llm-strings 1.1.2 → 1.2.0

package/dist/validate.cjs

@@ -24,34 +24,81 @@ __export(validate_exports, {
 });
 module.exports = __toCommonJS(validate_exports);
 
-// src/parse.ts
-function parse(connectionString) {
-  const url = new URL(connectionString);
-  if (url.protocol !== "llm:") {
-    throw new Error(
-      `Invalid scheme: expected "llm://", got "${url.protocol}//"`
-    );
+// src/provider-core.ts
+function hasOwn(object, key) {
+  return Object.prototype.hasOwnProperty.call(object, key);
+}
+var HOST_ALIASES = {
+  openai: "api.openai.com",
+  anthropic: "api.anthropic.com",
+  google: "generativelanguage.googleapis.com",
+  aistudio: "generativelanguage.googleapis.com",
+  mistral: "api.mistral.ai",
+  cohere: "api.cohere.com",
+  bedrock: "bedrock-runtime.us-east-1.amazonaws.com",
+  openrouter: "openrouter.ai",
+  vercel: "gateway.ai.vercel.app",
+  alibaba: "dashscope-intl.aliyuncs.com",
+  alibabacloud: "dashscope-intl.aliyuncs.com",
+  dashscope: "dashscope-intl.aliyuncs.com",
+  fireworks: "api.fireworks.ai",
+  fireworksai: "api.fireworks.ai",
+  venice: "api.venice.ai",
+  parasail: "api.parasail.io",
+  deepinfra: "api.deepinfra.com",
+  atlascloud: "api.atlascloud.ai",
+  novita: "api.novita.ai",
+  novitaai: "api.novita.ai",
+  grok: "api.x.ai",
+  xai: "api.x.ai",
+  wandb: "api.inference.wandb.ai",
+  weightsandbiases: "api.inference.wandb.ai",
+  baidu: "qianfan.baidubce.com",
+  qianfan: "qianfan.baidubce.com",
+  vertex: "aiplatform.googleapis.com",
+  xiaomi: "api.xiaomimimo.com",
+  minimax: "api.minimax.io"
+};
+function readProcessEnv() {
+  return typeof process !== "undefined" && process.env ? process.env : {};
+}
+function normalizeHostValue(value) {
+  const trimmed = value.trim();
+  if (!trimmed) return trimmed;
+  try {
+    if (trimmed.includes("://")) {
+      return new URL(trimmed).host;
+    }
+  } catch {
   }
-  const host = url.hostname;
-  const model = url.pathname.replace(/^\//, "");
-  const label = url.username || void 0;
-  const apiKey = url.password || void 0;
-  const params = {};
-  for (const [key, value] of url.searchParams) {
-    params[key] = value;
+  return trimmed.replace(/^\/\//, "").split("/")[0] ?? trimmed;
+}
+function envHostOverride(alias, env) {
+  const upper = alias.toUpperCase();
+  const override = env[`LLM_STRINGS_${upper}_HOST`] ?? env[`LLM_STRINGS_HOST_${upper}`];
+  return override?.trim() ? override : void 0;
+}
+function resolveHostAlias(host, env = readProcessEnv()) {
+  const normalizedHost = host.toLowerCase();
+  if (!hasOwn(HOST_ALIASES, normalizedHost)) {
+    return { host };
   }
+  const alias = normalizedHost;
+  const override = envHostOverride(alias, env);
   return {
-    raw: connectionString,
-    host,
-    model,
-    label,
-    apiKey,
-    params
+    host: normalizeHostValue(override ?? HOST_ALIASES[alias]),
+    alias
   };
 }
-
-// src/provider-core.ts
+function providerFromHostAlias(alias) {
+  const normalizedAlias = alias.toLowerCase();
+  if (hasOwn(PROVIDER_PARAMS, normalizedAlias)) {
+    return normalizedAlias;
+  }
+  return void 0;
+}
 function detectProvider(host) {
+  host = host.toLowerCase();
   if (host.includes("openrouter")) return "openrouter";
   if (host.includes("gateway.ai.vercel")) return "vercel";
   if (host.includes("amazonaws") || host.includes("bedrock")) return "bedrock";
@@ -214,11 +261,40 @@ var PROVIDER_PARAMS = {
 };
 var PARAM_SPECS = {
   openai: {
-    temperature: { type: "number", min: 0, max: 2, default: 0.7, description: "Controls randomness" },
-    max_tokens: { type: "number", min: 1, default: 4096, description: "Maximum output tokens" },
-    top_p: { type: "number", min: 0, max: 1, default: 1, description: "Nucleus sampling" },
-    frequency_penalty: { type: "number", min: -2, max: 2, default: 0, description: "Penalize frequent tokens" },
-    presence_penalty: { type: "number", min: -2, max: 2, default: 0, description: "Penalize repeated topics" },
