litellmts-core 2.0.0 → 2.1.1

package/dist/handlers/ollama.js

@@ -8,10 +8,10 @@ function toStreamingChunk(ollamaResponse, model, prompt) {
     return {
         model: model,
         created: (0, getUnixTimestamp_1.getUnixTimestamp)(),
-        usage: (0, toUsage_1.toUsage)(prompt, ollamaResponse.response),
+        usage: (0, toUsage_1.toUsage)(prompt, ollamaResponse.message.content),
         choices: [
             {
-                delta: { content: ollamaResponse.response, role: 'assistant' },
+                delta: { content: ollamaResponse.message.content, role: 'assistant' },
                 finish_reason: 'stop',
                 index: 0,
             },
@@ -53,15 +53,19 @@ async function* iterateResponse(response, model, prompt) {
         }
     }
 }
-async function getOllamaResponse(model, prompt, baseUrl) {
-    return fetch(`${baseUrl}/api/generate`, {
+async function getOllamaResponse(model, prompt, baseUrl, stream, apiKey) {
+    const headers = {
+        'Content-Type': 'application/json',
+    };
+    if (apiKey)
+        headers.Authorization = `Bearer ${apiKey}`;
+    return fetch(`${baseUrl}/api/chat`, {
         method: 'POST',
-        headers: {
-            'Content-Type': 'application/json',
-        },
+        headers,
         body: JSON.stringify({
             model,
-            prompt,
+            messages: [{ role: 'user', content: prompt }],
+            stream,
         }),
     });
 }
@@ -71,7 +75,7 @@ async function OllamaHandler(params) {
         ? params.model.slice(7)
         : params.model;
     const prompt = (0, combinePrompts_1.combinePrompts)(params.messages);
-    const res = await getOllamaResponse(model, prompt, baseUrl);
+    const res = await getOllamaResponse(model, prompt, baseUrl, !!params.stream, params.apiKey);
     if (!res.ok) {
         throw new Error(`Received an error with code ${res.status} from Ollama API.`);
     }
@@ -88,12 +92,16 @@ async function OllamaHandler(params) {
     return toResponse(message, model, prompt);
 }
 const registry_1 = require("../models/registry");
-(0, registry_1.registerModelProvider)('ollama', async () => {
-    const res = await fetch('http://127.0.0.1:11434/api/tags');
+(0, registry_1.registerModelProvider)('ollama', async ({ baseUrl, apiKey } = {}) => {
+    const url = baseUrl ?? 'http://127.0.0.1:11434';
+    const headers = {};
+    if (apiKey)
+        headers.Authorization = `Bearer ${apiKey}`;
+    const res = await fetch(`${url.replace(/\/+$/, '')}/api/tags`, { headers });
     if (!res.ok)
         return [];
-    const { models } = await res.json();
-    return (models ?? []).map((m) => ({ id: m.name, provider: 'ollama' }));
+    const { models } = (await res.json());
+    return (models ?? []).map((m) => ({ id: m.name.replace(/-cloud$/, ''), provider: 'ollama' }));
 });
 const registry_2 = require("../registry");
 (0, registry_2.registerCompletionHandler)('ollama/', OllamaHandler);

package/dist/models/registry.js

@@ -10,14 +10,58 @@ const CACHE_TTL = 5 * 60 * 1000;
 function registerModelProvider(provider, fetcher) {
     fetchers.set(provider, fetcher);
 }
+async function tryOllamaTags(baseUrl, provider, apiKey) {
+    try {
+        const headers = {};
+        if (apiKey)
+            headers.Authorization = `Bearer ${apiKey}`;
+        const res = await fetch(`${baseUrl}/api/tags`, { headers });
+        if (!res.ok)
+            return null;
+        const { models } = (await res.json());
+        if (!Array.isArray(models))
+            return null;
+        return models.map((m) => ({ id: m.name.replace(/-cloud$/, ''), provider }));
+    }
+    catch {
+        return null;
+    }
+}
+async function tryOpenAIModels(baseUrl, provider, apiKey) {
+    try {
+        const headers = {};
+        if (apiKey)
+            headers.Authorization = `Bearer ${apiKey}`;
+        const res = await fetch(`${baseUrl}/models`, { headers });
+        if (!res.ok)
+            return null;
+        const { data } = (await res.json());
+        if (!Array.isArray(data))
+            return null;
+        return data.map((m) => ({ id: m.id, provider }));
+    }
+    catch {
+        return null;
+    }
+}
 async function listModels(provider, opts) {
     const cached = cache.get(provider);
     if (cached && cached.expires > Date.now())
         return cached.data;
-    const fetcher = fetchers.get(provider);
-    if (!fetcher)
-        throw new Error(`Provider '${provider}' not found.`);
-    const data = await fetcher(opts);
+    let data = null;
+    if (opts?.baseUrl) {
+        const baseUrl = opts.baseUrl.replace(/\/+$/, '');
+        data = await tryOllamaTags(baseUrl, provider, opts.apiKey);
+        if (!data || data.length === 0) {
+            data = await tryOpenAIModels(baseUrl, provider, opts.apiKey);
+        }
+    }
+    if (!data || data.length === 0) {
+        const fetcher = fetchers.get(provider);
+        if (!fetcher)
+            throw new Error(`Provider '${provider}' not found.`);
+        data = await fetcher(opts);
+    }
     cache.set(provider, { data, expires: Date.now() + CACHE_TTL });
     return data;
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "litellmts-core",
-  "version": "2.0.0",
+  "version": "2.1.1",
   "description": "TypeScript implementation of LiteLLM — unified interface for 45+ LLM providers",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",