feat(ai): add Wafer Pass and Wafer Serverless providers
Wafer (https://wafer.ai) exposes a single OpenAI-compatible endpoint (`https://pass.wafer.ai/v1`) for two SKUs whose entitlement differs server-side, so we model them as two parallel providers — mirroring the firepass/fireworks split so a user with both subscriptions can switch without re-pasting: - `wafer-pass` — flat-rate. `/v1/models` is filtered to entries whose `wafer.tier === "pass_included"`. - `wafer-serverless` — pay-as-you-go superset of Pass. Both issue `wfr_…` keys. `/login wafer-pass` and `/login wafer-serverless` paste-and-validate via `/v1/models`. `WAFER_PASS_API_KEY` and `WAFER_SERVERLESS_API_KEY` are wired through `getEnvApiKey`. Bundled catalog: - `wafer-pass`: GLM-5.1, Qwen3.5-397B-A17B. - `wafer-serverless`: GLM-5.1, Qwen3.5-397B-A17B, Kimi-K2.6, Qwen3.6-35B-A3B. Dynamic discovery via `/v1/models` overlays additional models at runtime and folds the `wafer` envelope (tier, capabilities, cents/M pricing) into the canonical `Model<"openai-completions">` shape. GLM-family entries carry the zai-style thinking compat (`thinkingFormat: "zai"`, `reasoningContentField: "reasoning_content"`) so reasoning tokens land in the right field. Cents-per-million → dollars-per-million via /100. Tests (`packages/ai/test/wafer.test.ts`, 5 cases): bundled catalog contract for both providers and wire-id pass-through (case-sensitive, no rewrite — `GLM-5.1` must round-trip verbatim or upstream 404s). Optional `packages/ai/test/wafer.live.ts` exercises a real round-trip against `pass.wafer.ai` when `WAFER_PASS_API_KEY` is set.
This commit is contained in:
bench-local
@@ -251,13 +251,13 @@ Auth tags below: `oauth` signs in with your provider account, `plan` routes thro
|
||||
|
||||
Direct APIs and gateways. Mix providers per role.
|
||||
|
||||
Anthropic `oauth` · OpenAI · OpenAI Codex `oauth` · Google Gemini · Google Antigravity `oauth` · xAI · Mistral · Groq · Cerebras · Fireworks · Together · Hugging Face · NVIDIA · OpenRouter · Synthetic · Vercel AI Gateway · Cloudflare AI Gateway · Perplexity `oauth`
|
||||
Anthropic `oauth` · OpenAI · OpenAI Codex `oauth` · Google Gemini · Google Antigravity `oauth` · xAI · Mistral · Groq · Cerebras · Fireworks · Together · Hugging Face · NVIDIA · OpenRouter · Synthetic · Vercel AI Gateway · Cloudflare AI Gateway · Wafer Serverless · Perplexity `oauth`
|
||||
|
||||
### Coding plans
|
||||
|
||||
Subscription-routed. `/login` attaches the session.
|
||||
|
||||
Cursor `oauth` · GitHub Copilot `oauth` · GitLab Duo · Kimi Code `plan` · Moonshot · MiniMax Coding Plan `plan` · MiniMax Coding Plan CN `plan` · Alibaba Coding Plan `plan` · Qwen Portal · Z.AI / GLM Coding Plan `plan` · Xiaomi MiMo · Qianfan · NanoGPT · Venice · Kilo · ZenMux · OpenCode Go · OpenCode Zen
|
||||
Cursor `oauth` · GitHub Copilot `oauth` · GitLab Duo · Kimi Code `plan` · Moonshot · MiniMax Coding Plan `plan` · MiniMax Coding Plan CN `plan` · Alibaba Coding Plan `plan` · Qwen Portal · Z.AI / GLM Coding Plan `plan` · Xiaomi MiMo · Qianfan · NanoGPT · Venice · Kilo · ZenMux · Wafer Pass `plan` · OpenCode Go · OpenCode Zen
|
||||
|
||||
