mirror of
https://github.com/getcompanion-ai/co-mono.git
synced 2026-04-16 22:03:45 +00:00
refactor(ai): Implement unified model system with type-safe createLLM
- Add Model interface to types.ts with normalized structure - Create type-safe generic createLLM function with provider-specific model constraints - Generate models from OpenRouter API and models.dev data - Strip provider prefixes for direct providers (google, openai, anthropic, xai) - Keep full model IDs for OpenRouter-proxied models - Clean separation: types.ts (Model interface), models.ts (factory logic), models.generated.ts (data) - Remove old model scripts and unused dependencies - Rename GeminiLLM to GoogleLLM for consistency - Add tests for new providers (xAI, Groq, Cerebras, OpenRouter) - Support 181 tool-capable models across 7 providers with full type safety
This commit is contained in:
Mario Zechner
parent
3f36051bc6
commit
c7618db3f7
8 changed files with 409 additions and 418 deletions
|
|
@ -3,263 +3,260 @@
|
|||
import { readFileSync, writeFileSync } from "fs";
|
||||
import { join } from "path";
|
||||
|
||||
// Load the models.json file
|
||||
const data = JSON.parse(readFileSync(join(process.cwd(), "src/models.json"), "utf-8"));
|
||||
|
||||
// Categorize providers by their API type
|
||||
const openaiModels: Record<string, any> = {};
|
||||
const openaiCompatibleProviders: Record<string, any> = {};
|
||||
const anthropicModels: Record<string, any> = {};
|
||||
const geminiModels: Record<string, any> = {};
|
||||
|
||||
for (const [providerId, provider] of Object.entries(data)) {
|
||||
const p = provider as any;
|
||||
|
||||
if (providerId === "openai") {
|
||||
// All OpenAI models use the Responses API
|
||||
openaiModels[providerId] = p;
|
||||
} else if (providerId === "anthropic" || providerId === "google-vertex-anthropic") {
|
||||
// Anthropic direct and via Vertex
|
||||
anthropicModels[providerId] = p;
|
||||
} else if (providerId === "google" || providerId === "google-vertex") {
|
||||
// Google Gemini models
|
||||
geminiModels[providerId] = p;
|
||||
} else if (p.npm === "@ai-sdk/openai-compatible" ||
|
||||
p.npm === "@ai-sdk/groq" ||
|
||||
p.npm === "@ai-sdk/cerebras" ||
|
||||
p.npm === "@ai-sdk/fireworks" ||
|
||||
p.npm === "@ai-sdk/openrouter" ||
|
||||
p.npm === "@ai-sdk/openai" && providerId !== "openai" ||
|
||||
p.api?.includes("/v1") ||
|
||||
["together", "ollama", "llama", "github-models", "groq", "cerebras", "openrouter", "fireworks"].includes(providerId)) {
|
||||
// OpenAI-compatible providers - they all speak the OpenAI completions API
|
||||
// Set default base URLs for known providers
|
||||
if (!p.api) {
|
||||
switch (providerId) {
|
||||
case "groq": p.api = "https://api.groq.com/openai/v1"; break;
|
||||
case "cerebras": p.api = "https://api.cerebras.com/v1"; break;
|
||||
case "together": p.api = "https://api.together.xyz/v1"; break;
|
||||
case "fireworks": p.api = "https://api.fireworks.ai/v1"; break;
|
||||
}
|
||||
}
|
||||
openaiCompatibleProviders[providerId] = p;
|
||||
}
|
||||
interface ModelsDevModel {
|
||||
id: string;
|
||||
name: string;
|
||||
tool_call?: boolean;
|
||||
reasoning?: boolean;
|
||||
limit?: {
|
||||
context?: number;
|
||||
output?: number;
|
||||
};
|
||||
cost?: {
|
||||
input?: number;
|
||||
output?: number;
|
||||
cache_read?: number;
|
||||
