settings only update on config change, better progress reporting, changed name, added README and icon, added categories & keywords, changed activation event
This commit is contained in:
Nathan Hedge
parent
fd0f553738
commit
4f26bf3af3
18
README.md
18
README.md
|
|
@ -1,3 +1,17 @@
|
|||
# Ollama Coder
|
||||
# Ollama Autocoder
|
||||
|
||||
An ollama-based autocompletion engine.
|
||||
A simple to use Ollama autocompletion engine with options exposed.
|
||||
|
||||
## Requirements
|
||||
|
||||
- Ollama must be serving on the API endpoint applied in settings
|
||||
- For installation of Ollama, visit [ollama.ai](https://ollama.ai)
|
||||
- Ollama must have the model applied in settings installed.
|
||||
- For fastest results, an Nvidia GPU or Apple Silicon is recommended. CPU still works on small models.
|
||||
|
||||
## How to Use
|
||||
|
||||
1. In a text document, press space or go to a new line. The option `Autocomplete with Ollama` will appear. Press enter to start generation.
|
||||
2. After startup, the tokens will be streamed to your cursor.
|
||||
3. To stop the generation early, press the "Cancel" button on the "Ollama Autocoder" notification
|
||||
4. Once generation stops, the notification will disappear.
|
||||
|
|
|
|||
44
package.json
44
package.json
|
|
@ -1,48 +1,68 @@
|
|||
{
|
||||
"name": "ollama-coder",
|
||||
"displayName": "Ollama Coder",
|
||||
"name": "ollama-autocoder",
|
||||
"displayName": "Ollama Autocoder",
|
||||
"description": "An Ollama autocompletion engine with options exposed",
|
||||
"version": "0.0.1",
|
||||
"icon": "icon.png",
|
||||
"publisher": "10nates",
|
||||
"license": "MIT",
|
||||
"repository": {
|
||||
"type": "git",
|
||||
"url": "https://github.com/10Nates/ollama-coder"
|
||||
"url": "https://github.com/10Nates/ollama-autocoder"
|
||||
},
|
||||
"engines": {
|
||||
"vscode": "^1.73.0"
|
||||
},
|
||||
"categories": [
|
||||
"Other"
|
||||
"Machine Learning",
|
||||
"Snippets",
|
||||
"Programming Languages"
|
||||
],
|
||||
"keywords": [
|
||||
"llama",
|
||||
"ollama",
|
||||
"gpt",
|
||||
"coding",
|
||||
"autocomplete",
|
||||
"open source",
|
||||
"assistant",
|
||||
"ai",
|
||||
"llm"
|
||||
],
|
||||
"activationEvents": [
|
||||
"*"
|
||||
"onStartupFinished"
|
||||
],
|
||||
"main": "./out/extension.js",
|
||||
"contributes": {
|
||||
"configuration": {
|
||||
"title": "Ollama Coder",
|
||||
"title": "Ollama Autocoder",
|
||||
"properties": {
|
||||
"ollama-coder.endpoint": {
|
||||
"ollama-autocoder.endpoint": {
|
||||
"type": "string",
|
||||
"default": "http://localhost:11434/api/generate",
|
||||
"description": "The endpoint of the ollama REST API"
|
||||
},
|
||||
"ollama-coder.model": {
|
||||
"ollama-autocoder.model": {
|
||||
"type": "string",
|
||||
"default": "deepseek-coder",
|
||||
"default": "openhermes2.5-mistral:7b-q4_K_M",
|
||||
"description": "The model to use for generating completions"
|
||||
},
|
||||
"ollama-coder.system-message": {
|
||||
"ollama-autocoder.raw-input": {
|
||||
"type": "boolean",
|
||||
"default": false,
|
||||
"description": "Prompt the model without formatting. Disables system message."
|
||||
},
|
||||
"ollama-autocoder.system-message": {
|
||||
"type": "string",
|
||||
"default": "You are a code autocompletion engine. Respond with a continuation of the code provided and nothing else. Code should not be in a code block. Anything that is not code should be written as a code comment.",
|
||||
"description": "The system message to use for code completions. Type DEFAULT for Makefile."
|
||||
},
|
||||
"ollama-coder.max-tokens-predicted": {
|
||||
"ollama-autocoder.max-tokens-predicted": {
|
||||
"type": "integer",
|
||||
"default": 500,
|
||||
"description": "The maximum number of tokens generated by the model."
|
||||
},
|
||||
"ollama-coder.prompt-window-size": {
|
||||
"ollama-autocoder.prompt-window-size": {
|
||||
"type": "integer",
|
||||
"default": 2000,
|
||||
"description": "The size of the prompt in characters. NOT tokens, so can be set about 1.5-2x the max tokens of the model (varies)."