### Run it yourself
|
||||
|
||||
|
||||
@@ -74,6 +74,8 @@ These are consumed via `getEnvApiKey()` (`packages/ai/src/stream.ts`) unless not
|
||||
| `DEEPSEEK_API_KEY` | DeepSeek auth | Using DeepSeek models | |
|
||||
| `KILO_API_KEY` | Kilo auth | Using Kilo models | |
|
||||
| `OLLAMA_CLOUD_API_KEY` | Ollama Cloud auth | Using `ollama-cloud` provider | |
|
||||
| `WAFER_PASS_API_KEY` | Wafer Pass auth | Using `wafer-pass` provider | Flat-rate Wafer subscription; validated against `https://pass.wafer.ai/v1/models` |
|
||||
| `WAFER_SERVERLESS_API_KEY` | Wafer Serverless auth | Using `wafer-serverless` provider | Pay-as-you-go Wafer SKU; validated against `https://pass.wafer.ai/v1/models` |
|
||||
| `GITLAB_TOKEN` | GitLab Duo auth | Using `gitlab-duo` provider | |
|
||||
|
||||
### GitHub/Copilot token chains
|
||||
|
||||
@@ -5,6 +5,7 @@
|
||||
### Added
|
||||
|
||||
- Added `CheckCredentialsOptions.completionProbe` (and `completionTimeoutMs`) so `AuthStorage.checkCredentials` can additionally exercise each credential against the provider's chat-completion endpoint after refresh-on-expiry. Result lands on `CredentialHealthResult.completion` ({ok, reason?, modelId?, latencyMs?}) without disturbing the usage `ok` field. Public types: `CompletionProbe`, `CompletionProbeInput`, `CompletionProbeCredential`, `CredentialCompletionResult`. The probe is invoked even when no `UsageProvider` is registered for the row, and is skipped when OAuth refresh fails (the stale bytes would only mask the upstream failure).
|
||||
- Added Wafer Pass and Wafer Serverless providers (`wafer-pass`, `wafer-serverless`). OpenAI-compatible (`https://pass.wafer.ai/v1`), bearer auth, `wfr_…` keys. `/login wafer-pass` and `/login wafer-serverless` paste-and-validate the key against `/v1/models`. `WAFER_PASS_API_KEY` and `WAFER_SERVERLESS_API_KEY` environment variables wired into `getEnvApiKey`. Bundled catalog seeds `wafer-pass/{GLM-5.1, Qwen3.5-397B-A17B}` and `wafer-serverless/{GLM-5.1, Qwen3.5-397B-A17B, Kimi-K2.6, Qwen3.6-35B-A3B}`; dynamic discovery via `/v1/models` overlays additional models at runtime. Pass-tier discovery filters `wafer.tier === "pass_included"`. GLM-family entries carry the zai-style thinking compat (`thinkingFormat: "zai"`, `reasoningContentField: "reasoning_content"`).
|
||||
|
||||
### Changed
|
||||
|
||||
|
||||
@@ -62,6 +62,8 @@ Unified LLM API with automatic model discovery, provider configuration, token an
|
||||
- **Hugging Face Inference**
|
||||
- **xAI**
|
||||
- **Venice** (requires `VENICE_API_KEY`)
|
||||
- **Wafer Pass** (requires `WAFER_PASS_API_KEY`; flat-rate subscription, includes GLM-5.1 and Qwen3.5-397B-A17B)
|
||||
- **Wafer Serverless** (requires `WAFER_SERVERLESS_API_KEY`; pay-as-you-go)
|
||||
- **OpenRouter**
|
||||
- **Kilo Gateway** (supports OAuth `/login kilo` or `KILO_API_KEY`)
|
||||
- **LiteLLM** (requires `LITELLM_API_KEY`)
|
||||
|
||||
@@ -1619,6 +1619,18 @@ export class AuthStorage {
|
||||
await saveApiKeyCredential(apiKey);
|
||||
return;
|
||||
}
|
||||
case "wafer-pass": {
|
||||
const { loginWaferPass } = await import("./utils/oauth/wafer");
|
||||
const apiKey = await loginWaferPass(ctrl);
