_PATTERN.
+ *
+ * Example: Bedrock formats throttling errors as "ThrottlingException: Too many tokens,
+ * please wait before trying again." which would match the /too many tokens/i overflow
+ * pattern without this exclusion.
+ */
+const NON_OVERFLOW_PATTERNS = [
+    /^(Throttling error|Service unavailable):/i, // AWS Bedrock non-overflow errors (human-readable prefixes from formatBedrockError)
+    /rate limit/i, // Generic rate limiting
+    /too many requests/i, // Generic HTTP 429 style
+];
+/**
+ * Check if an assistant message represents a context overflow error.
+ *
+ * This handles two cases:
+ * 1. Error-based overflow: Most providers return stopReason "error" with a
+ *    specific error message pattern.
+ * 2. Silent overflow: Some providers accept overflow requests and return
+ *    successfully. For these, we check if usage.input exceeds the context window.
+ *
+ * ## Reliability by Provider
+ *
+ * **Reliable detection (returns error with detectable message):**
+ * - Anthropic: "prompt is too long: X tokens > Y maximum" or "request_too_large"
+ * - OpenAI (Completions & Responses): "exceeds the context window" or "exceeds the model's maximum context length of X tokens"
+ * - Google Gemini: "input token count exceeds the maximum"
+ * - xAI (Grok): "maximum prompt length is X but request contains Y"
+ * - Groq: "reduce the length of the messages"
+ * - Cerebras: 400/413 status code (no body)
+ * - Mistral: "Prompt contains X tokens ... too large for model with Y maximum context length"
+ * - OpenRouter (all backends): "maximum context length is X tokens"
+ * - Together AI: "The input (X tokens) is longer than the model's context length (Y tokens)."
+ * - llama.cpp: "exceeds the available context size"
+ * - LM Studio: "greater than the context length"
+ * - Kimi For Coding: "exceeded model token limit: X (requested: Y)"
+ *
+ * **Unreliable detection:**
+ * - z.ai: Sometimes accepts overflow silently (detectable via usage.input > contextWindow),
+ *   sometimes returns rate limit errors. Pass contextWindow param to detect silent overflow.
+ * - Xiaomi MiMo: Truncates input to fit contextWindow then returns stopReason "length" with
+ *   output=0. Pass contextWindow param to detect via the "filled context + zero output" signal.
+ * - Ollama: May truncate input silently for some setups, but may also return explicit
+ *   overflow errors that match the patterns above. Silent truncation still cannot be
+ *   detected here because we do not know the expected token count.
+ *
+ * ## Custom Providers
+ *
+ * If you've added custom models via settings.json, this function may not detect
+ * overflow errors from those providers. To add support:
+ *
+ * 1. Send a request that exceeds the model's context window
+ * 2. Check the errorMessage in the response
+ * 3. Create a regex pattern that matches the error
+ * 4. The pattern should be added to OVERFLOW_PATTERNS in this file, or
+ *    check the errorMessage yourself before calling this function
+ *
+ * @param message - The assistant message to check
+ * @param contextWindow - Optional context window size for detecting silent overflow (z.ai)
+ * @returns true if the message indicates a context overflow
+ */
+export function isContextOverflow(message, contextWindow) {
+    // Case 1: Check error message patterns
+    if (message.stopReason === "error" && message.errorMessage) {
+        // Skip messages matching known non-overflow patterns (e.g. throttling / rate-limit)
+        const isNonOverflow = NON_OVERFLOW_PATTERNS.some((p) => p.test(message.errorMessage));
+        if (!isNonOverflow && OVERFLOW_PATTERNS.some((p) => p.test(message.errorMessage))) {
+            return true;
+        }
+    }
+    // Case 2: Silent overflow (z.ai style) - successful but usage exceeds context
+    if (contextWindow && message.stopReason === "stop") {
+        const inputTokens = message.usage.input + message.usage.cacheRead;
+        if (inputTokens > contextWindow) {
+            return true;
+        }
+    }
+    // Case 3: Length-stop overflow (Xiaomi MiMo style) - server truncates oversized input
+    // to fit the context window, leaving no room for output. Returns stopReason "length"
+    // with output=0 and input+cacheRead filling the context window.
+    if (contextWindow && message.stopReason === "length" && message.usage.output === 0) {
+        const inputTokens = message.usage.input + message.usage.cacheRead;
+        if (inputTokens >= contextWindow * 0.99) {
+            return true;
+        }
+    }
+    return false;
+}
+/**
+ * Get the overflow patterns for testing purposes.
+ */
+export function getOverflowPatterns() {
+    return [...OVERFLOW_PATTERNS];
+}
+//# sourceMappingURL=overflow.js.map

