|
| 1 | +#!/usr/bin/env bun |
| 2 | + |
| 3 | +/** |
| 4 | + * Test script to verify Fireworks AI integration with minimax-m2.5. |
| 5 | + * |
| 6 | + * Usage: |
| 7 | + * # Test 1: Hit Fireworks API directly |
| 8 | + * bun scripts/test-fireworks.ts direct |
| 9 | + * |
| 10 | + * # Test 2: Hit our chat completions endpoint (requires running web server + valid API key) |
| 11 | + * CODEBUFF_API_KEY=<key> bun scripts/test-fireworks.ts endpoint |
| 12 | + * |
| 13 | + * # Run both tests |
| 14 | + * CODEBUFF_API_KEY=<key> bun scripts/test-fireworks.ts both |
| 15 | + */ |
| 16 | + |
| 17 | +const FIREWORKS_BASE_URL = 'https://api.fireworks.ai/inference/v1' |
| 18 | +const FIREWORKS_MODEL = 'accounts/fireworks/models/minimax-m2p5' |
| 19 | +const OPENROUTER_MODEL = 'minimax/minimax-m2.5' |
| 20 | + |
| 21 | +// Same pricing constants as web/src/llm-api/fireworks.ts |
| 22 | +const FIREWORKS_INPUT_COST_PER_TOKEN = 0.30 / 1_000_000 |
| 23 | +const FIREWORKS_CACHED_INPUT_COST_PER_TOKEN = 0.03 / 1_000_000 |
| 24 | +const FIREWORKS_OUTPUT_COST_PER_TOKEN = 1.20 / 1_000_000 |
| 25 | + |
| 26 | +function computeCost(usage: Record<string, unknown>): { cost: number; breakdown: string } { |
| 27 | + const inputTokens = typeof usage.prompt_tokens === 'number' ? usage.prompt_tokens : 0 |
| 28 | + const outputTokens = typeof usage.completion_tokens === 'number' ? usage.completion_tokens : 0 |
| 29 | + const promptDetails = usage.prompt_tokens_details as Record<string, unknown> | undefined |
| 30 | + const cachedTokens = typeof promptDetails?.cached_tokens === 'number' ? promptDetails.cached_tokens : 0 |
| 31 | + const nonCachedInput = Math.max(0, inputTokens - cachedTokens) |
| 32 | + |
| 33 | + const inputCost = nonCachedInput * FIREWORKS_INPUT_COST_PER_TOKEN |
| 34 | + const cachedCost = cachedTokens * FIREWORKS_CACHED_INPUT_COST_PER_TOKEN |
| 35 | + const outputCost = outputTokens * FIREWORKS_OUTPUT_COST_PER_TOKEN |
| 36 | + const totalCost = inputCost + cachedCost + outputCost |
| 37 | + |
| 38 | + const breakdown = [ |
| 39 | + `${nonCachedInput} input × $0.30/M = $${inputCost.toFixed(8)}`, |
| 40 | + `${cachedTokens} cached × $0.03/M = $${cachedCost.toFixed(8)}`, |
| 41 | + `${outputTokens} output × $1.20/M = $${outputCost.toFixed(8)}`, |
| 42 | + `Total: $${totalCost.toFixed(8)}`, |
| 43 | + ].join('\n ') |
| 44 | + |
| 45 | + return { cost: totalCost, breakdown } |
| 46 | +} |
| 47 | + |
| 48 | +const testPrompt = 'Say "hello world" and nothing else.' |
| 49 | + |
| 50 | +// ─── Direct Fireworks API Test ────────────────────────────────────────────── |
| 51 | + |
| 52 | +async function testFireworksDirect() { |
| 53 | + const apiKey = process.env.FIREWORKS_API_KEY |
| 54 | + if (!apiKey) { |
| 55 | + console.error('❌ FIREWORKS_API_KEY is not set. Add it to .env.local or pass it directly.') |
| 56 | + process.exit(1) |
| 57 | + } |
| 58 | + |
| 59 | + console.log('── Test 1: Fireworks API (non-streaming) ──') |
| 60 | + console.log(`Model: ${FIREWORKS_MODEL}`) |
| 61 | + console.log(`Prompt: "${testPrompt}"`) |
| 62 | + console.log() |
| 63 | + |
| 64 | + const startTime = Date.now() |
| 65 | + const response = await fetch(`${FIREWORKS_BASE_URL}/chat/completions`, { |
| 66 | + method: 'POST', |
| 67 | + headers: { |
| 68 | + Authorization: `Bearer ${apiKey}`, |
| 69 | + 'Content-Type': 'application/json', |
| 70 | + }, |
| 71 | + body: JSON.stringify({ |
| 72 | + model: FIREWORKS_MODEL, |
| 73 | + messages: [{ role: 'user', content: testPrompt }], |