|
||||
await saveApiKeyCredential(apiKey);
|
||||
return;
|
||||
}
|
||||
case "wafer-serverless": {
|
||||
const { loginWaferServerless } = await import("./utils/oauth/wafer");
|
||||
const apiKey = await loginWaferServerless(ctrl);
|
||||
await saveApiKeyCredential(apiKey);
|
||||
return;
|
||||
}
|
||||
case "zai": {
|
||||
const { loginZai } = await import("./utils/oauth/zai");
|
||||
const apiKey = await loginZai(ctrl);
|
||||
|
||||
@@ -70656,6 +70656,166 @@
|
||||
}
|
||||
}
|
||||
},
|
||||
"wafer-pass": {
|
||||
"GLM-5.1": {
|
||||
"id": "GLM-5.1",
|
||||
"name": "GLM-5.1",
|
||||
"api": "openai-completions",
|
||||
"provider": "wafer-pass",
|
||||
"baseUrl": "https://pass.wafer.ai/v1",
|
||||
"reasoning": true,
|
||||
"input": [
|
||||
"text"
|
||||
],
|
||||
"cost": {
|
||||
"input": 1.2,
|
||||
"output": 3.6,
|
||||
"cacheRead": 0.12,
|
||||
"cacheWrite": 0
|
||||
},
|
||||
"contextWindow": 202752,
|
||||
"maxTokens": 65536,
|
||||
"compat": {
|
||||
"supportsDeveloperRole": false,
|
||||
"thinkingFormat": "zai",
|
||||
"reasoningContentField": "reasoning_content"
|
||||
},
|
||||
"thinking": {
|
||||
"mode": "effort",
|
||||
"minLevel": "minimal",
|
||||
"maxLevel": "xhigh"
|
||||
}
|
||||
},
|
||||
"Qwen3.5-397B-A17B": {
|
||||
"id": "Qwen3.5-397B-A17B",
|
||||
"name": "Qwen3.5-397B-A17B",
|
||||
"api": "openai-completions",
|
||||
"provider": "wafer-pass",
|
||||
"baseUrl": "https://pass.wafer.ai/v1",
|
||||
"reasoning": false,
|
||||
"input": [
|
||||
"text",
|
||||
"image"
|
||||
],
|
||||
"cost": {
|
||||
"input": 0.48,
|
||||
"output": 2.88,
|
||||
"cacheRead": 0.05,
|
||||
"cacheWrite": 0
|
||||
},
|
||||
"contextWindow": 262144,
|
||||
"maxTokens": 65536,
|
||||
"compat": {
|
||||
"supportsDeveloperRole": false
|
||||
}
|
||||
}
|
||||
},
|
||||
"wafer-serverless": {
|
||||
"GLM-5.1": {
|
||||
"id": "GLM-5.1",
|
||||
"name": "GLM-5.1",
|
||||
"api": "openai-completions",
|
||||
"provider": "wafer-serverless",
|
||||
"baseUrl": "https://pass.wafer.ai/v1",
|
||||
"reasoning": true,
|
||||
"input": [
|
||||
"text"
|
||||
],
|
||||
"cost": {
|
||||
"input": 1.2,
|
||||
"output": 3.6,
|
||||
"cacheRead": 0.12,
|
||||
"cacheWrite": 0
|
||||
},
|
||||
"contextWindow": 202752,
|
||||
"maxTokens": 65536,
|
||||
"compat": {
|
||||
"supportsDeveloperRole": false,
|
||||
"thinkingFormat": "zai",
|
||||
"reasoningContentField": "reasoning_content"
|
||||
},
|
||||
"thinking": {
|
||||
"mode": "effort",
|
||||
"minLevel": "minimal",
|
||||
"maxLevel": "xhigh"
|
||||
}
|
||||
},
|
||||
"Kimi-K2.6": {
|
||||
"id": "Kimi-K2.6",
|
||||
"name": "Kimi-K2.6",
|
||||
"api": "openai-completions",
|
||||
"provider": "wafer-serverless",
|
||||
"baseUrl": "https://pass.wafer.ai/v1",
|
||||
"reasoning": true,
|
||||
"input": [
|
||||
"text",
|
||||
"image"
|
||||
],
|
||||
"cost": {
|
||||
"input": 0,
|
||||
"output": 0,
|
||||
"cacheRead": 0,
|
||||
"cacheWrite": 0
|
||||
},
|
||||
"contextWindow": 262144,
|
||||
"maxTokens": 65536,
|
||||
"compat": {
|
||||
"supportsDeveloperRole": false,
|
||||
"thinkingFormat": "zai",
|
||||
"reasoningContentField": "reasoning_content"
|
||||
},
|
||||
