Initial human archive app

src/lib/litellm.ts

@@ -0,0 +1,68 @@
+import { env } from "@/lib/env";
+
+type ChatMessage = {
+  role: "system" | "user" | "assistant";
+  content: string;
+};
+
+type ChatCompletion = {
+  choices?: Array<{
+    message?: {
+      content?: string;
+    };
+  }>;
+  usage?: {
+    total_tokens?: number;
+  };
+};
+
+export type LlmResult = {
+  content: string;
+  model: string;
+  tokenCount?: number;
+  raw?: unknown;
+};
+
+export async function runLlmChat(messages: ChatMessage[], model = env.DEFAULT_LLM_MODEL): Promise<LlmResult> {
+  if (!env.LITELLM_BASE_URL || !env.LITELLM_API_KEY) {
+    return {
+      content: localFallback(messages),
+      model: `${model}:fallback`,
+    };
+  }
+
+  const response = await fetch(`${env.LITELLM_BASE_URL.replace(/\/$/, "")}/chat/completions`, {
+    method: "POST",
+    headers: {
+      "content-type": "application/json",
+      authorization: `Bearer ${env.LITELLM_API_KEY}`,
+    },
+    body: JSON.stringify({
+      model,
+      messages,
+      temperature: 0.2,
+    }),
+  });
+
+  if (!response.ok) {
+    const errorBody = await response.text();
+    throw new Error(`LiteLLM request failed: ${response.status} ${errorBody}`);
+  }
+
+  const data = (await response.json()) as ChatCompletion;
+  return {
+    content: data.choices?.[0]?.message?.content?.trim() ?? "",
+    model,
+    tokenCount: data.usage?.total_tokens,
+    raw: data,
+  };
+}
+
+function localFallback(messages: ChatMessage[]) {
+  const latestUserMessage = [...messages].reverse().find((message) => message.role === "user")?.content ?? "";
+  return [
+    "Lokaler Fallback: Es ist kein LiteLLM-Endpunkt konfiguriert.",
+    "Die Anfrage wurde strukturiert angenommen und muss nach Konfiguration eines Modells erneut ausgeführt werden.",
+    `Auszug der Anfrage: ${latestUserMessage.slice(0, 420)}`,
+  ].join("\n\n");
+}