| 74 | + max_tokens: 64, |
| 75 | + }), |
| 76 | + }) |
| 77 | + |
| 78 | + if (!response.ok) { |
| 79 | + const errorText = await response.text() |
| 80 | + console.error(`❌ Fireworks API returned ${response.status}: ${errorText}`) |
| 81 | + process.exit(1) |
| 82 | + } |
| 83 | + |
| 84 | + const data = await response.json() |
| 85 | + const elapsed = Date.now() - startTime |
| 86 | + const content = data.choices?.[0]?.message?.content ?? '<no content>' |
| 87 | + const usage = data.usage ?? {} |
| 88 | + |
| 89 | + const { cost, breakdown } = computeCost(usage) |
| 90 | + console.log(`✅ Response (${elapsed}ms):`) |
| 91 | + console.log(` Content: ${content}`) |
| 92 | + console.log(` Model: ${data.model}`) |
| 93 | + console.log(` Usage: ${JSON.stringify(usage)}`) |
| 94 | + console.log(` Computed cost: $${cost.toFixed(8)}`) |
| 95 | + console.log(` ${breakdown}`) |
| 96 | + console.log() |
| 97 | + |
| 98 | + // Streaming test |
| 99 | + console.log('── Test 1b: Fireworks API (streaming) ──') |
| 100 | + const streamStart = Date.now() |
| 101 | + const streamResponse = await fetch(`${FIREWORKS_BASE_URL}/chat/completions`, { |
| 102 | + method: 'POST', |
| 103 | + headers: { |
| 104 | + Authorization: `Bearer ${apiKey}`, |
| 105 | + 'Content-Type': 'application/json', |
| 106 | + }, |
| 107 | + body: JSON.stringify({ |
| 108 | + model: FIREWORKS_MODEL, |
| 109 | + messages: [{ role: 'user', content: testPrompt }], |
| 110 | + max_tokens: 64, |
| 111 | + stream: true, |
| 112 | + stream_options: { include_usage: true }, |
| 113 | + }), |
| 114 | + }) |
| 115 | + |
| 116 | + if (!streamResponse.ok) { |
| 117 | + const errorText = await streamResponse.text() |
| 118 | + console.error(`❌ Fireworks streaming API returned ${streamResponse.status}: ${errorText}`) |
| 119 | + process.exit(1) |
| 120 | + } |
| 121 | + |
| 122 | + const reader = streamResponse.body?.getReader() |
| 123 | + if (!reader) { |
| 124 | + console.error('❌ No response body reader') |
| 125 | + process.exit(1) |
| 126 | + } |
| 127 | + |
| 128 | + const decoder = new TextDecoder() |
| 129 | + let streamContent = '' |
| 130 | + let streamUsage: Record<string, unknown> | null = null |
| 131 | + let chunkCount = 0 |
| 132 | + |
| 133 | + let done = false |
| 134 | + while (!done) { |
| 135 | + const result = await reader.read() |
| 136 | + done = result.done |
| 137 | + if (done) break |
| 138 | + |
| 139 | + const text = decoder.decode(result.value, { stream: true }) |
| 140 | + const lines = text.split('\n').filter((l) => l.startsWith('data: ')) |
| 141 | + |
| 142 | + for (const line of lines) { |
| 143 | + const raw = line.slice('data: '.length) |
| 144 | + if (raw === '[DONE]') continue |
| 145 | + |
| 146 | + try { |
| 147 | + const chunk = JSON.parse(raw) |
| 148 | + chunkCount++ |
| 149 | + const delta = chunk.choices?.[0]?.delta |
| 150 | + if (delta?.content) streamContent += delta.content |
| 151 | + if (delta?.reasoning_content) { |
| 152 | + console.log(` [reasoning chunk] ${delta.reasoning_content.slice(0, 80)}...`) |
| 153 | + } |
| 154 | + if (chunk.usage) streamUsage = chunk.usage |
| 155 | + } catch { |
| 156 | + // skip non-JSON lines |
| 157 | + } |
| 158 | + } |
| 159 | + } |
| 160 | + |
| 161 | + const streamElapsed = Date.now() - streamStart |
| 162 | + console.log(`✅ Stream response (${streamElapsed}ms, ${chunkCount} chunks):`) |
| 163 | + console.log(` Content: ${streamContent}`) |
| 164 | + if (streamUsage) { |