"thinking": {
|
||||
"mode": "effort",
|
||||
"minLevel": "minimal",
|
||||
"maxLevel": "xhigh"
|
||||
}
|
||||
},
|
||||
"Qwen3.5-397B-A17B": {
|
||||
"id": "Qwen3.5-397B-A17B",
|
||||
"name": "Qwen3.5-397B-A17B",
|
||||
"api": "openai-completions",
|
||||
"provider": "wafer-serverless",
|
||||
"baseUrl": "https://pass.wafer.ai/v1",
|
||||
"reasoning": false,
|
||||
"input": [
|
||||
"text",
|
||||
"image"
|
||||
],
|
||||
"cost": {
|
||||
"input": 0.48,
|
||||
"output": 2.88,
|
||||
"cacheRead": 0.05,
|
||||
"cacheWrite": 0
|
||||
},
|
||||
"contextWindow": 262144,
|
||||
"maxTokens": 65536,
|
||||
"compat": {
|
||||
"supportsDeveloperRole": false
|
||||
}
|
||||
},
|
||||
"Qwen3.6-35B-A3B": {
|
||||
"id": "Qwen3.6-35B-A3B",
|
||||
"name": "Qwen3.6-35B-A3B",
|
||||
"api": "openai-completions",
|
||||
"provider": "wafer-serverless",
|
||||
"baseUrl": "https://pass.wafer.ai/v1",
|
||||
"reasoning": false,
|
||||
"input": [
|
||||
"text"
|
||||
],
|
||||
"cost": {
|
||||
"input": 0,
|
||||
"output": 0,
|
||||
"cacheRead": 0,
|
||||
"cacheWrite": 0
|
||||
},
|
||||
"contextWindow": 32768,
|
||||
"maxTokens": 32768,
|
||||
"compat": {
|
||||
"supportsDeveloperRole": false
|
||||
}
|
||||
}
|
||||
},
|
||||
"xai": {
|
||||
"grok-2": {
|
||||
"id": "grok-2",
|
||||
|
||||
@@ -39,6 +39,8 @@ import {
|
||||
veniceModelManagerOptions,
|
||||
vercelAiGatewayModelManagerOptions,
|
||||
vllmModelManagerOptions,
|
||||
waferPassModelManagerOptions,
|
||||
waferServerlessModelManagerOptions,
|
||||
xaiModelManagerOptions,
|
||||
xaiOAuthModelManagerOptions,
|
||||
xiaomiModelManagerOptions,
|
||||
@@ -169,6 +171,18 @@ export const PROVIDER_DESCRIPTORS: readonly ProviderDescriptor[] = [
|
||||
// models.json would permanently drop the model from the catalog with no
|
||||
// automated mechanism to restore it.
|
||||
descriptor("firepass", "kimi-k2.6-turbo", config => firepassModelManagerOptions(config)),
|
||||
catalogDescriptor(
|
||||
"wafer-pass",
|
||||
"GLM-5.1",
|
||||
config => waferPassModelManagerOptions(config),
|
||||
catalog("Wafer Pass", ["WAFER_PASS_API_KEY"], { oauthProvider: "wafer-pass" }),
|
||||
),
|
||||
catalogDescriptor(
|
||||
"wafer-serverless",
|
||||
"GLM-5.1",
|
||||
config => waferServerlessModelManagerOptions(config),
|
||||
catalog("Wafer Serverless", ["WAFER_SERVERLESS_API_KEY"], { oauthProvider: "wafer-serverless" }),
|
||||
),
|
||||
descriptor("xai", "grok-4-fast-non-reasoning", config => xaiModelManagerOptions(config)),
|
||||
catalogDescriptor(
|
||||
"xai-oauth",
|
||||
|
||||
@@ -1037,6 +1037,136 @@ export function firepassModelManagerOptions(
|
||||
};
|
||||
}
|
||||
|
||||
// ---------------------------------------------------------------------------
|
||||
// 7.7 Wafer (Pass + Serverless)
|
||||
// ---------------------------------------------------------------------------
|
||||
|
||||
export interface WaferModelManagerConfig {
|
||||
apiKey?: string;
|
||||
baseUrl?: string;
|
||||
}
|
||||
|
||||
const WAFER_DEFAULT_BASE_URL = "https://pass.wafer.ai/v1";
|
||||
const WAFER_MAX_TOKENS_CAP = 65536;
|
||||
|
||||
/**
|
||||
* Shared mapper for Wafer's `/v1/models` records.
