SPB Git

spb/polyllm

Public
12.9 KB · 251 lines typescript
1/**2 * Real-API provider test matrix. Runs each capability against every provider with the3 * owner keys from .env and writes docs/provider-test-matrix.md. Each cell is ✅ only when4 * the real request succeeded. Cheap prompts, small max tokens.5 *6 *   pnpm providers:matrix              → all providers7 *   pnpm providers:matrix anthropic    → one provider8 */9import "./load-env";10import fs from "node:fs";11import zlib from "node:zlib";12import { getAdapter } from "@/lib/ai/providers";13import { PROVIDER_IDS, type PolyModel, type ProviderId, type UnifiedChatRequest, type UnifiedStreamEvent } from "@/lib/ai/core/types";1415const KEYS: Record<ProviderId, string | undefined> = {16  openai: process.env.OPENAI_API_KEY,17  anthropic: process.env.ANTHROPIC_API_KEY,18  gemini: process.env.GOOGLE_GEMINI_API_KEY,19  xai: process.env.XAI_API_KEY,20};2122/** Cheap, current models per provider used for the matrix. */23const TEST_MODELS: Record<ProviderId, { text: string; reasoning: string; vision: string }> = {24  openai: { text: "gpt-5.4-mini", reasoning: "gpt-5.4-mini", vision: "gpt-5.4-mini" },25  anthropic: { text: "claude-haiku-4-5", reasoning: "claude-sonnet-5", vision: "claude-haiku-4-5" },26  gemini: { text: "gemini-3.5-flash-lite", reasoning: "gemini-3.5-flash-lite", vision: "gemini-3.5-flash-lite" },27  xai: { text: "grok-4.20-0309-non-reasoning", reasoning: "grok-4.3", vision: "grok-4.20-0309-non-reasoning" },28};2930// 2x2 PNG (red, green / blue, white) — 32x32 needed for xAI minimums, so we scale via a bigger canvas.31function makePng(size: number): string {32  // Simple uncompressed PNG writer (RGB) for a solid red square with a white diagonal.33  const crcTable: number[] = [];34  for (let n = 0; n < 256; n++) {35    let c = n;36    for (let k = 0; k < 8; k++) c = c & 1 ? 0xedb88320 ^ (c >>> 1) : c >>> 1;37    crcTable[n] = c >>> 0;38  }39  const crc = (buf: Buffer) => {40    let c = 0xffffffff;41    for (const b of buf) c = crcTable[(c ^ b) & 0xff] ^ (c >>> 8);42    return (c ^ 0xffffffff) >>> 0;43  };44  const chunk = (type: string, data: Buffer) => {45    const len = Buffer.alloc(4);46    len.writeUInt32BE(data.length);47    const td = Buffer.concat([Buffer.from(type), data]);48    const c = Buffer.alloc(4);49    c.writeUInt32BE(crc(td));50    return Buffer.concat([len, td, c]);51  };52  const ihdr = Buffer.alloc(13);53  ihdr.writeUInt32BE(size, 0);54  ihdr.writeUInt32BE(size, 4);55  ihdr[8] = 8;56  ihdr[9] = 2;57  const raw = Buffer.alloc((size * 3 + 1) * size);58  for (let y = 0; y < size; y++) {59    raw[y * (size * 3 + 1)] = 0;60    for (let x = 0; x < size; x++) {61      const o = y * (size * 3 + 1) + 1 + x * 3;62      const diag = Math.abs(x - y) < 2;63      raw[o] = diag ? 255 : 220;64      raw[o + 1] = diag ? 255 : 30;65      raw[o + 2] = diag ? 255 : 30;66    }67  }68  return Buffer.concat([Buffer.from([0x89, 0x50, 0x4e, 0x47, 0x0d, 0x0a, 0x1a, 0x0a]), chunk("IHDR", ihdr), chunk("IDAT", zlib.deflateSync(raw)), chunk("IEND", Buffer.alloc(0))]).toString("base64");69}7071type Cell = "✅" | "❌" | "—";72const ROWS = ["Auth", "Model list", "Text", "Streaming", "System", "Vision", "Tools", "Structured output", "Reasoning", "Token usage", "Invalid key", "Error normalization", "Long response"] as const;73type Row = (typeof ROWS)[number];7475async function collect(req: UnifiedChatRequest, attempt = 1): Promise<Awaited<ReturnType<typeof collectOnce>>> {76  try {77    return await collectOnce(req);78  } catch (e) {79    const poly = (e as { poly?: { code: string; retryAfterMs?: number } }).poly;80    if (poly && (poly.code === "RATE_LIMITED" || poly.code === "PROVIDER_UNAVAILABLE") && attempt <= 3) {81      await new Promise((r) => setTimeout(r, Math.max(3000, poly.retryAfterMs ?? 