cache_write?: number;
|
||||
};
|
||||
modalities?: {
|
||||
input?: string[];
|
||||
};
|
||||
}
|
||||
|
||||
// Generate the TypeScript file
|
||||
let output = `// This file is auto-generated by scripts/generate-models.ts
|
||||
interface NormalizedModel {
|
||||
id: string;
|
||||
name: string;
|
||||
provider: string;
|
||||
reasoning: boolean;
|
||||
input: ("text" | "image")[];
|
||||
cost: {
|
||||
input: number;
|
||||
output: number;
|
||||
cacheRead: number;
|
||||
cacheWrite: number;
|
||||
};
|
||||
contextWindow: number;
|
||||
maxTokens: number;
|
||||
}
|
||||
|
||||
async function fetchOpenRouterModels(): Promise<NormalizedModel[]> {
|
||||
try {
|
||||
console.log("🌐 Fetching models from OpenRouter API...");
|
||||
const response = await fetch("https://openrouter.ai/api/v1/models");
|
||||
const data = await response.json();
|
||||
|
||||
const models: NormalizedModel[] = [];
|
||||
|
||||
for (const model of data.data) {
|
||||
// Only include models that support tools
|
||||
if (!model.supported_parameters?.includes("tools")) continue;
|
||||
|
||||
// Parse provider from model ID
|
||||
const [providerPrefix] = model.id.split("/");
|
||||
let provider = "";
|
||||
let modelKey = model.id;
|
||||
|
||||
// Map provider prefixes to our provider names
|
||||
if (model.id.startsWith("google/")) {
|
||||
provider = "google";
|
||||
modelKey = model.id.replace("google/", "");
|
||||
} else if (model.id.startsWith("openai/")) {
|
||||
provider = "openai";
|
||||
modelKey = model.id.replace("openai/", "");
|
||||
} else if (model.id.startsWith("anthropic/")) {
|
||||
provider = "anthropic";
|
||||
modelKey = model.id.replace("anthropic/", "");
|
||||
} else if (model.id.startsWith("x-ai/")) {
|
||||
provider = "xai";
|
||||
modelKey = model.id.replace("x-ai/", "");
|
||||
} else {
|
||||
// All other models go through OpenRouter
|
||||
provider = "openrouter";
|
||||
modelKey = model.id; // Keep full ID for OpenRouter
|
||||
}
|
||||
|
||||
// Skip if not one of our supported providers
|
||||
if (!["google", "openai", "anthropic", "xai", "openrouter"].includes(provider)) {
|
||||
continue;
|
||||
}
|
||||
|
||||
// Parse input modalities
|
||||
const input: ("text" | "image")[] = ["text"];
|
||||
if (model.architecture?.modality?.includes("image")) {
|
||||
input.push("image");
|
||||
}
|
||||
|
||||
// Convert pricing from $/token to $/million tokens
|
||||
const inputCost = parseFloat(model.pricing?.prompt || "0") * 1_000_000;
|
||||
const outputCost = parseFloat(model.pricing?.completion || "0") * 1_000_000;
|
||||
const cacheReadCost = parseFloat(model.pricing?.input_cache_read || "0") * 1_000_000;
|
||||
const cacheWriteCost = parseFloat(model.pricing?.input_cache_write || "0") * 1_000_000;
|
||||
|
||||
models.push({
|
||||
id: modelKey,
|
||||
name: model.name,
|
||||
provider,
|
||||
reasoning: model.supported_parameters?.includes("reasoning") || false,
|
||||
input,
|
||||
cost: {
|
||||
input: inputCost,
|
||||
output: outputCost,
|
||||
cacheRead: cacheReadCost,
|
||||
cacheWrite: cacheWriteCost,
|
||||
},
|
||||
contextWindow: model.context_length || 4096,
|
||||
maxTokens: model.top_provider?.max_completion_tokens || 4096,
|
||||
});
|
||||
}
|
||||
|
||||
console.log(`✅ Fetched ${models.length} tool-capable models from OpenRouter`);
|
||||
return models;
|
||||
} catch (error) {
