first commit as Fableous coder main changes commited

README.md

@@ -1,8 +1,7 @@
-# Ollama Autocoder
+# Fabelous Autocoder
 
 A simple to use Ollama autocompletion engine with options exposed and streaming functionality
 
-![example](example.gif)
 
 ## Requirements
 

package.json

@@ -1,23 +1,23 @@
 {
-	"name": "ollama-autocoder",
+	"name": "fabelous-autocoder",
-	"displayName": "Ollama Autocoder",
+	"displayName": "Fabelous Autocoder",
 	"description": "A simple to use Ollama autocompletion engine with options exposed and streaming functionality",
-	"version": "0.0.8",
+	"version": "0.0.1",
 	"icon": "icon.png",
-	"publisher": "10nates",
+	"publisher": "fabel",
-	"license": "MIT",
+	"license": "CC BY-ND 4.0",
 	"bugs": {
-		"url": "https://github.com/10Nates/ollama-autocoder/issues"
+		"url": "https://gitea.fabelous.app/fabel/Fabelous-Autocoder/issues"
 	},
 	"sponsor": {
 		"url": "https://ko-fi.com/natehedge"
 	},
 	"repository": {
 		"type": "git",
-		"url": "https://github.com/10Nates/ollama-autocoder"
+		"url": "https://gitea.fabelous.app/fabel/Fabelous-Autocoder"
 	},
 	"engines": {
-		"vscode": "^1.73.0"
+		"vscode": "^1.89.0"
 	},
 	"categories": [
 		"Machine Learning",
@@ -25,7 +25,6 @@
 		"Programming Languages"
 	],
 	"keywords": [
-		"llama",
 		"ollama",
 		"gpt",
 		"coding",
@@ -51,6 +50,11 @@
 					"default": "http://localhost:11434/api/generate",
 					"description": "The endpoint of the ollama REST API"
 				},
+				"ollama-autocoder.authentication": {
+					"type": "string",
+					"default": "",
+					"description": "Authorization Token for Ollama"
+				},
 				"ollama-autocoder.model": {
 					"type": "string",
 					"default": "openhermes2.5-mistral:7b-q4_K_M",
@@ -102,31 +106,77 @@
 					"type": "number",
 					"default": 0.5,
 					"description": "Temperature of the model. It is recommended to set it lower than you would for dialogue."
+				},
+				"ollama-autocoder.keep alive": {
+					"type": "number",
+					"default": 10,
+					"description": "Time in minutes before Ollama unloads the model."
+				},
+				"ollama-autocoder.top k": {
+					"type": "integer",
+					"description": "Top k sampling for the model."
+				},
+				"ollama-autocoder.top p": {
+					"type": "number",
+					"description": "Top p sampling for the model."
+				},
+				"ollama-autocoder.tfs z": {
+					"type": "number",
+					"description": "TFS z sampling for the model."
+				},
+				"ollama-autocoder.typical p": {
+					"type": "number",
+					"description": "Typical p sampling for the model."
+				},
+				"ollama-autocoder.repeat last n": {
+					"type": "number",
+					"description": "Repeat the last n tokens of the prompt. This can be useful for code completion as it allows the model to have more context."
+				},
+				"ollama-autocoder.repeat penalty": {
+					"type": "number",
+					"description": "Repetition penalty for the model."
+				},
+				"ollama-autocoder.presence penalty": {
+					"type": "number",
+					"description": "Presence penalty for the model."
+				},
+				"ollama-autocoder.frequency penalty": {
+					"type": "number",
+					"description": "Frequency penalty for the model."
+				},
+				"ollama-autocoder.num batch": {
+					"type": "number",
+					"description": "Batch size."
+				},
+				"ollama-autocoder.num keep": {
+					"type": "number",
+					"description": "Number of keep tokens for the model."
 				}
 			}
 		},
 		"commands": [
 			{
-			  "command": "ollama-autocoder.autocomplete",
+			  "command": "fabelous-autocoder.autocomplete",
-			  "title": "Autocomplete with Ollama"
+			  "title": "Autocompletion with Ollama"
 			}
 		  ]
 	},
 	"scripts": {
 		"vscode:prepublish": "npm run compile",
-		"compile": "tsc -p ./",
+		"compile": "tsc --skipLibCheck -p ./",
+		"package": "npm run compile && vsce package",
 		"lint": "eslint \"src/**/*.ts\"",
-		"watch": "tsc -watch -p ./"
+		"watch": "tsc --skipLibCheck -watch -p ./"
 	},
 	"devDependencies": {
-		"@types/node": "^16.18.34",
+		"@types/node": "^20.12.8",
-		"@types/vscode": "^1.73.0",
+		"@types/vscode": "^1.89.0",
-		"@typescript-eslint/eslint-plugin": "^6.7.0",
+		"@typescript-eslint/eslint-plugin": "^7.8.0",
-		"@typescript-eslint/parser": "^6.7.0",
+		"@typescript-eslint/parser": "^7.8.0",
-		"eslint": "^8.26.0",
+		"eslint": "^8.57.0",
-		"typescript": "^5.3.2"
+		"typescript": "^5.4.5"
 	},
 	"dependencies": {
-		"axios": "^1.6.2"
+		"axios": "^1.6.8"
 	}
 }