+    temperature: {
+      type: "number",
+      min: 0,
+      max: 2,
+      default: 0.7,
+      description: "Controls randomness"
+    },
+    max_tokens: {
+      type: "number",
+      min: 1,
+      default: 4096,
+      description: "Maximum output tokens"
+    },
+    top_p: {
+      type: "number",
+      min: 0,
+      max: 1,
+      default: 1,
+      description: "Nucleus sampling"
+    },
+    frequency_penalty: {
+      type: "number",
+      min: -2,
+      max: 2,
+      default: 0,
+      description: "Penalize frequent tokens"
+    },
+    presence_penalty: {
+      type: "number",
+      min: -2,
+      max: 2,
+      default: 0,
+      description: "Penalize repeated topics"
+    },
     stop: { type: "string", description: "Stop sequences" },
     n: { type: "number", min: 1, default: 1, description: "Completions count" },
     seed: { type: "number", description: "Random seed" },
@@ -231,73 +307,288 @@ var PARAM_SPECS = {
     }
   },
   anthropic: {
-    temperature: { type: "number", min: 0, max: 1, default: 0.7, description: "Controls randomness" },
-    max_tokens: { type: "number", min: 1, default: 4096, description: "Maximum output tokens" },
-    top_p: { type: "number", min: 0, max: 1, default: 1, description: "Nucleus sampling" },
-    top_k: { type: "number", min: 0, default: 40, description: "Top-K sampling" },
+    temperature: {
+      type: "number",
+      min: 0,
+      max: 1,
+      default: 0.7,
+      description: "Controls randomness"
+    },
+    max_tokens: {
+      type: "number",
+      min: 1,
+      default: 4096,
+      description: "Maximum output tokens"
+    },
+    top_p: {
+      type: "number",
+      min: 0,
+      max: 1,
+      default: 1,
+      description: "Nucleus sampling"
+    },
+    top_k: {
+      type: "number",
+      min: 0,
+      default: 40,
+      description: "Top-K sampling"
+    },
     stop_sequences: { type: "string", description: "Stop sequences" },
     stream: { type: "boolean", default: false, description: "Stream response" },
-    effort: { type: "string", values: ["low", "medium", "high", "max"], default: "medium", description: "Thinking effort" },
-    cache_control: { type: "string", values: ["ephemeral"], default: "ephemeral", description: "Cache control" },
-    cache_ttl: { type: "string", values: ["5m", "1h"], default: "5m", description: "Cache TTL" }
+    effort: {
+      type: "string",
+      values: ["low", "medium", "high", "max"],
+      default: "medium",
+      description: "Thinking effort"
+    },
+    cache_control: {
+      type: "string",
+      values: ["ephemeral"],
+      default: "ephemeral",
+      description: "Cache control"
+    },
+    cache_ttl: {
+      type: "string",
+      values: ["5m", "1h"],
+      default: "5m",
+      description: "Cache TTL"
+    }
   },
   google: {
-    temperature: { type: "number", min: 0, max: 2, default: 0.7, description: "Controls randomness" },
-    maxOutputTokens: { type: "number", min: 1, default: 4096, description: "Maximum output tokens" },
-    topP: { type: "number", min: 0, max: 1, default: 1, description: "Nucleus sampling" },
-    topK: { type: "number", min: 0, default: 40, description: "Top-K sampling" },
-    frequencyPenalty: { type: "number", min: -2, max: 2, default: 0, description: "Penalize frequent tokens" },
-    presencePenalty: { type: "number", min: -2, max: 2, default: 0, description: "Penalize repeated topics" },
+    temperature: {
+      type: "number",
+      min: 0,
+      max: 2,
+      default: 0.7,
+      description: "Controls randomness"
+    },
+    maxOutputTokens: {
+      type: "number",
+      min: 1,
+      default: 4096,
+      description: "Maximum output tokens"
+    },
+    topP: {
+      type: "number",
+      min: 0,
+      max: 1,
+      default: 1,
+      description: "Nucleus sampling"
+    },
+    topK: {