| 165 | + const { cost: streamCost, breakdown: streamBreakdown } = computeCost(streamUsage as Record<string, unknown>) |
| 166 | + console.log(` Usage: ${JSON.stringify(streamUsage)}`) |
| 167 | + console.log(` Computed cost: $${streamCost.toFixed(8)}`) |
| 168 | + console.log(` ${streamBreakdown}`) |
| 169 | + } |
| 170 | + console.log() |
| 171 | +} |
| 172 | + |
| 173 | +// ─── Chat Completions Endpoint Test ───────────────────────────────────────── |
| 174 | + |
| 175 | +async function testChatCompletionsEndpoint() { |
| 176 | + const codebuffApiKey = process.env.CODEBUFF_API_KEY |
| 177 | + if (!codebuffApiKey) { |
| 178 | + console.error('❌ CODEBUFF_API_KEY is not set. Pass it as an env var.') |
| 179 | + console.error(' Example: CODEBUFF_API_KEY=<key> bun scripts/test-fireworks.ts endpoint') |
| 180 | + process.exit(1) |
| 181 | + } |
| 182 | + |
| 183 | + const appUrl = process.env.NEXT_PUBLIC_CODEBUFF_APP_URL ?? 'http://localhost:3000' |
| 184 | + const endpoint = `${appUrl}/api/v1/chat/completions` |
| 185 | + |
| 186 | + console.log('── Test 2: Chat Completions Endpoint (non-streaming) ──') |
| 187 | + console.log(`Endpoint: ${endpoint}`) |
| 188 | + console.log(`Model: ${OPENROUTER_MODEL} (should route to Fireworks)`) |
| 189 | + console.log(`Prompt: "${testPrompt}"`) |
| 190 | + console.log() |
| 191 | + |
| 192 | + // We need a valid run_id. This is tricky without a full setup, |
| 193 | + // so we'll just fire the request and check the error to confirm routing. |
| 194 | + // If you have a valid run_id, set it via RUN_ID env var. |
| 195 | + const runId = process.env.RUN_ID ?? 'test-run-id-fireworks' |
| 196 | + |
| 197 | + const startTime = Date.now() |
| 198 | + const response = await fetch(endpoint, { |
| 199 | + method: 'POST', |
| 200 | + headers: { |
| 201 | + Authorization: `Bearer ${codebuffApiKey}`, |
| 202 | + 'Content-Type': 'application/json', |
| 203 | + }, |
| 204 | + body: JSON.stringify({ |
| 205 | + model: OPENROUTER_MODEL, |
| 206 | + messages: [{ role: 'user', content: testPrompt }], |
| 207 | + max_tokens: 64, |
| 208 | + stream: false, |
| 209 | + codebuff_metadata: { |
| 210 | + run_id: runId, |
| 211 | + client_id: 'test-fireworks-script', |
| 212 | + cost_mode: 'free', |
| 213 | + }, |
| 214 | + }), |
| 215 | + }) |
| 216 | + |
| 217 | + const elapsed = Date.now() - startTime |
| 218 | + const data = await response.json() |
| 219 | + |
| 220 | + if (response.ok) { |
| 221 | + const content = data.choices?.[0]?.message?.content ?? '<no content>' |
| 222 | + console.log(`✅ Response (${elapsed}ms):`) |
| 223 | + console.log(` Content: ${content}`) |
| 224 | + console.log(` Model: ${data.model}`) |
| 225 | + console.log(` Provider: ${data.provider}`) |
| 226 | + console.log(` Usage: ${JSON.stringify(data.usage)}`) |
| 227 | + } else { |
| 228 | + // Even an auth/validation error confirms the endpoint is reachable |
| 229 | + console.log(`⚠️ Response ${response.status} (${elapsed}ms):`) |
| 230 | + console.log(` ${JSON.stringify(data)}`) |
| 231 | + if (response.status === 400 && data.message?.includes('runId')) { |
| 232 | + console.log(' ℹ️ This is expected if you don\'t have a valid run_id.') |
| 233 | + console.log(' ℹ️ The request reached the endpoint successfully — routing is wired up.') |
| 234 | + } else if (response.status === 401) { |
| 235 | + console.log(' ℹ️ Auth failed. Make sure CODEBUFF_API_KEY is valid.') |
| 236 | + } |
| 237 | + } |
| 238 | + console.log() |
| 239 | + |
| 240 | + // Streaming test |
| 241 | + console.log('── Test 2b: Chat Completions Endpoint (streaming) ──') |