package/node_modules/@earendil-works/pi-ai/dist/utils/overflow.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"overflow.js","sourceRoot":"","sources":["../../src/utils/overflow.ts"],"names":[],"mappings":"AAEA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA6BG;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,4BAA4B;IACrE,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;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAgDG;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.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 * - 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 * - 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 (all backends)\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 (all backends): \"maximum context length is X 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"]}

package/node_modules/@earendil-works/pi-ai/package.json

@@ -0,0 +1,107 @@
+{
+	"name": "@earendil-works/pi-ai",
+	"version": "0.74.1",
+	"description": "Unified LLM API with automatic model discovery and provider configuration",
+	"type": "module",
+	"main": "./dist/index.js",
+	"types": "./dist/index.d.ts",
+	"exports": {
+		".": {
+			"types": "./dist/index.d.ts",
+			"import": "./dist/index.js"
+		},
+		"./anthropic": {
+			"types": "./dist/providers/anthropic.d.ts",
+			"import": "./dist/providers/anthropic.js"
+		},
+		"./azure-openai-responses": {
+			"types": "./dist/providers/azure-openai-responses.d.ts",
+			"import": "./dist/providers/azure-openai-responses.js"
+		},
+		"./google": {
+			"types": "./dist/providers/google.d.ts",
+			"import": "./dist/providers/google.js"
+		},
+		"./google-vertex": {
+			"types": "./dist/providers/google-vertex.d.ts",
+			"import": "./dist/providers/google-vertex.js"
+		},
+		"./mistral": {
+			"types": "./dist/providers/mistral.d.ts",
+			"import": "./dist/providers/mistral.js"
+		},
+		"./openai-codex-responses": {
+			"types": "./dist/providers/openai-codex-responses.d.ts",
+			"import": "./dist/providers/openai-codex-responses.js"
+		},
+		"./openai-completions": {
+			"types": "./dist/providers/openai-completions.d.ts",
+			"import": "./dist/providers/openai-completions.js"
+		},
+		"./openai-responses": {
+			"types": "./dist/providers/openai-responses.d.ts",
+			"import": "./dist/providers/openai-responses.js"
+		},
+		"./oauth": {
+			"types": "./dist/oauth.d.ts",
+			"import": "./dist/oauth.js"
+		},
+		"./bedrock-provider": {
+			"types": "./dist/bedrock-provider.d.ts",
+			"import": "./dist/bedrock-provider.js"
+		}
+	},
+	"bin": {
+		"pi-ai": "./dist/cli.js"
+	},
+	"files": [
+		"dist",
+		"README.md"
+	],
+	"scripts": {
+		"clean": "shx rm -rf dist",
+		"generate-models": "npx tsx scripts/generate-models.ts",
+		"generate-image-models": "npx tsx scripts/generate-image-models.ts",
+		"build": "npm run generate-models && npm run generate-image-models && tsgo -p tsconfig.build.json",
+		"dev": "tsgo -p tsconfig.build.json --watch --preserveWatchOutput",
+		"dev:tsc": "tsgo -p tsconfig.build.json --watch --preserveWatchOutput",
+		"test": "vitest --run",
+		"prepublishOnly": "npm run clean && npm run build"
+	},
+	"dependencies": {
+		"@anthropic-ai/sdk": "^0.91.1",
+		"@aws-sdk/client-bedrock-runtime": "^3.1030.0",
+		"@google/genai": "^1.40.0",
+		"@mistralai/mistralai": "^2.2.0",
+		"http-proxy-agent": "^7.0.2",
+		"https-proxy-agent": "^7.0.6",
+		"openai": "6.26.0",
+		"partial-json": "^0.1.7",
+		"typebox": "^1.1.24"
+	},
+	"keywords": [
+		"ai",
+		"llm",
+		"openai",
+		"anthropic",
+		"gemini",
+		"bedrock",
+		"unified",
+		"api"
+	],
+	"author": "Mario Zechner",
+	"license": "MIT",
+	"repository": {
+		"type": "git",
+		"url": "git+https://github.com/earendil-works/pi-mono.git",
+		"directory": "packages/ai"
+	},
+	"engines": {
+		"node": ">=20.0.0"
+	},
+	"devDependencies": {
+		"@types/node": "^24.3.0",
+		"canvas": "^3.2.0",
+		"vitest": "^3.2.4"
+	}
+}