+      type: "number",
+      min: 0,
+      default: 40,
+      description: "Top-K sampling"
+    },
+    frequencyPenalty: {
+      type: "number",
+      min: -2,
+      max: 2,
+      default: 0,
+      description: "Penalize frequent tokens"
+    },
+    presencePenalty: {
+      type: "number",
+      min: -2,
+      max: 2,
+      default: 0,
+      description: "Penalize repeated topics"
+    },
     stopSequences: { type: "string", description: "Stop sequences" },
-    candidateCount: { type: "number", min: 1, default: 1, description: "Candidate count" },
+    candidateCount: {
+      type: "number",
+      min: 1,
+      default: 1,
+      description: "Candidate count"
+    },
     stream: { type: "boolean", default: false, description: "Stream response" },
     seed: { type: "number", description: "Random seed" },
     responseMimeType: { type: "string", description: "Response MIME type" },
     responseSchema: { type: "string", description: "Response schema" }
   },
   mistral: {
-    temperature: { type: "number", min: 0, max: 1, default: 0.7, description: "Controls randomness" },
-    max_tokens: { type: "number", min: 1, default: 4096, description: "Maximum output tokens" },
-    top_p: { type: "number", min: 0, max: 1, default: 1, description: "Nucleus sampling" },
-    frequency_penalty: { type: "number", min: -2, max: 2, default: 0, description: "Penalize frequent tokens" },
-    presence_penalty: { type: "number", min: -2, max: 2, default: 0, description: "Penalize repeated topics" },
+    temperature: {
+      type: "number",
+      min: 0,
+      max: 1,
+      default: 0.7,
+      description: "Controls randomness"
+    },
+    max_tokens: {
+      type: "number",
+      min: 1,
+      default: 4096,
+      description: "Maximum output tokens"
+    },
+    top_p: {
+      type: "number",
+      min: 0,
+      max: 1,
+      default: 1,
+      description: "Nucleus sampling"
+    },
+    frequency_penalty: {
+      type: "number",
+      min: -2,
+      max: 2,
+      default: 0,
+      description: "Penalize frequent tokens"
+    },
+    presence_penalty: {
+      type: "number",
+      min: -2,
+      max: 2,
+      default: 0,
+      description: "Penalize repeated topics"
+    },
     stop: { type: "string", description: "Stop sequences" },
     n: { type: "number", min: 1, default: 1, description: "Completions count" },
     random_seed: { type: "number", description: "Random seed" },
     stream: { type: "boolean", default: false, description: "Stream response" },
-    safe_prompt: { type: "boolean", default: false, description: "Enable safe prompt" },
-    min_tokens: { type: "number", min: 0, default: 0, description: "Minimum tokens" }
+    safe_prompt: {
+      type: "boolean",
+      default: false,
+      description: "Enable safe prompt"
+    },
+    min_tokens: {
+      type: "number",
+      min: 0,
+      default: 0,
+      description: "Minimum tokens"
+    }
   },
   cohere: {
-    temperature: { type: "number", min: 0, max: 1, default: 0.7, description: "Controls randomness" },
-    max_tokens: { type: "number", min: 1, default: 4096, description: "Maximum output tokens" },
-    p: { type: "number", min: 0, max: 1, default: 1, description: "Nucleus sampling (p)" },
-    k: { type: "number", min: 0, max: 500, default: 40, description: "Top-K sampling (k)" },
-    frequency_penalty: { type: "number", min: 0, max: 1, default: 0, description: "Penalize frequent tokens" },
-    presence_penalty: { type: "number", min: 0, max: 1, default: 0, description: "Penalize repeated topics" },
+    temperature: {
+      type: "number",
+      min: 0,
+      max: 1,
+      default: 0.7,
+      description: "Controls randomness"
+    },
+    max_tokens: {
+      type: "number",
+      min: 1,
+      default: 4096,
+      description: "Maximum output tokens"
+    },
+    p: {
+      type: "number",
+      min: 0,
+      max: 1,