|
||||
*
|
||||
* Wafer wraps each entry with a `wafer` envelope describing tier, capabilities,
|
||||
* and cents-per-million pricing. The mapper folds that metadata into the
|
||||
* canonical `Model<"openai-completions">` shape and applies zai-family thinking
|
||||
* compat when the entry advertises reasoning support (GLM-family on the Pass
|
||||
* SKU). Cents-per-million → dollars-per-million via /100.
|
||||
*/
|
||||
interface WaferRecord {
|
||||
context_length?: unknown;
|
||||
tier?: unknown;
|
||||
capabilities?: { vision?: unknown; reasoning?: unknown; tools?: unknown };
|
||||
pricing?: {
|
||||
input_cents_per_million?: unknown;
|
||||
output_cents_per_million?: unknown;
|
||||
cache_read_cents_per_million?: unknown;
|
||||
};
|
||||
display_name?: unknown;
|
||||
}
|
||||
|
||||
function readWaferRecord(entry: OpenAICompatibleModelRecord): WaferRecord | undefined {
|
||||
const raw = (entry as { wafer?: unknown }).wafer;
|
||||
return raw && typeof raw === "object" ? (raw as WaferRecord) : undefined;
|
||||
}
|
||||
|
||||
function mapWaferModel(
|
||||
providerId: "wafer-pass" | "wafer-serverless",
|
||||
baseUrl: string,
|
||||
entry: OpenAICompatibleModelRecord,
|
||||
defaults: Model<"openai-completions">,
|
||||
): Model<"openai-completions"> {
|
||||
const wafer = readWaferRecord(entry);
|
||||
const capabilities = wafer?.capabilities ?? {};
|
||||
const reasoning = capabilities.reasoning === true;
|
||||
const vision = capabilities.vision === true;
|
||||
const contextWindow = toPositiveNumber(
|
||||
wafer?.context_length,
|
||||
toPositiveNumber((entry as { max_model_len?: unknown }).max_model_len, defaults.contextWindow),
|
||||
);
|
||||
const maxTokens = Math.min(contextWindow, WAFER_MAX_TOKENS_CAP);
|
||||
const pricing = wafer?.pricing ?? {};
|
||||
// Wafer publishes cents-per-million; OMP catalog stores dollars-per-million.
|
||||
const cost = {
|
||||
input: toPositiveNumber(pricing.input_cents_per_million, 0) / 100,
|
||||
output: toPositiveNumber(pricing.output_cents_per_million, 0) / 100,
|
||||
cacheRead: toPositiveNumber(pricing.cache_read_cents_per_million, 0) / 100,
|
||||
cacheWrite: 0,
|
||||
};
|
||||
const name = toModelName(wafer?.display_name, defaults.name);
|
||||
const base: Model<"openai-completions"> = {
|
||||
...defaults,
|
||||
id: defaults.id,
|
||||
name,
|
||||
api: "openai-completions",
|
||||
provider: providerId,
|
||||
baseUrl,
|
||||
reasoning,
|
||||
input: vision ? (["text", "image"] as const) : ["text"],
|
||||
cost,
|
||||
contextWindow,
|
||||
maxTokens,
|
||||
};
|
||||
if (reasoning) {
|
||||
return {
|
||||
...base,
|
||||
compat: {
|
||||
thinkingFormat: "zai",
|
||||
reasoningContentField: "reasoning_content",
|
||||
supportsDeveloperRole: false,
|
||||
},
|
||||
};
|
||||
}
|
||||
return {
|
||||
...base,
|
||||
compat: { supportsDeveloperRole: false },
|
||||
};
|
||||
}
|
||||
|
||||
function createWaferOptions(
|
||||
providerId: "wafer-pass" | "wafer-serverless",
|
||||
config: WaferModelManagerConfig | undefined,
|
||||
): ModelManagerOptions<"openai-completions"> {
|
||||
const apiKey = config?.apiKey;
|
||||