|
||||
console.error("❌ Failed to fetch OpenRouter models:", error);
|
||||
return [];
|
||||
}
|
||||
}
|
||||
|
||||
function loadModelsDevData(): NormalizedModel[] {
|
||||
try {
|
||||
console.log("📁 Loading models from models.json...");
|
||||
const data = JSON.parse(readFileSync(join(process.cwd(), "src/models.json"), "utf-8"));
|
||||
|
||||
const models: NormalizedModel[] = [];
|
||||
|
||||
// Process Groq models
|
||||
if (data.groq?.models) {
|
||||
for (const [modelId, model] of Object.entries(data.groq.models)) {
|
||||
const m = model as ModelsDevModel;
|
||||
if (m.tool_call !== true) continue;
|
||||
|
||||
models.push({
|
||||
id: modelId,
|
||||
name: m.name || modelId,
|
||||
provider: "groq",
|
||||
reasoning: m.reasoning === true,
|
||||
input: m.modalities?.input?.includes("image") ? ["text", "image"] : ["text"],
|
||||
cost: {
|
||||
input: m.cost?.input || 0,
|
||||
output: m.cost?.output || 0,
|
||||
cacheRead: m.cost?.cache_read || 0,
|
||||
cacheWrite: m.cost?.cache_write || 0,
|
||||
},
|
||||
contextWindow: m.limit?.context || 4096,
|
||||
maxTokens: m.limit?.output || 4096,
|
||||
});
|
||||
}
|
||||
}
|
||||
|
||||
// Process Cerebras models
|
||||
if (data.cerebras?.models) {
|
||||
for (const [modelId, model] of Object.entries(data.cerebras.models)) {
|
||||
const m = model as ModelsDevModel;
|
||||
if (m.tool_call !== true) continue;
|
||||
|
||||
models.push({
|
||||
id: modelId,
|
||||
name: m.name || modelId,
|
||||
provider: "cerebras",
|
||||
reasoning: m.reasoning === true,
|
||||
input: m.modalities?.input?.includes("image") ? ["text", "image"] : ["text"],
|
||||
cost: {
|
||||
input: m.cost?.input || 0,
|
||||
output: m.cost?.output || 0,
|
||||
cacheRead: m.cost?.cache_read || 0,
|
||||
cacheWrite: m.cost?.cache_write || 0,
|
||||
},
|
||||
contextWindow: m.limit?.context || 4096,
|
||||
maxTokens: m.limit?.output || 4096,
|
||||
});
|
||||
}
|
||||
}
|
||||
|
||||
console.log(`✅ Loaded ${models.length} tool-capable models from models.dev`);
|
||||
return models;
|
||||
} catch (error) {
|
||||
console.error("❌ Failed to load models.dev data:", error);
|
||||
return [];
|
||||
}
|
||||
}
|
||||
|
||||
async function generateModels() {
|
||||
// Fetch all models
|
||||
const openRouterModels = await fetchOpenRouterModels();
|
||||
const modelsDevModels = loadModelsDevData();
|
||||
|
||||
// Combine models (models.dev takes priority for Groq/Cerebras)
|
||||
const allModels = [...modelsDevModels, ...openRouterModels];
|
||||
|
||||
// Group by provider
|
||||
const providers: Record<string, NormalizedModel[]> = {};
|
||||
for (const model of allModels) {
|
||||
if (!providers[model.provider]) {
|
||||
providers[model.provider] = [];
|
||||
}
|
||||
providers[model.provider].push(model);
|
||||
}
|
||||
|
||||
// Generate TypeScript file
|
||||
let output = `// This file is auto-generated by scripts/generate-models.ts
|
||||
// Do not edit manually - run 'npm run generate-models' to update
|
||||
|
||||
import type { ModalityInput, ModalityOutput } from "./models.js";
|
||||
|
||||
export interface ModelData {
|
||||
id: string;
|
||||
name: string;
|
||||
reasoning: boolean;
|
||||
tool_call: boolean;
|
||||
attachment: boolean;
|
||||
temperature: boolean;
|
||||
knowledge?: string;
|
||||
release_date: string;
|
||||
last_updated: string;
|
||||