+      default: 1,
+      description: "Nucleus sampling (p)"
+    },
+    k: {
+      type: "number",
+      min: 0,
+      max: 500,
+      default: 40,
+      description: "Top-K sampling (k)"
+    },
+    frequency_penalty: {
+      type: "number",
+      min: 0,
+      max: 1,
+      default: 0,
+      description: "Penalize frequent tokens"
+    },
+    presence_penalty: {
+      type: "number",
+      min: 0,
+      max: 1,
+      default: 0,
+      description: "Penalize repeated topics"
+    },
     stop_sequences: { type: "string", description: "Stop sequences" },
     stream: { type: "boolean", default: false, description: "Stream response" },
     seed: { type: "number", description: "Random seed" }
   },
   bedrock: {
     // Converse API inferenceConfig params
-    temperature: { type: "number", min: 0, max: 1, default: 0.7, description: "Controls randomness" },
-    maxTokens: { type: "number", min: 1, default: 4096, description: "Maximum output tokens" },
-    topP: { type: "number", min: 0, max: 1, default: 1, description: "Nucleus sampling" },
-    topK: { type: "number", min: 0, default: 40, description: "Top-K sampling" },
+    temperature: {
+      type: "number",
+      min: 0,
+      max: 1,
+      default: 0.7,
+      description: "Controls randomness"
+    },
+    maxTokens: {
+      type: "number",
+      min: 1,
+      default: 4096,
+      description: "Maximum output tokens"
+    },
+    topP: {
+      type: "number",
+      min: 0,
+      max: 1,
+      default: 1,
+      description: "Nucleus sampling"
+    },
+    topK: {
+      type: "number",
+      min: 0,
+      default: 40,
+      description: "Top-K sampling"
+    },
     stopSequences: { type: "string", description: "Stop sequences" },
     stream: { type: "boolean", default: false, description: "Stream response" },
-    cache_control: { type: "string", values: ["ephemeral"], default: "ephemeral", description: "Cache control" },
-    cache_ttl: { type: "string", values: ["5m", "1h"], default: "5m", description: "Cache TTL" }
+    cache_control: {
+      type: "string",
+      values: ["ephemeral"],
+      default: "ephemeral",
+      description: "Cache control"
+    },
+    cache_ttl: {
+      type: "string",
+      values: ["5m", "1h"],
+      default: "5m",
+      description: "Cache TTL"
+    }
   },
   openrouter: {
     // Loose validation — proxies to many providers with varying ranges
-    temperature: { type: "number", min: 0, max: 2, default: 0.7, description: "Controls randomness" },
-    max_tokens: { type: "number", min: 1, default: 4096, description: "Maximum output tokens" },
-    top_p: { type: "number", min: 0, max: 1, default: 1, description: "Nucleus sampling" },
-    top_k: { type: "number", min: 0, default: 40, description: "Top-K sampling" },
-    frequency_penalty: { type: "number", min: -2, max: 2, default: 0, description: "Penalize frequent tokens" },
-    presence_penalty: { type: "number", min: -2, max: 2, default: 0, description: "Penalize repeated topics" },
+    temperature: {
+      type: "number",
+      min: 0,
+      max: 2,
+      default: 0.7,
+      description: "Controls randomness"
+    },
+    max_tokens: {
+      type: "number",
+      min: 1,
+      default: 4096,
+      description: "Maximum output tokens"
+    },
+    top_p: {
+      type: "number",
+      min: 0,
+      max: 1,
+      default: 1,
+      description: "Nucleus sampling"
+    },
+    top_k: {
+      type: "number",
+      min: 0,
+      default: 40,
+      description: "Top-K sampling"
+    },
+    frequency_penalty: {
+      type: "number",
+      min: -2,
+      max: 2,
+      default: 0,
+      description: "Penalize frequent tokens"
+    },
+    presence_penalty: {
+      type: "number",
+      min: -2,
+      max: 2,
+      default: 0,
+      description: "Penalize repeated topics"
+    },
     stop: { type: "string", description: "Stop sequences" },
