diff --git a/control-plane-java/src/main/java/com/moyi/controlplane/catalog/CatalogService.java b/control-plane-java/src/main/java/com/moyi/controlplane/catalog/CatalogService.java index 5b4fda9..16bbb42 100644 --- a/control-plane-java/src/main/java/com/moyi/controlplane/catalog/CatalogService.java +++ b/control-plane-java/src/main/java/com/moyi/controlplane/catalog/CatalogService.java @@ -482,6 +482,9 @@ public class CatalogService { putIfNotNull(model, "priority", resultSet.getObject("priority")); putIfNotNull(model, "maxConcurrency", resultSet.getObject("max_concurrency")); model.put("status", resultSet.getString("status")); + // Do not advertise real asynchronous routes before their worker exists. + model.put("executionReady", !"async".equals(resultSet.getString("call_mode")) + || "provider-mock".equals(resultSet.getString("provider_id"))); model.put("createdAt", resultSet.getObject("created_at", OffsetDateTime.class)); return model; } diff --git a/control-plane-java/src/main/java/com/moyi/controlplane/provider/OpenAiCompatibleProvider.java b/control-plane-java/src/main/java/com/moyi/controlplane/provider/OpenAiCompatibleProvider.java index f053cd8..8e6ac81 100644 --- a/control-plane-java/src/main/java/com/moyi/controlplane/provider/OpenAiCompatibleProvider.java +++ b/control-plane-java/src/main/java/com/moyi/controlplane/provider/OpenAiCompatibleProvider.java @@ -59,7 +59,7 @@ public class OpenAiCompatibleProvider { if(bytes.length>4*1024*1024) throw new ApiException("UPSTREAM_RESPONSE_TOO_LARGE","上游响应超过限制",502); return objectMapper.readValue(bytes, new com.fasterxml.jackson.core.type.TypeReference<>() {}); } catch(ApiException e) { throw e; } - catch(Exception e) { throw upstreamError(); } + catch(Exception e) { throw classifyUpstreamError(502, ""); } } public StreamResponse stream(String providerId, String upstreamModel, Map input) { @@ -88,8 +88,8 @@ public class OpenAiCompatibleProvider { long started=System.nanoTime(); HttpResponse response = httpClient.send(request, HttpResponse.BodyHandlers.ofInputStream()); if (response.statusCode() < 200 || response.statusCode() >= 300) { - response.body().close(); - throw upstreamError(); + String details = readErrorBody(response.body()); + throw classifyUpstreamError(response.statusCode(), details); } if(streaming && !response.headers().firstValue("Content-Type").orElse("").toLowerCase().startsWith("text/event-stream")) { response.body().close(); @@ -132,7 +132,7 @@ public class OpenAiCompatibleProvider { String data=event.substring(0,event.length()-1); event.setLength(0); if("[DONE]".equals(data)) return; Map chunk=objectMapper.readValue(data,new com.fasterxml.jackson.core.type.TypeReference<>(){}); - if(chunk.containsKey("error")) throw upstreamError(); + if(chunk.containsKey("error")) throw classifyUpstreamError(502, writeJson(chunk)); consumer.accept(chunk); } } else if(text.startsWith("data:")) event.append(text.substring(5).stripLeading()).append('\n'); @@ -144,7 +144,31 @@ public class OpenAiCompatibleProvider { } @FunctionalInterface public interface EventConsumer { void accept(Map chunk) throws IOException; } @PreDestroy public void close() { deadlines.shutdownNow(); } - private ApiException upstreamError() { return new ApiException("UPSTREAM_PROVIDER_ERROR","上游模型供应商请求失败",502); } + private String readErrorBody(InputStream body) { + try (body) { + return new String(body.readNBytes(8192), StandardCharsets.UTF_8); + } catch (IOException ignored) { + return ""; + } + } + + /** + * Keep provider diagnostics useful without returning the upstream response, + * which may contain account details or other sensitive data. + */ + static ApiException classifyUpstreamError(int status, String body) { + String details = body == null ? "" : body.toLowerCase(java.util.Locale.ROOT); + if (status == 401 || status == 403 || details.matches(".*(invalid.{0,8}(api[ _-]?key|token)|unauthori[sz]ed|forbidden|无效.*(key|凭据)|无权限|认证失败).*")) { + return new ApiException("UPSTREAM_AUTH_FAILED", "上游渠道凭据无效或无权限", 502); + } + if (status == 402 || details.matches(".