modalities: {
|
||||
input: ModalityInput[];
|
||||
output: ModalityOutput[];
|
||||
};
|
||||
open_weights: boolean;
|
||||
limit: {
|
||||
context: number;
|
||||
output: number;
|
||||
};
|
||||
cost?: {
|
||||
input: number;
|
||||
output: number;
|
||||
cache_read?: number;
|
||||
cache_write?: number;
|
||||
};
|
||||
}
|
||||
|
||||
export interface ProviderData {
|
||||
id: string;
|
||||
name: string;
|
||||
baseUrl?: string;
|
||||
env?: string[];
|
||||
models: Record<string, ModelData>;
|
||||
}
|
||||
import type { Model } from "./types.js";
|
||||
|
||||
export const PROVIDERS = {
|
||||
`;
|
||||
|
||||
// Generate OpenAI models
|
||||
output += `// OpenAI models - all use OpenAIResponsesLLM\n`;
|
||||
output += `export const OPENAI_MODELS = {\n`;
|
||||
for (const [providerId, provider] of Object.entries(openaiModels)) {
|
||||
const p = provider as any;
|
||||
for (const [modelId, model] of Object.entries(p.models || {})) {
|
||||
const m = model as any;
|
||||
output += ` "${modelId}": ${JSON.stringify(m, null, 8).split('\n').join('\n ')},\n`;
|
||||
}
|
||||
}
|
||||
output += `} as const;\n\n`;
|
||||
// Generate provider sections
|
||||
for (const [providerId, models] of Object.entries(providers)) {
|
||||
output += `\t${providerId}: {\n`;
|
||||
output += `\t\tmodels: {\n`;
|
||||
|
||||
// Generate OpenAI-compatible providers
|
||||
output += `// OpenAI-compatible providers - use OpenAICompletionsLLM\n`;
|
||||
output += `export const OPENAI_COMPATIBLE_PROVIDERS = {\n`;
|
||||
for (const [providerId, provider] of Object.entries(openaiCompatibleProviders)) {
|
||||
const p = provider as any;
|
||||
output += ` "${providerId}": {\n`;
|
||||
output += ` id: "${providerId}",\n`;
|
||||
output += ` name: "${p.name}",\n`;
|
||||
if (p.api) {
|
||||
output += ` baseUrl: "${p.api}",\n`;
|
||||
}
|
||||
if (p.env) {
|
||||
output += ` env: ${JSON.stringify(p.env)},\n`;
|
||||
}
|
||||
output += ` models: {\n`;
|
||||
for (const [modelId, model] of Object.entries(p.models || {})) {
|
||||
const m = model as any;
|
||||
output += ` "${modelId}": ${JSON.stringify(m, null, 12).split('\n').join('\n ')},\n`;
|
||||
}
|
||||
output += ` }\n`;
|
||||
output += ` },\n`;
|
||||
}
|
||||
output += `} as const;\n\n`;
|
||||
for (const model of models) {
|
||||
output += `\t\t\t"${model.id}": {\n`;
|
||||
output += `\t\t\t\tid: "${model.id}",\n`;
|
||||
output += `\t\t\t\tname: "${model.name}",\n`;
|
||||
output += `\t\t\t\tprovider: "${model.provider}",\n`;
|
||||
output += `\t\t\t\treasoning: ${model.reasoning},\n`;
|
||||
output += `\t\t\t\tinput: ${JSON.stringify(model.input)},\n`;
|
||||
output += `\t\t\t\tcost: {\n`;
|
||||
output += `\t\t\t\t\tinput: ${model.cost.input},\n`;
|
||||
output += `\t\t\t\t\toutput: ${model.cost.output},\n`;
|
||||
output += `\t\t\t\t\tcacheRead: ${model.cost.cacheRead},\n`;
|
||||
output += `\t\t\t\t\tcacheWrite: ${model.cost.cacheWrite},\n`;
|
||||
output += `\t\t\t\t},\n`;
|
||||
output += `\t\t\t\tcontextWindow: ${model.contextWindow},\n`;
|
||||
output += `\t\t\t\tmaxTokens: ${model.maxTokens},\n`;
|
||||
output += `\t\t\t} satisfies Model,\n`;
|
||||
}
|
||||
|
||||
// Generate Anthropic models (avoiding duplicates)
|
||||
output += `// Anthropic models - use AnthropicLLM\n`;
|
||||
output += `export const ANTHROPIC_MODELS = {\n`;
|
||||