0) * attempt));82      return collect(req, attempt + 1);83    }84    throw e;85  }86}8788async function collectOnce(req: UnifiedChatRequest) {89  const adapter = getAdapter(req.provider);90  const events: UnifiedStreamEvent[] = [];91  for await (const ev of adapter.streamChat(req)) events.push(ev);92  const err = events.find((e) => e.type === "error");93  if (err && err.type === "error") throw Object.assign(new Error(err.error.message), { poly: err.error });94  const text = events.filter((e) => e.type === "text-delta").map((e) => (e as { text: string }).text).join("");95  const reasoning = events.filter((e) => e.type === "reasoning-delta").map((e) => (e as { text: string }).text).join("");96  const usage = events.find((e) => e.type === "usage") as { usage: { inputTokens: number; outputTokens: number; reasoningTokens?: number } } | undefined;97  const tools = events.filter((e) => e.type === "tool-end") as { name: string; arguments: Record<string, unknown> }[];98  const deltas = events.filter((e) => e.type === "text-delta" || e.type === "reasoning-delta").length;99  return { text, reasoning, usage: usage?.usage, tools, deltas, finish: (events.find((e) => e.type === "finish") as { reason?: string } | undefined)?.reason };100}101102async function runProvider(p: ProviderId, results: Record<Row, Record<ProviderId, Cell>>, notes: string[]) {103  const key = KEYS[p];104  const adapter = getAdapter(p);105  const set = (r: Row, v: Cell) => (results[r][p] = v);106  if (!key) {107    for (const r of ROWS) set(r, "—");108    notes.push(`${p}: no key in env`);109    return;110  }111  const note = (s: string) => notes.push(`${p}: ${s}`);112  const m = TEST_MODELS[p];113  let models: PolyModel[] = [];114  try {115    const v = await adapter.validateApiKey(key);116    set("Auth", v.ok ? "✅" : "❌");117    if (!v.ok) note(`auth: ${v.error?.code} ${v.error?.message}`);118  } catch (e) {119    set("Auth", "❌");120    note(`auth threw: ${(e as Error).message}`);121  }122  try {123    models = await adapter.listModels(key);124    set("Model list", models.length ? "✅" : "❌");125    note(`models: ${models.length} (${models.slice(0, 5).map((x) => x.id).join(", ")}…)`);126  } catch (e) {127    set("Model list", "❌");128    note(`list: ${(e as Error).message}`);129  }130  // Aliases such as `claude-haiku-4-5` resolve to the dated snapshot listed by the provider.131  const info = (id: string) => models.find((x) => x.id === id) ?? models.find((x) => x.id.startsWith(`${id}-`));132  const base = (model: string): UnifiedChatRequest => ({ provider: p, model, apiKey: key, messages: [{ role: "user", content: [{ type: "text", text: "Reply with exactly: OK" }] }], settings: { maxTokens: 2000 }, modelInfo: info(model) });133134  try {135    const r = await collect(base(m.text));136    set("Text", r.text.trim().length ? "✅" : "❌");137    set("Streaming", r.deltas >= 1 ? "✅" : "❌");138    set("Token usage", r.usage && r.usage.inputTokens > 0 && r.usage.outputTokens > 0 ? "✅" : "❌");139    