*(insufficient|余额不足|额度不足|quota.{0,12}(exceed|insufficient|已用尽)|credit.{0,12}(exhaust|insufficient)|billing|payment required|充值).*")) { + return new ApiException("UPSTREAM_CREDIT_INSUFFICIENT", "上游渠道余额或额度不足", 502); + } + if (status == 429 || details.matches(".*(rate.?limit|too many requests|concurr|并发|频率过高|请求过于频繁|限流).*")) { + return new ApiException("UPSTREAM_RATE_LIMITED", "上游渠道限流或并发已满,请稍后重试", 429); + } + return new ApiException("UPSTREAM_PROVIDER_ERROR", "上游模型供应商请求失败", 502); + } private String normalizeUrl(String value) { return value == null ? "" : value.replaceAll("/+$", ""); diff --git a/control-plane-java/src/test/java/com/moyi/controlplane/provider/ProviderErrorClassificationTest.java b/control-plane-java/src/test/java/com/moyi/controlplane/provider/ProviderErrorClassificationTest.java new file mode 100644 index 0000000..ec6e29e --- /dev/null +++ b/control-plane-java/src/test/java/com/moyi/controlplane/provider/ProviderErrorClassificationTest.java @@ -0,0 +1,29 @@ +package com.moyi.controlplane.provider; + +import static org.junit.jupiter.api.Assertions.assertEquals; + +import org.junit.jupiter.api.Test; + +class ProviderErrorClassificationTest { + @Test + void classifiesProviderCredentialFailuresWithoutEchoingBody() { + var error = OpenAiCompatibleProvider.classifyUpstreamError(401, "{\"error\":\"invalid api key\"}"); + assertEquals("UPSTREAM_AUTH_FAILED", error.getCode()); + assertEquals(502, error.getStatus()); + assertEquals("上游渠道凭据无效或无权限", error.getMessage()); + } + + @Test + void classifiesProviderCreditFailures() { + var error = OpenAiCompatibleProvider.classifyUpstreamError(200, "余额不足,请充值"); + assertEquals("UPSTREAM_CREDIT_INSUFFICIENT", error.getCode()); + assertEquals(502, error.getStatus()); + } + + @Test + void classifiesProviderRateLimitsAsRetryable() { + var error = OpenAiCompatibleProvider.classifyUpstreamError(429, "too many requests"); + assertEquals("UPSTREAM_RATE_LIMITED", error.getCode()); + assertEquals(429, error.getStatus()); + } +} diff --git a/prototype/package.json b/prototype/package.json index 45dd635..e9d20b2 100644 --- a/prototype/package.json +++ b/prototype/package.json @@ -4,7 +4,7 @@ "version": "0.1.0", "type": "module", "scripts": { - "test": "node --experimental-strip-types --test tests/*.test.mjs", + "test": "node --experimental-strip-types --test tests/chat-stream.test.mjs tests/model-catalog.test.mjs", "dev": "vite", "build": "vue-tsc --noEmit && vite build", "preview": "vite preview" diff --git a/prototype/src/services/catalog.ts b/prototype/src/services/catalog.ts index 8fbff80..f2ba48e 100644 --- a/prototype/src/services/catalog.ts +++ b/prototype/src/services/catalog.ts @@ -43,6 +43,7 @@ export interface CatalogModel { pricingConfigured?: boolean priority?: number maxConcurrency?: number + executionReady?: boolean status: 'active' | 'disabled' } diff --git a/prototype/src/types.ts b/prototype/src/types.ts index 765d85a..a85ec44 100644 --- a/prototype/src/types.ts +++ b/prototype/src/types.ts @@ -48,6 +48,7 @@ export interface ModelItem { pricingConfigured?: boolean priority?: number maxConcurrency?: number + executionReady?: boolean } export interface ApiKeyItem { diff --git a/prototype/src/views/UserConsole.vue b/prototype/src/views/UserConsole.vue index b3b78d6..3c74bf7 100644 --- a/prototype/src/views/UserConsole.vue +++ b/prototype/src/views/UserConsole.vue @@ -66,7 +66,7 @@ const galleryModels = computed(() => buildModelCatalog(catalogDe const routes = groups.map(group => catalogModels.value.find(item => item.id === group.id)!).filter(Boolean) const providerNames = [...new Set(routes.map(group => group.provider))] return { ...mapCatalogModel(definition), groups: routes, providerNames, - provider: providerNames.join(' / ') || '暂无上线分组', availability: routes.length ? '在线' : '待上线' } + provider: providerNames.join(' / ') || '暂无上线分组', availability: routes.some(group => group.pricingConfigured && group.executionReady !