|
| 1 | +import { PanelProvider } from '../panel-provider' |
| 2 | +import { SetRecordingStateMessage } from '../../types/messages' |
| 3 | +import { spawn } from 'child_process' |
| 4 | +import { Logger } from '@shared/utils/logger' |
| 5 | +import { |
| 6 | + ModelProvidersManager, |
| 7 | + ToolConfig |
| 8 | +} from '@/services/model-providers-manager' |
| 9 | +import { PROVIDERS } from '@shared/constants/providers' |
| 10 | +import * as vscode from 'vscode' |
| 11 | +import { apply_reasoning_effort } from '@/utils/apply-reasoning-effort' |
| 12 | +import axios from 'axios' |
| 13 | +import { make_api_request } from '@/utils/make-api-request' |
| 14 | + |
| 15 | +const start_recording = (panel_provider: PanelProvider) => { |
| 16 | + panel_provider.audio_chunks = [] |
| 17 | + try { |
| 18 | + panel_provider.recording_process = spawn('rec', ['-q', '-t', 'wav', '-']) |
| 19 | + |
| 20 | + panel_provider.recording_process.stdout.on('data', (chunk: Buffer) => { |
| 21 | + panel_provider.audio_chunks.push(chunk) |
| 22 | + }) |
| 23 | + } catch (error: any) { |
| 24 | + Logger.error({ |
| 25 | + function_name: 'start_recording', |
| 26 | + message: 'Failed to start recording', |
| 27 | + data: { error } |
| 28 | + }) |
| 29 | + } |
| 30 | +} |
| 31 | + |
| 32 | +const stop_recording = async (panel_provider: PanelProvider) => { |
| 33 | + if (panel_provider.recording_process) { |
| 34 | + panel_provider.recording_process.kill() |
| 35 | + panel_provider.recording_process = null |
| 36 | + |
| 37 | + const audio_buffer = Buffer.concat(panel_provider.audio_chunks) |
| 38 | + const base64_audio = audio_buffer.toString('base64') |
| 39 | + |
| 40 | + panel_provider.audio_chunks = [] |
| 41 | + |
| 42 | + try { |
| 43 | + const model_providers_manager = new ModelProvidersManager( |
| 44 | + panel_provider.context |
| 45 | + ) |
| 46 | + const configs = |
| 47 | + await model_providers_manager.get_voice_input_tool_configs() |
| 48 | + |
| 49 | + if (configs.length == 0) { |
| 50 | + vscode.window.showWarningMessage( |
| 51 | + 'No Voice Input configuration found. Please configure it in settings.', |
| 52 | + { modal: true } |
| 53 | + ) |
| 54 | + return |
| 55 | + } |
| 56 | + |
| 57 | + let config: ToolConfig | undefined = |
| 58 | + await model_providers_manager.get_default_voice_input_config() |
| 59 | + |
| 60 | + if (!config) { |
| 61 | + if (configs.length == 1) { |
| 62 | + config = configs[0] |
| 63 | + } else { |
| 64 | + const selected = await vscode.window.showQuickPick( |
| 65 | + configs.map((c) => ({ |
| 66 | + label: c.model, |
| 67 | + description: c.provider_name, |
| 68 | + config: c |
| 69 | + })), |
| 70 | + { |
| 71 | + title: 'Configurations', |
| 72 | + placeHolder: 'Select a configuration' |
| 73 | + } |
| 74 | + ) |
| 75 | + if (selected) { |
| 76 | + config = selected.config |
| 77 | + } else { |
| 78 | + return |
| 79 | + } |
| 80 | + } |
| 81 | + } |
| 82 | + |
| 83 | + const provider = await model_providers_manager.get_provider( |
| 84 | + config!.provider_name |
| 85 | + ) |
| 86 | + |
| 87 | + if (!provider) { |
| 88 | + vscode.window.showErrorMessage( |
| 89 | + `Provider ${config.provider_name} not found.` |
| 90 | + ) |
| 91 | + return |
| 92 | + } |
| 93 | + |
| 94 | + let endpoint_url = '' |
| 95 | + if (provider.type == 'built-in') { |
| 96 | + const provider_info = PROVIDERS[provider.name as keyof typeof PROVIDERS] |
| 97 | + if (provider_info) { |
| 98 | + endpoint_url = provider_info.base_url |
| 99 | + } |
| 100 | + } else { |
| 101 | + endpoint_url = provider.base_url |
| 102 | + } |
| 103 | + |
| 104 | + const body: { [key: string]: any } = { |
| 105 | + model: config.model, |
| 106 | + messages: [ |
| 107 | + { |
| 108 | + role: 'user', |
| 109 | + content: [ |
| 110 | + { |
| 111 | + type: 'text', |
| 112 | + text: 'Transcribe this audio' |
| 113 | + }, |
| 114 | + { |
| 115 | + type: 'input_audio', |
| 116 | + input_audio: { |
| 117 | + data: base64_audio, |
| 118 | + format: 'wav' |
| 119 | + } |
| 120 | + } |
| 121 | + ] |
| 122 | + } |
| 123 | + ] |
| 124 | + } |
| 125 | + |
| 126 | + apply_reasoning_effort(body, provider, config.reasoning_effort) |
| 127 | + |
| 128 | + const cancel_token_source = axios.CancelToken.source() |
| 129 | + |
| 130 | + const result = await make_api_request({ |
| 131 | + endpoint_url, |
| 132 | + api_key: provider.api_key, |
| 133 | + body, |
| 134 | + cancellation_token: cancel_token_source.token |
| 135 | + }) |
| 136 | + |
| 137 | + if (result?.response) { |
| 138 | + panel_provider.add_text_at_cursor_position(result.response) |
| 139 | + } |
| 140 | + } catch (error: any) { |
| 141 | + Logger.error({ |
| 142 | + function_name: 'stop_recording', |
| 143 | + message: 'Failed to process audio', |
| 144 | + data: { error } |
| 145 | + }) |
| 146 | + vscode.window.showErrorMessage( |
| 147 | + 'Failed to process audio: ' + error.message |
| 148 | + ) |
| 149 | + } |
| 150 | + } |
| 151 | +} |
| 152 | + |
| 153 | +export const handle_voice_input = async ( |
| 154 | + panel_provider: PanelProvider, |
| 155 | + message: SetRecordingStateMessage |
| 156 | +) => { |
| 157 | + if (panel_provider.is_recording === message.is_recording) { |
| 158 | + return |
| 159 | + } |
| 160 | + |
| 161 | + panel_provider.is_recording = message.is_recording |
| 162 | + panel_provider.send_message({ |
| 163 | + command: 'RECORDING_STATE', |
| 164 | + is_recording: panel_provider.is_recording |
| 165 | + }) |
| 166 | + |
| 167 | + if (panel_provider.is_recording) { |
| 168 | + start_recording(panel_provider) |
| 169 | + } else { |
| 170 | + await stop_recording(panel_provider) |
| 171 | + } |
| 172 | +} |
0 commit comments