|
||||
|
|
|
|||
|
|
@ -1,36 +1,56 @@
|
|||
// Significant help from GPT4
|
||||
|
||||
import * as vscode from "vscode";
|
||||
import axios, { AxiosResponse } from "axios";
|
||||
import axios from "axios";
|
||||
|
||||
const VSConfig = vscode.workspace.getConfiguration("ollama-coder");
|
||||
const apiEndpoint: string = VSConfig.get("apiEndpoint") || "http://localhost:11434/api/generate";
|
||||
const apiModel: string = VSConfig.get("model") || "deepseek-coder";
|
||||
let apiSystemMessage: string | undefined = VSConfig.get("system-message");
|
||||
if (apiSystemMessage == "DEFAULT") apiSystemMessage = undefined;
|
||||
const numPredict: number = VSConfig.get("max-tokens-predicted") || 500;
|
||||
const promptWindowSize: number = VSConfig.get("prompt-window-size") || 2000;
|
||||
let VSConfig: vscode.WorkspaceConfiguration;
|
||||
let apiEndpoint: string;
|
||||
let apiModel: string;
|
||||
let apiSystemMessage: string | undefined;
|
||||
let numPredict: number;
|
||||
let promptWindowSize: number;
|
||||
let rawInput: boolean;
|
||||
|
||||
// Function called on ollama-coder.autocomplete
|
||||
function updateVSConfig() {
|
||||
VSConfig = vscode.workspace.getConfiguration("ollama-autocoder");
|
||||
apiEndpoint = VSConfig.get("apiEndpoint") || "http://localhost:11434/api/generate";
|
||||
apiModel = VSConfig.get("model") || "openhermes2.5-mistral:7b-q4_K_M";
|
||||
apiSystemMessage = VSConfig.get("system-message");
|
||||
numPredict = VSConfig.get("max-tokens-predicted") || 500;
|
||||
promptWindowSize = VSConfig.get("prompt-window-size") || 2000;
|
||||
rawInput = VSConfig.get("raw-input") || false;
|
||||
|
||||
if (apiSystemMessage == "DEFAULT" || rawInput) apiSystemMessage = undefined;
|
||||
}
|
||||
|
||||
updateVSConfig();
|
||||
|
||||
// No need for restart for any of these settings
|
||||
vscode.workspace.onDidChangeConfiguration(updateVSConfig);
|
||||
|
||||
// Function called on ollama-autocoder.autocomplete
|
||||
async function autocompleteCommand(document: vscode.TextDocument, position: vscode.Position, prompt: string, cancellationToken: vscode.CancellationToken) {
|
||||
// Show a progress message
|
||||
vscode.window.withProgress(
|
||||
{
|
||||
location: vscode.ProgressLocation.Notification,
|
||||
title: "Getting a completion from Ollama...",
|
||||
title: "Ollama Autocoder",
|
||||
cancellable: true,
|
||||
},
|
||||
async (progress, progressCancellationToken) => {
|
||||
try {
|
||||
progress.report({ message: "Starting model..." });
|
||||
|
||||
// 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
|
||||
prompt: prompt,
|
||||
stream: true,
|
||||
system: apiSystemMessage,
|
||||
raw: rawInput,
|
||||
options: {
|
||||
num_predict: numPredict
|
||||
},
|
||||
}
|
||||
}, {
|
||||
cancelToken: new axios.CancelToken((c) => {
|
||||
const cancelPost = function () {
|
||||
|
|
@ -48,6 +68,8 @@ async function autocompleteCommand(document: vscode.TextDocument, position: vsco
|
|||
let currentPosition = position;
|
||||
|
||||
response.data.on('data', async (d: Uint8Array) => {
|
||||
progress.report({ message: "Generating..." });
|
||||
|
||||
// Get a completion from the response
|
||||
const completion: string = JSON.parse(d.toString()).response;
|
||||
|
||||
|
|
@ -75,7 +97,7 @@ async function autocompleteCommand(document: vscode.TextDocument, position: vsco
|
|||
currentPosition = newPosition;
|
||||
|
||||
// completion bar
|
||||
progress.report({ increment: 1 / (numPredict/100) });
|
||||
progress.report({ message: "Generating...", increment: 1 / (numPredict/100) });
|
||||
|
||||
// move cursor
|
||||
const editor = vscode.window.activeTextEditor;
|
||||
|
|
@ -97,12 +119,13 @@ async function autocompleteCommand(document: vscode.TextDocument, position: vsco
|
|||
vscode.window.showErrorMessage(
|
||||
"Ollama encountered an error: " + err.message
|
||||
);
|
||||
console.log(err);
|
||||
}
|
||||
}
|
||||
);
|
||||
}
|
||||
|
||||
// This method is called when your extension is activated
|
||||
// This method is called when extension is activated
|
||||
function activate(context: vscode.ExtensionContext) {
|
||||
// Register a completion provider for JavaScript files
|
||||
const provider = vscode.languages.registerCompletionItemProvider("*", {
|
||||
|
|
@ -120,7 +143,7 @@ function activate(context: vscode.ExtensionContext) {
|
|||
item.documentation = new vscode.MarkdownString('Press `Enter` to get a completion from Ollama');
|
||||
// Set the command to trigger the completion
|
||||
item.command = {
|
||||
command: 'ollama-coder.autocomplete',
|
||||
command: 'ollama-autocoder.autocomplete',
|
||||
title: 'Ollama',
|
||||
arguments: [document, position, prompt, cancellationToken]
|
||||
};
|
||||
|
|
@ -137,7 +160,7 @@ function activate(context: vscode.ExtensionContext) {
|
|||
|
||||
// Register a command for getting a completion from Ollama
|
||||
const disposable = vscode.commands.registerCommand(
|
||||
"ollama-coder.autocomplete",
|
||||
"ollama-autocoder.autocomplete",
|
||||
autocompleteCommand
|
||||
);
|
||||
|
||||
|
|
@ -145,7 +168,7 @@ function activate(context: vscode.ExtensionContext) {
|
|||
context.subscriptions.push(disposable);
|
||||
}
|
||||
|
||||
// This method is called when your extension is deactivated
|
||||
// This method is called when extension is deactivated
|
||||
function deactivate() { }
|
||||
|
||||
module.exports = {
|
||||
|
|
|
|||
Loading…
Reference in New Issue