const seenAnthropicModels = new Set<string>();
|
||||
for (const [providerId, provider] of Object.entries(anthropicModels)) {
|
||||
const p = provider as any;
|
||||
for (const [modelId, model] of Object.entries(p.models || {})) {
|
||||
if (!seenAnthropicModels.has(modelId)) {
|
||||
seenAnthropicModels.add(modelId);
|
||||
const m = model as any;
|
||||
output += ` "${modelId}": ${JSON.stringify(m, null, 8).split('\n').join('\n ')},\n`;
|
||||
}
|
||||
}
|
||||
}
|
||||
output += `} as const;\n\n`;
|
||||
output += `\t\t}\n`;
|
||||
output += `\t},\n`;
|
||||
}
|
||||
|
||||
// Generate Gemini models (avoiding duplicates)
|
||||
output += `// Gemini models - use GeminiLLM\n`;
|
||||
output += `export const GEMINI_MODELS = {\n`;
|
||||
const seenGeminiModels = new Set<string>();
|
||||
for (const [providerId, provider] of Object.entries(geminiModels)) {
|
||||
const p = provider as any;
|
||||
for (const [modelId, model] of Object.entries(p.models || {})) {
|
||||
if (!seenGeminiModels.has(modelId)) {
|
||||
seenGeminiModels.add(modelId);
|
||||
const m = model as any;
|
||||
output += ` "${modelId}": ${JSON.stringify(m, null, 8).split('\n').join('\n ')},\n`;
|
||||
}
|
||||
}
|
||||
}
|
||||
output += `} as const;\n\n`;
|
||||
output += `} as const;
|
||||
|
||||
// Generate type helpers
|
||||
output += `// Type helpers\n`;
|
||||
output += `export type OpenAIModel = keyof typeof OPENAI_MODELS;\n`;
|
||||
output += `export type OpenAICompatibleProvider = keyof typeof OPENAI_COMPATIBLE_PROVIDERS;\n`;
|
||||
output += `export type AnthropicModel = keyof typeof ANTHROPIC_MODELS;\n`;
|
||||
output += `export type GeminiModel = keyof typeof GEMINI_MODELS;\n\n`;
|
||||
|
||||
// Generate the factory function
|
||||
output += `// Factory function implementation\n`;
|
||||
output += `import { OpenAIResponsesLLM } from "./providers/openai-responses.js";\n`;
|
||||
output += `import { OpenAICompletionsLLM } from "./providers/openai-completions.js";\n`;
|
||||
output += `import { AnthropicLLM } from "./providers/anthropic.js";\n`;
|
||||
output += `import { GeminiLLM } from "./providers/gemini.js";\n`;
|
||||
output += `import type { LLM, LLMOptions } from "./types.js";\n\n`;
|
||||
|
||||
output += `export interface CreateLLMOptions {
|
||||
apiKey?: string;
|
||||
baseUrl?: string;
|
||||
}
|
||||
|
||||
// Overloads for type safety
|
||||
export function createLLM(
|
||||
provider: "openai",
|
||||
model: OpenAIModel,
|
||||
options?: CreateLLMOptions
|
||||
): OpenAIResponsesLLM;
|
||||
|
||||
export function createLLM(
|
||||
provider: OpenAICompatibleProvider,
|
||||
model: string, // We'll validate at runtime
|
||||
options?: CreateLLMOptions
|
||||
): OpenAICompletionsLLM;
|
||||
|
||||
export function createLLM(
|
||||
provider: "anthropic",
|
||||
model: AnthropicModel,
|
||||
options?: CreateLLMOptions
|
||||
): AnthropicLLM;
|
||||
|
||||
export function createLLM(
|
||||
provider: "gemini",
|
||||
model: GeminiModel,
|
||||
options?: CreateLLMOptions
|
||||
): GeminiLLM;
|
||||
|
||||
// Implementation
|
||||
export function createLLM(
|
||||
provider: string,
|
||||
model: string,
|
||||
options?: CreateLLMOptions
|
||||
): LLM<LLMOptions> {
|
||||
const apiKey = options?.apiKey || process.env[getEnvVar(provider)];
|
||||
|
||||
if (provider === "openai") {
|
||||
return new OpenAIResponsesLLM(model, apiKey);
|