src/extension.ts

@@ -5,6 +5,7 @@ import axios from "axios";
 
 let VSConfig: vscode.WorkspaceConfiguration;
 let apiEndpoint: string;
+let apiAuthentication: string;
 let apiModel: string;
 let apiMessageHeader: string;
 let apiTemperature: number;
@@ -15,10 +16,22 @@ let responsePreview: boolean | undefined;
 let responsePreviewMaxTokens: number;
 let responsePreviewDelay: number;
 let continueInline: boolean | undefined;
+let keepAlive: number | undefined;
+let topK: number | undefined;
+let topP: number | undefined;
+let tfsZ: number | undefined;
+let typicalP: number | undefined;
+let repeatLastN: number | undefined;
+let repeatPenalty: number | undefined;
+let presencePenalty: number | undefined;
+let frequencyPenalty: number | undefined;
+let numBatch: number | undefined;
+let numKeep: number | undefined;
 
 function updateVSConfig() {
 	VSConfig = vscode.workspace.getConfiguration("ollama-autocoder");
 	apiEndpoint = VSConfig.get("endpoint") || "http://localhost:11434/api/generate";
+	apiAuthentication = VSConfig.get("authentication") || "";
 	apiModel = VSConfig.get("model") || "openhermes2.5-mistral:7b-q4_K_M"; // The model I tested with
 	apiMessageHeader = VSConfig.get("message header") || "";
 	numPredict = VSConfig.get("max tokens predicted") || 1000;
@@ -29,13 +42,22 @@ function updateVSConfig() {
 	responsePreviewDelay = VSConfig.get("preview delay") || 0; // Must be || 0 instead of || [default] because of truthy
 	continueInline = VSConfig.get("continue inline");
 	apiTemperature = VSConfig.get("temperature") || 0.5;
+	keepAlive = VSConfig.get("keep alive") || undefined;
+	topK = VSConfig.get("top k") || undefined;
+	topP = VSConfig.get("top p") || undefined;
+	tfsZ = VSConfig.get("tfs z") || undefined;
+	typicalP = VSConfig.get("typical p") || undefined;
+	repeatLastN = VSConfig.get("repeat last n") || undefined;
+	repeatPenalty = VSConfig.get("repeat penalty") || undefined;
+	presencePenalty = VSConfig.get("presence penalty") || undefined;
+	frequencyPenalty = VSConfig.get("frequency penalty") || undefined;
+	numBatch = VSConfig.get("num batch") || undefined;
+	numKeep = VSConfig.get("num keep") || undefined;
 }
 
