@unified-llm/core
Version:
Unified LLM interface.
270 lines • 10.8 kB
JavaScript
;
Object.defineProperty(exports, "__esModule", { value: true });
exports.CLIENT_PRESETS = exports.ClientManager = void 0;
const llm_client_1 = require("./llm-client");
const client_repository_1 = require("./database/client-repository");
/**
* ClientManager provides methods to manage LLMClient configurations,
*/
class ClientManager {
/**
* Create a new LLMClient configuration
*/
static async create(config) {
return await llm_client_1.LLMClient.save(config);
}
/**
* Create a new LLMClient configuration preset
*/
static async createPreset(preset, customConfig) {
const presetConfig = exports.CLIENT_PRESETS[preset];
if (!presetConfig) {
throw new Error(`Unknown preset: ${preset}`);
}
const config = {
...presetConfig,
...customConfig,
name: (customConfig === null || customConfig === void 0 ? void 0 : customConfig.name) || presetConfig.name,
};
return await llm_client_1.LLMClient.save(config);
}
/**
* Clone an existing LLMClient configuration
*/
static async clone(sourceId, newName, updates) {
const sourceConfig = await llm_client_1.LLMClient.getConfig(sourceId);
if (!sourceConfig) {
throw new Error(`LLMClient ${sourceId} not found`);
}
const clonedConfig = {
...sourceConfig,
...updates,
id: undefined, // Generate new ID
name: newName,
};
return await llm_client_1.LLMClient.save(clonedConfig);
}
/**
* Import an LLMClient configuration from JSON
*/
static async import(configJson) {
try {
const config = JSON.parse(configJson);
// IDがある場合は削除（新しいIDを生成）
if (config.id) {
delete config.id;
}
return await llm_client_1.LLMClient.save(config);
}
catch (error) {
throw new Error(`Failed to import agent configuration: ${error}`);
}
}
/**
* Export an LLMClient configuration to JSON
*/
static async export(id) {
const config = await llm_client_1.LLMClient.getConfig(id);
if (!config) {
throw new Error(`LLMClient ${id} not found`);
}
// APIキーは除外してエクスポート
const exportConfig = { ...config };
delete exportConfig.apiKey;
return JSON.stringify(exportConfig, null, 2);
}
/**
* Search for LLMClients based on various criteria
*/
static async search(query) {
let results = await llm_client_1.LLMClient.list();
if (query.name) {
const stored = await client_repository_1.clientRepository.findByName(query.name);
results = stored.map(s => client_repository_1.ClientRepository.toConfig(s));
}
if (query.provider) {
results = results.filter(a => a.provider === query.provider);
}
if (query.tags && query.tags.length > 0) {
results = results.filter(a => a.tags && query.tags && query.tags.some(tag => a.tags ? a.tags.includes(tag) : false));
}
if (query.model) {
results = results.filter(a => a.model === query.model);
}
return results;
}
/**
* Batch operations for creating, updating, deleting, activating, or deactivating LLMClients
*/
static async batch(operations) {
const results = [];
for (const op of operations) {
try {
let result;
switch (op.type) {
case 'create':
if (!op.config)
throw new Error('Config required for create operation');
result = await llm_client_1.LLMClient.save(op.config);
break;
case 'update':
if (!op.id || !op.updates)
throw new Error('ID and updates required for update operation');
result = await llm_client_1.LLMClient.update(op.id, op.updates);
break;
case 'delete':
if (!op.id)
throw new Error('ID required for delete operation');
result = await llm_client_1.LLMClient.delete(op.id);
break;
case 'activate':
if (!op.id)
throw new Error('ID required for activate operation');
result = await llm_client_1.LLMClient.update(op.id, { isActive: true });
break;
case 'deactivate':
if (!op.id)
throw new Error('ID required for deactivate operation');
result = await llm_client_1.LLMClient.update(op.id, { isActive: false });
break;
default:
throw new Error(`Unknown operation type: ${op.type}`);
}
results.push({
operation: op,
success: true,
result
});
}
catch (error) {
results.push({
operation: op,
success: false,
error: error instanceof Error ? error.message : String(error)
});
}
}
return results;
}
/**
* Get statistics about LLMClients
*/