note(`text(${m.text}): "${r.text.trim().slice(0, 40)}" deltas=${r.deltas} usage=${JSON.stringify(r.usage)} finish=${r.finish}`);140  } catch (e) {141    set("Text", "❌");142    set("Streaming", "❌");143    set("Token usage", "❌");144    note(`text: ${(e as Error).message}`);145  }146  try {147    const r = await collect({ ...base(m.text), system: "You are a pirate. Always start your reply with 'Arr'.", messages: [{ role: "user", content: [{ type: "text", text: "Say hello in five words." }] }] });148    set("System", /arr/i.test(r.text) ? "✅" : "❌");149    note(`system: "${r.text.trim().slice(0, 60)}"`);150  } catch (e) {151    set("System", "❌");152    note(`system: ${(e as Error).message}`);153  }154  try {155    const r = await collect({ ...base(m.vision), messages: [{ role: "user", content: [{ type: "image", mimeType: "image/png", data: makePng(64) }, { type: "text", text: "What is the dominant color of this image? One word." }] }] });156    set("Vision", /red/i.test(r.text) ? "✅" : "❌");157    note(`vision: "${r.text.trim().slice(0, 40)}"`);158  } catch (e) {159    set("Vision", "❌");160    note(`vision: ${(e as Error).message}`);161  }162  try {163    const r = await collect({ ...base(m.text), messages: [{ role: "user", content: [{ type: "text", text: "What is 1234 * 5678? Use the calculator tool." }] }], tools: [{ name: "calculator", description: "Evaluate an arithmetic expression", parameters: { type: "object", properties: { expression: { type: "string" } }, required: ["expression"], additionalProperties: false }, strict: true }] });164    set("Tools", r.tools.length && r.tools[0].name === "calculator" ? "✅" : "❌");165    note(`tools: ${JSON.stringify(r.tools.map((t) => ({ n: t.name, a: t.arguments })))} finish=${r.finish}`);166  } catch (e) {167    set("Tools", "❌");168    note(`tools: ${(e as Error).message}`);169  }170  try {171    const r = await collect({ ...base(m.text), settings: { maxTokens: 2000, responseFormat: { type: "json_schema", schema: { type: "object", properties: { city: { type: "string" }, country: { type: "string" } }, required: ["city", "country"], additionalProperties: false }, schemaName: "city" } }, messages: [{ role: "user", content: [{ type: "text", text: "Give the capital of Canada as JSON." }] }] });172    let ok = false;173    try {174      const j = JSON.parse(r.text);175      ok = typeof j.city === "string" && typeof j.country === "string";176    } catch {177      ok = false;178    }179    set("Structured output", ok ? "✅" : "❌");180    note(`structured: ${r.text.trim().slice(0, 80)}`);181  } catch (e) {182    set("Structured output", "❌");183    note(`structured: ${(e as Error).message}`);184  }185  try {186    const r = await collect({ ...base(m.reasoning), settings: { maxTokens: 6000, reasoningEffort: p === "anthropic" ? "high" : "low", includeReasoning: true }, messages: [{ role: "user", content: [{ type: "text", text: "How many prime numbers are there strictly between 1000 and 1100? Think it through carefully, then answer with just the number." }] }] });187    const hasReasoning = r.reasoning.length > 0 || (r.usage?.reasoningTokens ?? 