const baseUrl = config?.baseUrl ?? WAFER_DEFAULT_BASE_URL;
|
||||
const passOnly = providerId === "wafer-pass";
|
||||
return {
|
||||
providerId,
|
||||
...(apiKey && {
|
||||
fetchDynamicModels: () =>
|
||||
fetchOpenAICompatibleModels({
|
||||
api: "openai-completions",
|
||||
provider: providerId,
|
||||
baseUrl,
|
||||
apiKey,
|
||||
filterModel: entry => {
|
||||
if (!passOnly) return true;
|
||||
const wafer = readWaferRecord(entry);
|
||||
return wafer?.tier === "pass_included";
|
||||
},
|
||||
mapModel: (entry, defaults) => mapWaferModel(providerId, baseUrl, entry, defaults),
|
||||
}),
|
||||
}),
|
||||
};
|
||||
}
|
||||
|
||||
export function waferPassModelManagerOptions(
|
||||
config?: WaferModelManagerConfig,
|
||||
): ModelManagerOptions<"openai-completions"> {
|
||||
return createWaferOptions("wafer-pass", config);
|
||||
}
|
||||
|
||||
export function waferServerlessModelManagerOptions(
|
||||
config?: WaferModelManagerConfig,
|
||||
): ModelManagerOptions<"openai-completions"> {
|
||||
return createWaferOptions("wafer-serverless", config);
|
||||
}
|
||||
|
||||
// ---------------------------------------------------------------------------
|
||||
// 7. Mistral
|
||||
// ---------------------------------------------------------------------------
|
||||
|
||||
@@ -183,6 +183,8 @@ const serviceProviderMap: Record<string, KeyResolver> = {
|
||||
"xai-oauth": () => $pickenv("XAI_OAUTH_TOKEN", "XAI_API_KEY"),
|
||||
fireworks: "FIREWORKS_API_KEY",
|
||||
firepass: "FIREPASS_API_KEY",
|
||||
"wafer-pass": "WAFER_PASS_API_KEY",
|
||||
"wafer-serverless": "WAFER_SERVERLESS_API_KEY",
|
||||
openrouter: "OPENROUTER_API_KEY",
|
||||
kilo: "KILO_API_KEY",
|
||||
"vercel-ai-gateway": "AI_GATEWAY_API_KEY",
|
||||
|
||||
@@ -142,6 +142,8 @@ export type KnownProvider =
|
||||
| "venice"
|
||||
| "vllm"
|
||||
| "xiaomi"
|
||||
| "wafer-pass"
|
||||
| "wafer-serverless"
|
||||
| "zenmux"
|
||||
| "lm-studio";
|
||||
export type Provider = KnownProvider | string;
|
||||
|
||||
@@ -235,6 +235,16 @@ const builtInOAuthProviders: OAuthProviderInfo[] = [
|
||||
name: "xAI Grok OAuth (SuperGrok Subscription)",
|
||||
available: true,
|
||||
},
|
||||
{
|
||||
id: "wafer-pass",
|
||||
name: "Wafer Pass (flat-rate subscription)",
|
||||
available: true,
|
||||
},
|
||||
{
|
||||
id: "wafer-serverless",
|
||||
name: "Wafer Serverless (pay-as-you-go)",
|
||||
available: true,
|
||||
},
|
||||
];
|
||||
|
||||
const customOAuthProviders = new Map<string, OAuthProviderInterface>();
|
||||
@@ -359,6 +369,8 @@ export async function refreshOAuthToken(
|
||||
case "cloudflare-ai-gateway":
|
||||
case "vercel-ai-gateway":
|
||||
case "qwen-portal":
|
||||
case "wafer-pass":
|
||||
case "wafer-serverless":
|
||||
case "zenmux":
|
||||
case "vllm":
|
||||
// API keys / static bearer tokens don't expire, return as-is
|
||||
|
||||
@@ -48,6 +48,8 @@ export type OAuthProvider =
|
||||
| "together"
|
||||
| "venice"
|
||||
| "vercel-ai-gateway"
|
||||
| "wafer-pass"
|
||||
| "wafer-serverless"
|
||||
| "vllm"
|
||||
| "xai-oauth"
|
||||
| "xiaomi"
|
||||
|
||||
@@ -0,0 +1,50 @@
|
||||
/**
|
||||
* Wafer login flows.