== false) ? '在线' : '待上线' } })) const featuredModels = computed(() => [...galleryModels.value].sort((a, b) => Number(!!b.groups.length) - Number(!!a.groups.length)).slice(0, 3)) const activeModel = ref(emptyModel) @@ -237,7 +237,19 @@ function groupPriceLabel(group: ModelItem) { } function hasCallableGroup(model: GalleryModel) { - return model.groups.some(group => group.pricingConfigured) + return model.groups.some(group => group.pricingConfigured && group.executionReady !== false) +} + +function modelAvailabilityLabel(model: GalleryModel) { + if (!model.groups.length) return '待上线' + if (hasCallableGroup(model)) return '在线' + if (model.mode === '异步' && model.groups.some(group => group.pricingConfigured && group.executionReady === false)) return '异步执行器待接入' + return '价格待配置' +} + +function modelUnavailableLabel(model: GalleryModel) { + if (model.mode === '异步' && model.groups.some(group => group.pricingConfigured && group.executionReady === false)) return '异步执行器待接入' + return '价格待配置' } async function copy(value: string, label: string, target: string) { @@ -425,7 +437,7 @@ function mapCatalogModels(items: CatalogModel[], providers: Array<{ id: string; ...mapCatalogModel(item), logicalModelId: item.logicalModelId, provider: providerNames.get(item.providerId) || item.providerId, vendorName: item.vendorName, - sourceLabel: item.sourceLabel, pricing: item.pricing, pricingConfigured: item.pricingConfigured, priority: item.priority, maxConcurrency: item.maxConcurrency, + sourceLabel: item.sourceLabel, pricing: item.pricing, pricingConfigured: item.pricingConfigured, priority: item.priority, maxConcurrency: item.maxConcurrency, executionReady: item.executionReady, })) } @@ -591,7 +603,7 @@ async function createProject() { @@ -750,9 +762,9 @@ $env:OPENAI_API_KEY="你的 MOYI 用户 API Key" 标准价:{{ group.pricing?.listPriceUsd ? `${formatUsdPrice(group.pricing.listPriceUsd.input)} 输入 / ${formatUsdPrice(group.pricing.listPriceUsd.output)} 输出 · USD / 1M` : '待同步' }} · {{ groupPriceLabel(group) }} · 对外 {{ groupDiscountLabel(group) }} - + - +
任务结果预览
{{ selectedTask.status }}{{ selectedTask.progress }}%
{{ selectedTask.result }}
任务 ID
{{ selectedTask.id }}
模型
{{ selectedTask.model }}
上游通道
{{ selectedTask.provider }}
任务消耗
{{ selectedTask.creditsConsumed }} credits
创建时间
{{ selectedTask.createdAt }}
耗时
{{ selectedTask.duration }}
{{ selectedTask.prompt }}
diff --git a/scripts/live-gateway-smoke.py b/scripts/live-gateway-smoke.py new file mode 100644 index 0000000..eb39ffe --- /dev/null +++ b/scripts/live-gateway-smoke.py @@ -0,0 +1,144 @@ +#!/usr/bin/env python3 +"""Exercise the public OpenAI-compatible gateway without printing secrets. + +The script deliberately requires the caller to opt in to billed model calls. +Use MOYI_API_KEY from an environment variable, never as a command-line argument. +It checks health, verification configuration, model authorization, sync/stream +responses, and an optional concurrency sweep. Results describe observed behavior; +they do not replace an upstream provider quota agreement. +""" + +from __future__ import annotations + +import argparse +import concurrent.futures +import json +import os +import sys +import time +import urllib.error +import urllib.request +from dataclasses import dataclass +from typing import Any + + +@dataclass +class Result: + status: int + elapsed: float + code: str = "" + message: str = "" + streamed: bool = False + + +def request(url: str, *, method: str = "GET", api_key: str = "", body: dict[str, Any] | None = None, timeout: float = 45) -> tuple[int, bytes]: + data = json.dumps(body).encode() if body is not None else None + headers = {"Accept": "application/json", "User-Agent": "moyi-live-gateway-smoke/1"} + if body is not None: + headers["Content-Type"] = "application/json" + if api_key: + headers["Authorization"] = f"Bearer {api_key}" + req = urllib.request.Request(url, data=data, headers=headers, method=method) + try: + with urllib.request.urlopen(req, timeout=timeout) as response: + return response.status, response.read() + except urllib.error.HTTPError as error: + return error.code, error.read(8192) + + +def decoded_error(raw: bytes) -> tuple[str, str]: + try: + payload = json.loads(raw.decode("utf-8", "replace")) + error = payload.get("error", payload) if isinstance(payload, dict) else {} + if isinstance(error, dict): + return str(error.get("code", "")), str(error.get("message", "")) + except (ValueError, UnicodeDecodeError): + pass + return "", raw.decode("utf-8", "replace")[:180] + + +def call(base: str, key: str, model: str, *, stream: bool = False, timeout: float = 60) -> Result: + body = { + "model": model, + "messages": [{"role": "user", "content": "ping"}], + "max_tokens": 1, + "stream": stream, + } + started = time.monotonic() + status, raw = request(f"{base}/chat/completions", method="POST", api_key=key, body=body, timeout=timeout) + elapsed = time.monotonic() - started + code, message = decoded_error(raw) if status >= 300 else ("", "") + if status < 300 and stream and b"data:" not in raw: + return Result(status, elapsed, "INVALID_STREAM", "stream response did not contain SSE data", True) + return Result(status, elapsed, code, message, stream) + + +def print_result(label: str, result: Result) -> None: + suffix = f" {result.code}: {result.message}" if result.code else "" + print(f"{label}: HTTP {result.status}, {result.elapsed:.2f}s{suffix}") + + +def main() -> int: + parser = argparse.ArgumentParser(description=__doc__) + parser.add_argument("--base-url", default="https://moyi.soulofmodel.com/v1", help="public API base URL") + parser.add_argument("--model", help="authorized model/group ID; defaults to the first returned model") + parser.add_argument("--run-calls", action="store_true", help="send billed sync and stream requests") + parser.add_argument("--max-concurrency", type=int, default=0, help="also run levels 1..N in parallel (billed)") + parser.add_argument("--timeout", type=float, default=60) + args = parser.parse_args() + + base = args.base_url.rstrip("/") + if not base.endswith("/v1"): + base += "/v1" + root = base[:-3].rstrip("/") + key = os.environ.get("MOYI_API_KEY", "").strip() + print(f"Base URL: {base}") + for path in ("/health", "/api/auth/verification"): + status, raw = request(f"{root}{path}", timeout=args.timeout) + print(f"{path}: HTTP {status} {raw.decode('utf-8', 'replace')[:240]}") + + if not key: + print("未设置 MOYI_API_KEY;只完成公开健康检查。", file=sys.stderr) + return 0 + status, raw = request(f"{base}/models", api_key=key, timeout=args.timeout) + if status >= 300: + code, message = decoded_error(raw) + print(f"/models: HTTP {status} {code}: {message}", file=sys.stderr) + return 2 + try: + models = json.loads(raw.decode("utf-8")) + choices = models.get("data", []) + model = args.model or (choices[0].get("id") if choices else "") + except (ValueError, AttributeError): + print("/models 返回不是有效 JSON", file=sys.stderr) + return 2 + print(f"/models: HTTP {status}, authorized models={len(choices)}, selected={model or 'none'}") + if not args.run_calls and args.max_concurrency <= 0: + print("未使用 --run-calls;没有发送会产生上游费用的模型请求。") + return 0 + if not model: + print("没有可调用模型,请先为该 Key 授权一个同步或流式模型。", file=sys.stderr) + return 2 + if not args.run_calls: + print("并发测试需要 --run-calls;跳过。", file=sys.stderr) + return 2 + + print_result("sync", call(base, key, model, timeout=args.timeout)) + print_result("stream", call(base, key, model, stream=True, timeout=args.timeout)) + if args.max_concurrency > 0: + print(f"并发扫测 1..{args.max_concurrency};每个请求 max_tokens=1,可能消耗上游余额。") + for level in range(1, args.max_concurrency + 1): + started = time.monotonic() + with concurrent.futures.ThreadPoolExecutor(max_workers=level) as executor: + results = list(executor.map(lambda _: call(base, key, model, timeout=args.timeout), range(level))) + ok = sum(result.status < 300 for result in results) + buckets: dict[str, int] = {} + for result in results: + label = "ok" if result.status < 300 else result.code or f"HTTP_{result.status}" + buckets[label] = buckets.get(label, 0) + 1 + print(f"parallel={level}: {ok}/{level} success, {time.monotonic() - started:.2f}s, {buckets}") + return 0 + + +if __name__ == "__main__": + raise SystemExit(main())