 updateVSConfig();
-
 // No need for restart for any of these settings
 vscode.workspace.onDidChangeConfiguration(updateVSConfig);
-
 // Give model additional information
 function messageHeaderSub(document: vscode.TextDocument) {
 	const sub = apiMessageHeader
@@ -44,16 +66,13 @@ function messageHeaderSub(document: vscode.TextDocument) {
 		.replace("{PROJECT_NAME}", vscode.workspace.name || "Untitled");
 	return sub;
 }
-
 // internal function for autocomplete, not directly exposed
 async function autocompleteCommand(textEditor: vscode.TextEditor, cancellationToken?: vscode.CancellationToken) {
 	const document = textEditor.document;
 	const position = textEditor.selection.active;
-
 	// Get the current prompt
 	let prompt = document.getText(new vscode.Range(document.lineAt(0).range.start, position));
 	prompt = prompt.substring(Math.max(0, prompt.length - promptWindowSize), prompt.length);
-
 	// Show a progress message
 	vscode.window.withProgress(
 		{
@@ -64,7 +83,6 @@ async function autocompleteCommand(textEditor: vscode.TextEditor, cancellationTo
 		async (progress, progressCancellationToken) => {
 			try {
 				progress.report({ message: "Starting model..." });
-
 				let axiosCancelPost: () => void;
 				const axiosCancelToken = new axios.CancelToken((c) => {
 					const cancelPost = function () {
@@ -75,7 +93,6 @@ async function autocompleteCommand(textEditor: vscode.TextEditor, cancellationTo
 					progressCancellationToken.onCancellationRequested(cancelPost);
 					vscode.workspace.onDidCloseTextDocument(cancelPost);
 				});
-
 				// Make a request to the ollama.ai REST API
 				const response = await axios.post(apiEndpoint, {
 					model: apiModel, // Change this to the model you want to use
@@ -89,35 +106,33 @@ async function autocompleteCommand(textEditor: vscode.TextEditor, cancellationTo
 					}
 				}, {
 					cancelToken: axiosCancelToken,
-					responseType: 'stream'
+					responseType: 'stream',
+					headers: {
+						'Authorization': apiAuthentication
+					}
 				}
 				);
-
 				//tracker
 				let currentPosition = position;
-
 				response.data.on('data', async (d: Uint8Array) => {
 					progress.report({ message: "Generating..." });
-
 					// Check for user input (cancel)
 					if (currentPosition.line != textEditor.selection.end.line || currentPosition.character != textEditor.selection.end.character) {
 						axiosCancelPost(); // cancel axios => cancel finished promise => close notification
 						return;
 					}
-
 					// Get a completion from the response
 					const completion: string = JSON.parse(d.toString()).response;
 					// lastToken = completion;
 					