0) > 0;188    set("Reasoning", hasReasoning ? "✅" : "❌");189    note(`reasoning(${m.reasoning}): text="${r.text.trim().slice(0, 30)}" reasoningChars=${r.reasoning.length} reasoningTokens=${r.usage?.reasoningTokens}`);190  } catch (e) {191    set("Reasoning", "❌");192    note(`reasoning: ${(e as Error).message}`);193  }194  try {195    const bad = await adapter.validateApiKey(p === "gemini" ? "AIzaSyInvalidKey0000000000000000000000" : p === "anthropic" ? "sk-ant-api03-invalid" : p === "xai" ? "xai-invalid" : "sk-invalid");196    set("Invalid key", !bad.ok && bad.error?.code === "INVALID_API_KEY" ? "✅" : "❌");197    note(`invalid key → ${bad.error?.code} (${bad.error?.status})`);198  } catch (e) {199    set("Invalid key", "❌");200    note(`invalid key threw: ${(e as Error).message}`);201  }202  try {203    await collect(base("definitely-not-a-model-xyz"));204    set("Error normalization", "❌");205  } catch (e) {206    const poly = (e as { poly?: { code: string } }).poly;207    set("Error normalization", poly?.code === "MODEL_NOT_FOUND" || poly?.code === "INVALID_PARAMETER" ? "✅" : "❌");208    note(`unknown model → ${poly?.code}: ${(e as Error).message.slice(0, 80)}`);209  }210  try {211    const r = await collect({ ...base(m.text), settings: { maxTokens: 3000 }, messages: [{ role: "user", content: [{ type: "text", text: "Write a 400-word story about a lighthouse keeper. Plain prose." }] }] });212    set("Long response", r.text.split(/\s+/).length > 250 ? "✅" : "❌");213    note(`long: ${r.text.split(/\s+/).length} words, ${r.deltas} deltas`);214  } catch (e) {215    set("Long response", "❌");216    note(`long: ${(e as Error).message}`);217  }218}219220async function main() {221  const only = process.argv[2] as ProviderId | undefined;222  const providers = only ? [only] : [...PROVIDER_IDS];223  const results = Object.fromEntries(ROWS.map((r) => [r, Object.fromEntries(PROVIDER_IDS.map((p) => [p, "—" as Cell]))])) as Record<Row, Record<ProviderId, Cell>>;224  const notes: string[] = [];225  await Promise.all(providers.map((p) => runProvider(p, results, notes)));226  // Merge with a previous full matrix so single-provider runs do not blank the others.227  if (only && fs.existsSync("docs/provider-test-matrix.md")) {228    const prev = fs.readFileSync("docs/provider-test-matrix.md", "utf8");229    for (const r of ROWS) {230      const m = prev.match(new RegExp(`^\\| ${r.replace(/[.*+?^${}()|[\]\\]/g, "\\$&")} \\| (.+) \\|$`, "m"));231      if (!m) continue;232      const cells = m[1].split(" | ") as Cell[];233      PROVIDER_IDS.forEach((pp, i) => {234        if (pp !== only && cells[i]) results[r][pp] = cells[i];235      });236    }237  }238  const date = new Date().toISOString().slice(0, 10);239  const header = `| Capability | OpenAI | Anthropic | Gemini | xAI |\n| --- | :-: | :-: | :-: | :-: |`;240  const lines = ROWS.map((r) => `| ${r} | ${PROVIDER_IDS.map((p) => results[r][p]).join(" | ")} |`);241  const md = `# Provider test matrix\n\nGenerated by \`pnpm providers:matrix\` against the REAL provider APIs on ${date}. A cell is ✅ only when the live request succeeded and the assertion passed. “—” = not run (no key).\n\nTest models: ${PROVIDER_IDS.map((p) => `${p}: ${TEST_MODELS[p].text} / reasoning ${TEST_MODELS[p].reasoning}`).join("; ")}.\n\n${header}\n${lines.join("\n")}\n\n## Notes\n\n${notes.map((n) => `- ${n.replace(/\n/g, " ")}`).join("\n")}\n`;242  fs.mkdirSync("docs", { recursive: true });243  fs.writeFileSync("docs/provider-test-matrix.md", md);244  console.log(md);245}246247main().catch((e) => {248  console.error(e);249  process.exit(1);250});251
Files · 573449927df925a6fe97f81f5be20cd588e0a8dd
Menu

Theme