FEAT: Added support for Bearer token in header for protected endpoints

package.json

@@ -76,6 +76,11 @@
             "description": "Ollama Server Endpoint. Empty for local instance. Example: http://192.168.0.100:11434",
             "order": 1
           },
+          "inference.bearerToken": {
+            "type": "string",
+            "default": "",
+            "description": "Auth Bearer token that should be used for secure requests. Leave empty if not desired."
+          },
           "inference.model": {
             "type": "string",
             "enum": [

src/config.ts

@@ -15,6 +15,7 @@ class Config {
         if (endpoint === '') {
             endpoint = 'http://127.0.0.1:11434';
         }
+        let bearerToken = config.get('bearerToken') as string;
 
         // Load general paremeters
         let maxLines = config.get('maxLines') as number;
@@ -39,6 +40,7 @@ class Config {
 
         return {
             endpoint,
+            bearerToken,
             maxLines,
             maxTokens,
             temperature,

src/modules/lineGenerator.ts

@@ -1,14 +1,16 @@
-export async function* lineGenerator(url: string, data: any): AsyncGenerator<string> {
-
+export async function* lineGenerator(url: string, data: any, bearerToken: string): AsyncGenerator<string> {
     // Request
     const controller = new AbortController();
     let res = await fetch(url, {
-        method: 'POST',
-        body: JSON.stringify(data),
-        headers: {
-            "Content-Type": "application/json",
-        },
-        signal: controller.signal
+      method: 'POST',
+      body: JSON.stringify(data),
+      headers: bearerToken ? {
+            'Content-Type': 'application/json',
+            Authorization: `Bearer ${bearerToken}`,
+          } : {
+            'Content-Type': 'application/json',
+          },
+      signal: controller.signal,
     });
     if (!res.ok || !res.body) {
         throw Error('Unable to connect to backend');

src/modules/ollamaCheckModel.ts

@@ -1,9 +1,12 @@
 import { info } from "./log";
 
-export async function ollamaCheckModel(endpoint: string, model: string) {
-
+export async function ollamaCheckModel(endpoint: string, model: string, bearerToken: string) {
     // Check if exists
-    let res = await fetch(endpoint + '/api/tags');
+    let res = await fetch(endpoint + '/api/tags', {
+      headers: bearerToken ? {
+            Authorization: `Bearer ${bearerToken}`,
+          } : {},
+    });
     if (!res.ok) {
         info(await res.text());
         info(endpoint + '/api/tags');

src/modules/ollamaDownloadModel.ts

@@ -1,9 +1,9 @@
 import { lineGenerator } from "./lineGenerator";
 import { info } from "./log";
 
-export async function ollamaDownloadModel(endpoint: string, model: string) {
+export async function ollamaDownloadModel(endpoint: string, model: string, bearerToken: string) {
     info('Downloading model from ollama: ' + model);
-    for await (let line of lineGenerator(endpoint + '/api/pull', { name: model })) {
+    for await (let line of lineGenerator(endpoint + '/api/pull', { name: model }, bearerToken)) {
         info('[DOWNLOAD] ' + line);
     }
 }

src/modules/ollamaTokenGenerator.ts

@@ -7,8 +7,8 @@ export type OllamaToken = {
     done: boolean
 };
 
-export async function* ollamaTokenGenerator(url: string, data: any): AsyncGenerator<OllamaToken> {
-    for await (let line of lineGenerator(url, data)) {
+export async function* ollamaTokenGenerator(url: string, data: any, bearerToken: string): AsyncGenerator<OllamaToken> {
+    for await (let line of lineGenerator(url, data, bearerToken)) {
         info('Receive line: ' + line);
         let parsed: OllamaToken;
         try {

src/prompts/autocomplete.ts

@@ -5,6 +5,7 @@ import { ModelFormat, adaptPrompt } from './processors/models';
 
 export async function autocomplete(args: {
     endpoint: string,
+    bearerToken: string,
     model: string,
     format: ModelFormat,
     prefix: string,
@@ -33,7 +34,7 @@ export async function autocomplete(args: {
     let res = '';
     let totalLines = 1;
     let blockStack: ('[' | '(' | '{')[] = [];
-    outer: for await (let tokens of ollamaTokenGenerator(args.endpoint + '/api/generate', data)) {
+    outer: for await (let tokens of ollamaTokenGenerator(args.endpoint + '/api/generate', data, args.bearerToken)) {
         if (args.canceled && args.canceled()) {
             break;
         }

src/prompts/provider.ts

@@ -122,7 +122,7 @@ export class PromptProvider implements vscode.InlineCompletionItemProvider {
                     try {
 
                         // Check model exists
-                        let modelExists = await ollamaCheckModel(inferenceConfig.endpoint, inferenceConfig.modelName);
+                        let modelExists = await ollamaCheckModel(inferenceConfig.endpoint, inferenceConfig.modelName, inferenceConfig.bearerToken);
                         if (token.isCancellationRequested) {
                             info(`Canceled after AI completion.`);
                             return;
@@ -147,7 +147,7 @@ export class PromptProvider implements vscode.InlineCompletionItemProvider {
 
                             // Perform download
                             this.update('sync~spin', 'Downloading');
-                            await ollamaDownloadModel(inferenceConfig.endpoint, inferenceConfig.modelName);
+                            await ollamaDownloadModel(inferenceConfig.endpoint, inferenceConfig.modelName, inferenceConfig.bearerToken);
                             this.update('sync~spin', 'Llama Coder')
                         }
                         if (token.isCancellationRequested) {
@@ -161,6 +161,7 @@ export class PromptProvider implements vscode.InlineCompletionItemProvider {
                             prefix: prepared.prefix,
                             suffix: prepared.suffix,
                             endpoint: inferenceConfig.endpoint,
+                            bearerToken: inferenceConfig.bearerToken,
                             model: inferenceConfig.modelName,
                             format: inferenceConfig.modelFormat,
                             maxLines: inferenceConfig.maxLines,