+
 					if (completion === "") {
 						return;
 					}
-
 					//complete edit for token
 					const edit = new vscode.WorkspaceEdit();
 					edit.insert(document.uri, currentPosition, completion);
 					await vscode.workspace.applyEdit(edit);
-
 					// Move the cursor to the end of the completion
 					const completionLines = completion.split("\n");
 					const newPosition = new vscode.Position(
@@ -129,14 +144,11 @@ async function autocompleteCommand(textEditor: vscode.TextEditor, cancellationTo
 						newPosition
 					);
 					currentPosition = newPosition;
-
 					// completion bar
 					progress.report({ message: "Generating...", increment: 1 / (numPredict / 100) });
-
 					// move cursor
 					textEditor.selection = newSelection;
 				});
-
 				// Keep cancel window available
 				const finished = new Promise((resolve) => {
 					response.data.on('end', () => {
@@ -147,9 +159,7 @@ async function autocompleteCommand(textEditor: vscode.TextEditor, cancellationTo
 						resolve(false);
 					});
 				});
-
 				await finished;
-
 			} catch (err: any) {
 				// Show an error message
 				vscode.window.showErrorMessage(
@@ -160,22 +170,17 @@ async function autocompleteCommand(textEditor: vscode.TextEditor, cancellationTo
 		}
 	);
 }
-
 // Completion item provider callback for activate
 async function provideCompletionItems(document: vscode.TextDocument, position: vscode.Position, cancellationToken: vscode.CancellationToken) {
-
 	// Create a completion item
 	const item = new vscode.CompletionItem("Autocomplete with Ollama");
-
 	// Set the insert text to a placeholder
 	item.insertText = new vscode.SnippetString('${1:}');
-
 	// Wait before initializing Ollama to reduce compute usage
 	if (responsePreview) await new Promise(resolve => setTimeout(resolve, responsePreviewDelay * 1000));
 	if (cancellationToken.isCancellationRequested) {
 		return [ item ];
 	}
-
 	// Set the label & inset text to a shortened, non-stream response
 	if (responsePreview) {
 		let prompt = document.getText(new vscode.Range(document.lineAt(0).range.start, position));
@@ -188,7 +193,18 @@ async function provideCompletionItems(document: vscode.TextDocument, position: v
 			options: {
 				num_predict: responsePreviewMaxTokens, // reduced compute max
 				temperature: apiTemperature,
-				stop: ['\n', '```']
+				stop: ['\n', '```'],
+				...keepAlive && { keep_alive: keepAlive },
+				...topK && { top_k: topK },
+				...topP && { top_p: topP },
+				...tfsZ && { tfs_z: tfsZ },
+				...typicalP && { typical_p: typicalP },
+				...repeatLastN && { repeat_last_n: repeatLastN },
+				...repeatPenalty && { repeat_penalty: repeatPenalty },
+				...presencePenalty && { presence_penalty: presencePenalty },
+				...frequencyPenalty && { frequency_penalty: frequencyPenalty },
+				...numBatch && { num_batch: numBatch },
+				...numKeep && { num_keep: numKeep },
 			}
 		}, {
 			cancelToken: new axios.CancelToken((c) => {
@@ -198,13 +214,11 @@ async function provideCompletionItems(document: vscode.TextDocument, position: v
 				cancellationToken.onCancellationRequested(cancelPost);
 			})
 		});
-
 		if (response_preview.data.response.trim() != "") { // default if empty
 			item.label = response_preview.data.response.trimStart(); // tended to add whitespace at the beginning
 			item.insertText = response_preview.data.response.trimStart();
 		}
 	}
-
 	// Set the documentation to a message
 	item.documentation = new vscode.MarkdownString('Press `Enter` to get an autocompletion from Ollama');
 	// Set the command to trigger the completion
@@ -216,7 +230,6 @@ async function provideCompletionItems(document: vscode.TextDocument, position: v
 	// Return the completion item
 	return [item];
 }
-
 // This method is called when extension is activated
 function activate(context: vscode.ExtensionContext) {
 	// Register a completion provider for JavaScript files
@@ -225,7 +238,6 @@ function activate(context: vscode.ExtensionContext) {
 	},
 		...completionKeys.split("")
 	);
-
 	// Register a command for getting a completion from Ollama through command/keybind
 	const externalAutocompleteCommand = vscode.commands.registerTextEditorCommand(
 		"ollama-autocoder.autocomplete",
@@ -234,14 +246,13 @@ function activate(context: vscode.ExtensionContext) {
 			autocompleteCommand(textEditor, cancellationToken);
 		}
 	);
-
 	// Add the commands & completion provider to the context
 	context.subscriptions.push(completionProvider);
 	context.subscriptions.push(externalAutocompleteCommand);
-
 }
 
 // This method is called when extension is deactivated
+// eslint-disable-next-line @typescript-eslint/no-empty-function
 function deactivate() { }
 
 module.exports = {