| 242 | + const streamStart = Date.now() |
| 243 | + const streamResponse = await fetch(endpoint, { |
| 244 | + method: 'POST', |
| 245 | + headers: { |
| 246 | + Authorization: `Bearer ${codebuffApiKey}`, |
| 247 | + 'Content-Type': 'application/json', |
| 248 | + }, |
| 249 | + body: JSON.stringify({ |
| 250 | + model: OPENROUTER_MODEL, |
| 251 | + messages: [{ role: 'user', content: testPrompt }], |
| 252 | + max_tokens: 64, |
| 253 | + stream: true, |
| 254 | + codebuff_metadata: { |
| 255 | + run_id: runId, |
| 256 | + client_id: 'test-fireworks-script', |
| 257 | + cost_mode: 'free', |
| 258 | + }, |
| 259 | + }), |
| 260 | + }) |
| 261 | + |
| 262 | + const streamElapsed = Date.now() - streamStart |
| 263 | + |
| 264 | + if (streamResponse.ok) { |
| 265 | + const reader = streamResponse.body?.getReader() |
| 266 | + if (!reader) { |
| 267 | + console.error('❌ No response body reader') |
| 268 | + process.exit(1) |
| 269 | + } |
| 270 | + |
| 271 | + const decoder = new TextDecoder() |
| 272 | + let streamContent = '' |
| 273 | + let chunkCount = 0 |
| 274 | + |
| 275 | + let done = false |
| 276 | + while (!done) { |
| 277 | + const result = await reader.read() |
| 278 | + done = result.done |
| 279 | + if (done) break |
| 280 | + |
| 281 | + const text = decoder.decode(result.value, { stream: true }) |
| 282 | + const lines = text.split('\n').filter((l) => l.startsWith('data: ')) |
| 283 | + |
| 284 | + for (const line of lines) { |
| 285 | + const raw = line.slice('data: '.length) |
| 286 | + if (raw === '[DONE]') continue |
| 287 | + |
| 288 | + try { |
| 289 | + const chunk = JSON.parse(raw) |
| 290 | + chunkCount++ |
| 291 | + const delta = chunk.choices?.[0]?.delta |
| 292 | + if (delta?.content) streamContent += delta.content |
| 293 | + } catch { |
| 294 | + // skip non-JSON lines |
| 295 | + } |
| 296 | + } |
| 297 | + } |
| 298 | + |
| 299 | + console.log(`✅ Stream response (${streamElapsed}ms, ${chunkCount} chunks):`) |
| 300 | + console.log(` Content: ${streamContent}`) |
| 301 | + } else { |
| 302 | + const data = await streamResponse.json() |
| 303 | + console.log(`⚠️ Response ${streamResponse.status} (${streamElapsed}ms):`) |
| 304 | + console.log(` ${JSON.stringify(data)}`) |
| 305 | + if (streamResponse.status === 400 && data.message?.includes('runId')) { |
| 306 | + console.log(' ℹ️ Expected without a valid run_id. Endpoint is reachable and routing works.') |
| 307 | + } |
| 308 | + } |
| 309 | + console.log() |
| 310 | +} |
| 311 | + |
| 312 | +// ─── Main ─────────────────────────────────────────────────────────────────── |
| 313 | + |
| 314 | +async function main() { |
| 315 | + const mode = process.argv[2] ?? 'direct' |
| 316 | + |
| 317 | + console.log('🔥 Fireworks Integration Test') |
| 318 | + console.log('='.repeat(50)) |
| 319 | + console.log() |
| 320 | + |
| 321 | + switch (mode) { |
| 322 | + case 'direct': |
| 323 | + await testFireworksDirect() |
| 324 | + break |
| 325 | + case 'endpoint': |
| 326 | + await testChatCompletionsEndpoint() |
| 327 | + break |
| 328 | + case 'both': |
| 329 | + await testFireworksDirect() |
| 330 | + await testChatCompletionsEndpoint() |
| 331 | + break |
| 332 | + default: |
| 333 | + console.error(`Unknown mode: ${mode}`) |
| 334 | + console.error('Usage: bun scripts/test-fireworks.ts [direct|endpoint|both]') |
| 335 | + process.exit(1) |
| 336 | + } |
| 337 | + |
| 338 | + console.log('Done!') |
| 339 | +} |
| 340 | + |
| 341 | +main() |
0 commit comments