||||
}
|
||||
|
||||
if (provider === "anthropic") {
|
||||
return new AnthropicLLM(model, apiKey);
|
||||
}
|
||||
|
||||
if (provider === "gemini") {
|
||||
return new GeminiLLM(model, apiKey);
|
||||
}
|
||||
|
||||
// OpenAI-compatible providers
|
||||
if (provider in OPENAI_COMPATIBLE_PROVIDERS) {
|
||||
const providerData = OPENAI_COMPATIBLE_PROVIDERS[provider as OpenAICompatibleProvider];
|
||||
const baseUrl = options?.baseUrl || providerData.baseUrl;
|
||||
return new OpenAICompletionsLLM(model, apiKey, baseUrl);
|
||||
}
|
||||
|
||||
throw new Error(\`Unknown provider: \${provider}\`);
|
||||
}
|
||||
|
||||
// Helper to get the default environment variable for a provider
|
||||
function getEnvVar(provider: string): string {
|
||||
switch (provider) {
|
||||
case "openai": return "OPENAI_API_KEY";
|
||||
case "anthropic": return "ANTHROPIC_API_KEY";
|
||||
case "gemini": return "GEMINI_API_KEY";
|
||||
case "groq": return "GROQ_API_KEY";
|
||||
case "cerebras": return "CEREBRAS_API_KEY";
|
||||
case "together": return "TOGETHER_API_KEY";
|
||||
case "openrouter": return "OPENROUTER_API_KEY";
|
||||
default: return \`\${provider.toUpperCase()}_API_KEY\`;
|
||||
}
|
||||
}
|
||||
// Helper type to extract models for each provider
|
||||
export type ProviderModels = {
|
||||
[K in keyof typeof PROVIDERS]: typeof PROVIDERS[K]["models"]
|
||||
};
|
||||
`;
|
||||
|
||||
// Write the generated file
|
||||
writeFileSync(join(process.cwd(), "src/models.generated.ts"), output);
|
||||
console.log("✅ Generated src/models.generated.ts");
|
||||
// Write file
|
||||
writeFileSync(join(process.cwd(), "src/models.generated.ts"), output);
|
||||
console.log("✅ Generated src/models.generated.ts");
|
||||
|
||||
// Count statistics
|
||||
const openaiCount = Object.values(openaiModels).reduce((acc, p: any) => acc + Object.keys(p.models || {}).length, 0);
|
||||
const compatCount = Object.values(openaiCompatibleProviders).reduce((acc, p: any) => acc + Object.keys(p.models || {}).length, 0);
|
||||
const anthropicCount = Object.values(anthropicModels).reduce((acc, p: any) => acc + Object.keys(p.models || {}).length, 0);
|
||||
const geminiCount = Object.values(geminiModels).reduce((acc, p: any) => acc + Object.keys(p.models || {}).length, 0);
|
||||
// Print statistics
|
||||
const totalModels = allModels.length;
|
||||
const reasoningModels = allModels.filter(m => m.reasoning).length;
|
||||
|
||||
console.log(`\nModel counts:`);
|
||||
console.log(` OpenAI (Responses API): ${openaiCount} models`);
|
||||
console.log(` OpenAI-compatible: ${compatCount} models across ${Object.keys(openaiCompatibleProviders).length} providers`);
|
||||
console.log(` Anthropic: ${anthropicCount} models`);
|
||||
console.log(` Gemini: ${geminiCount} models`);
|
||||
console.log(` Total: ${openaiCount + compatCount + anthropicCount + geminiCount} models`);
|
||||
console.log(`\n📊 Model Statistics:`);
|
||||
console.log(` Total tool-capable models: ${totalModels}`);
|
||||
console.log(` Reasoning-capable models: ${reasoningModels}`);
|
||||
|
||||
for (const [provider, models] of Object.entries(providers)) {
|
||||
console.log(` ${provider}: ${models.length} models`);
|
||||
}
|
||||
}
|
||||
|
||||
// Run the generator
|
||||
generateModels().catch(console.error);
|
||||
Loading…
Add table
Add a link
Reference in a new issue