     n: { type: "number", min: 1, default: 1, description: "Completions count" },
     seed: { type: "number", description: "Random seed" },
@@ -311,12 +602,46 @@ var PARAM_SPECS = {
   },
   vercel: {
     // Loose validation — proxies to many providers with varying ranges
-    temperature: { type: "number", min: 0, max: 2, default: 0.7, description: "Controls randomness" },
-    max_tokens: { type: "number", min: 1, default: 4096, description: "Maximum output tokens" },
-    top_p: { type: "number", min: 0, max: 1, default: 1, description: "Nucleus sampling" },
-    top_k: { type: "number", min: 0, default: 40, description: "Top-K sampling" },
-    frequency_penalty: { type: "number", min: -2, max: 2, default: 0, description: "Penalize frequent tokens" },
-    presence_penalty: { type: "number", min: -2, max: 2, default: 0, description: "Penalize repeated topics" },
+    temperature: {
+      type: "number",
+      min: 0,
+      max: 2,
+      default: 0.7,
+      description: "Controls randomness"
+    },
+    max_tokens: {
+      type: "number",
+      min: 1,
+      default: 4096,
+      description: "Maximum output tokens"
+    },
+    top_p: {
+      type: "number",
+      min: 0,
+      max: 1,
+      default: 1,
+      description: "Nucleus sampling"
+    },
+    top_k: {
+      type: "number",
+      min: 0,
+      default: 40,
+      description: "Top-K sampling"
+    },
+    frequency_penalty: {
+      type: "number",
+      min: -2,
+      max: 2,
+      default: 0,
+      description: "Penalize frequent tokens"
+    },
+    presence_penalty: {
+      type: "number",
+      min: -2,
+      max: 2,
+      default: 0,
+      description: "Penalize repeated topics"
+    },
     stop: { type: "string", description: "Stop sequences" },
     n: { type: "number", min: 1, default: 1, description: "Completions count" },
     seed: { type: "number", description: "Random seed" },
@@ -343,7 +668,13 @@ function detectGatewaySubProvider(model) {
   const slash = model.indexOf("/");
   if (slash < 1) return void 0;
   const prefix = model.slice(0, slash);
-  const direct = ["openai", "anthropic", "google", "mistral", "cohere"];
+  const direct = [
+    "openai",
+    "anthropic",
+    "google",
+    "mistral",
+    "cohere"
+  ];
   return direct.find((p) => p === prefix);
 }
 var REASONING_MODEL_UNSUPPORTED = /* @__PURE__ */ new Set([
@@ -403,9 +734,36 @@ var CACHE_TTLS = {
 };
 var DURATION_RE = /^\d+[mh]$/;
 
+// src/parse.ts
+function parse(connectionString) {
+  const url = new URL(connectionString);
+  if (url.protocol !== "llm:") {
+    throw new Error(
+      `Invalid scheme: expected "llm://", got "${url.protocol}//"`
+    );
+  }
+  const { host, alias: hostAlias } = resolveHostAlias(url.host);
+  const model = url.pathname.replace(/^\//, "");
+  const label = url.username || void 0;
+  const apiKey = url.password || void 0;
+  const params = {};
+  for (const [key, value] of url.searchParams) {
+    params[key] = value;
+  }
+  return {
+    raw: connectionString,
+    host,
+    hostAlias,
+    model,
+    label,
+    apiKey,
+    params
+  };
+}
+
 // src/normalize.ts
 function normalize(config, options = {}) {
-  const provider = detectProvider(config.host);
+  const provider = (config.hostAlias ? providerFromHostAlias(config.hostAlias) : void 0) ?? detectProvider(config.host);
   const subProvider = provider && isGatewayProvider(provider) ? detectGatewaySubProvider(config.model) : void 0;
   const changes = [];
   const params = {};
@@ -590,11 +948,7 @@ function validate(connectionString, options = {}) {
     }
     let spec = specs[key];
     if (subProvider && gatewayReverseMap && !spec) {
-      const result = lookupSubProviderSpec(
-        key,
-        gatewayReverseMap,
-        subProvider
-      );
+      const result = lookupSubProviderSpec(key, gatewayReverseMap, subProvider);
       spec = result.spec;
     }
     if (!spec) continue;