static async getStats() {
const all = await llm_client_1.LLMClient.list({ includeInactive: true });
const active = all.filter(a => a.isActive !== false);
const byProvider = all.reduce((acc, a) => {
acc[a.provider] = (acc[a.provider] || 0) + 1;
return acc;
}, {});
const allTags = all
.flatMap(a => a.tags || [])
.reduce((acc, tag) => {
acc[tag] = (acc[tag] || 0) + 1;
return acc;
}, {});
return {
total: all.length,
active: active.length,
inactive: all.length - active.length,
byProvider,
popularTags: Object.entries(allTags)
.sort(([, a], [, b]) => b - a)
.slice(0, 10)
.map(([tag, count]) => ({ tag, count })),
recentlyUpdated: all
.filter(a => a.updatedAt)
.sort((a, b) => {
const dateA = a.updatedAt ? new Date(a.updatedAt).getTime() : 0;
const dateB = b.updatedAt ? new Date(b.updatedAt).getTime() : 0;
return dateB - dateA;
})
.slice(0, 5)
.map(a => ({ id: a.id || '', name: a.name, updatedAt: a.updatedAt }))
};
}
}
exports.ClientManager = ClientManager;
exports.CLIENT_PRESETS = {
'coding-agent': {
name: 'Coding Assistant',
description: 'Helps with programming tasks, code review, and debugging',
provider: 'openai',
model: 'gpt-4.1-mini',
systemPrompt: 'You are a helpful coding agent. Provide clear, well-commented code examples and explain your reasoning.',
instructions: 'Focus on code quality, best practices, and security. Always explain your solutions.',
generationConfig: {
temperature: 0.1,
max_tokens: 2000,
},
tools: ['cat', 'tree', 'getProjectInfo'],
tags: ['coding', 'development', 'programming'],
},
'creative-writer': {
name: 'Creative Writer',
description: 'Assists with creative writing, storytelling, and content creation',
provider: 'anthropic',
model: 'claude-3-sonnet-20240229',
systemPrompt: 'You are a creative writing agent. Help users craft engaging stories, poems, and creative content.',
instructions: 'Be imaginative, encourage creativity, and provide constructive feedback on writing.',
generationConfig: {
temperature: 0.8,
max_tokens: 4000,
},
tags: ['writing', 'creative', 'content'],
},
'data-analyst': {
name: 'Data Analyst',
description: 'Helps with data analysis, statistics, and insights',
provider: 'openai',
model: 'gpt-4.1',
systemPrompt: 'You are a data analyst. Help users understand data, create visualizations, and derive insights.',
instructions: 'Focus on accuracy, provide step-by-step analysis, and suggest actionable insights.',
generationConfig: {
temperature: 0.2,
max_tokens: 3000,
},
tools: ['getCurrentTime', 'getProjectInfo'],
tags: ['data', 'analysis', 'statistics'],
},
'translator': {
name: 'Translator',
description: 'Provides accurate translations between languages',
provider: 'google',
model: 'gemini-1.5-pro',
systemPrompt: 'You are a professional translator. Provide accurate, natural translations while preserving meaning and context.',
instructions: 'Consider cultural nuances, provide alternative translations when appropriate.',
generationConfig: {
temperature: 0.3,
max_tokens: 2000,
},
tags: ['translation', 'language', 'localization'],
},
'customer-support': {
name: 'Customer Support',
description: 'Provides helpful customer service responses',
provider: 'anthropic',
model: 'claude-3-haiku-20240307',
systemPrompt: 'You are a friendly customer support representative. Be helpful, empathetic, and solution-focused.',
instructions: 'Always be polite, acknowledge concerns, and provide clear next steps.',
generationConfig: {
temperature: 0.4,
max_tokens: 1000,
},
tags: ['support', 'customer-service', 'help'],
},
'research-agent': {
name: 'Research Assistant',
description: 'Helps with research, fact-checking, and information gathering',
provider: 'openai',
model: 'gpt-4.1',
systemPrompt: 'You are a research agent. Help users find accurate information, cite sources, and organize research.',
instructions: 'Be thorough, fact-check information, and provide credible sources when possible.',
generationConfig: {
temperature: 0.2,
max_tokens: 3000,
},
tools: ['getCurrentTime'],
tags: ['research', 'information', 'analysis'],
},
};
//# sourceMappingURL=client-manager.js.map