|
||||
*
|
||||
* Wafer (https://wafer.ai) exposes a single OpenAI-compatible base URL
|
||||
* (`https://pass.wafer.ai/v1`) for two SKUs:
|
||||
*
|
||||
* - **Wafer Pass** — flat-rate subscription. The key authorizes models whose
|
||||
* catalog entries carry `wafer.tier = "pass_included"`.
|
||||
* - **Wafer Serverless** — pay-as-you-go. Superset of Pass; the same `/v1/models`
|
||||
* endpoint returns the full per-account model list.
|
||||
*
|
||||
* Both SKUs issue `wfr_…` keys. The key prefix alone does not distinguish
|
||||
* tiers — the entitlement is per-account on the server side — so we expose
|
||||
* two parallel logins / env vars (`WAFER_PASS_API_KEY`, `WAFER_SERVERLESS_API_KEY`)
|
||||
* mirroring the firepass/fireworks split, letting users with both
|
||||
* subscriptions switch between them without re-pasting.
|
||||
*
|
||||
* Validation uses the shared `/v1/models` endpoint, which works for both
|
||||
* tiers and is cheap (no token spend).
|
||||
*/
|
||||
import { createApiKeyLogin } from "./api-key-login";
|
||||
|
||||
const WAFER_AUTH_URL = "https://wafer.ai/dashboard";
|
||||
const WAFER_MODELS_URL = "https://pass.wafer.ai/v1/models";
|
||||
|
||||
export const loginWaferPass = createApiKeyLogin({
|
||||
providerLabel: "Wafer Pass",
|
||||
authUrl: WAFER_AUTH_URL,
|
||||
instructions: "Create or copy your Wafer Pass API key from the Wafer dashboard",
|
||||
promptMessage: "Paste your Wafer Pass API key",
|
||||
placeholder: "wfr_...",
|
||||
validation: {
|
||||
kind: "models-endpoint",
|
||||
provider: "Wafer Pass",
|
||||
modelsUrl: WAFER_MODELS_URL,
|
||||
},
|
||||
});
|
||||
|
||||
export const loginWaferServerless = createApiKeyLogin({
|
||||
providerLabel: "Wafer Serverless",
|
||||
authUrl: WAFER_AUTH_URL,
|
||||
instructions: "Create or copy your Wafer Serverless API key from the Wafer dashboard",
|
||||
promptMessage: "Paste your Wafer Serverless API key",
|
||||
placeholder: "wfr_...",
|
||||
validation: {
|
||||
kind: "models-endpoint",
|
||||
provider: "Wafer Serverless",
|
||||
modelsUrl: WAFER_MODELS_URL,
|
||||
},
|
||||
});
|
||||
@@ -0,0 +1,93 @@
|
||||
/**
|
||||
* Live Wafer Pass smoke. NOT part of the bun test suite — run manually:
|
||||
* WAFER_PASS_API_KEY=wfr_... bun packages/ai/test/wafer.live.ts
|
||||
*
|
||||
* Validates that the bundled `wafer-pass/GLM-5.1` entry round-trips a real
|
||||
* streaming chat completion against `https://pass.wafer.ai/v1`, with the wire
|
||||
* `model` field preserved verbatim (`GLM-5.1`, not lowercased) and a non-empty
|
||||
* assistant text returned.
|
||||
*/
|
||||
import { getBundledModel } from "../src/models";
|
||||
import { streamOpenAICompletions } from "../src/providers/openai-completions";
|
||||
import type { Context, Model } from "../src/types";
|
||||
|
||||
const apiKey = process.env.WAFER_PASS_API_KEY ?? process.env.WAFER_SERVERLESS_API_KEY;
|
||||
if (!apiKey) {
|
||||
console.error("WAFER_PASS_API_KEY (or WAFER_SERVERLESS_API_KEY) env var is required");
|
||||
process.exit(2);
|
||||
}
|
||||
|
||||
const providerId = process.env.WAFER_PASS_API_KEY ? "wafer-pass" : "wafer-serverless";
|
||||
const model = getBundledModel<"openai-completions">(providerId, "GLM-5.1");
|
||||
console.log(`Model: ${model.provider}/${model.id} -> ${model.baseUrl}`);
|
||||
console.log(`compat.thinkingFormat: ${model.compat?.thinkingFormat ?? "(none)"}`);
|
||||
|
||||
interface CapturedRequest {
|
||||
url: string;
|
||||
body: string | null;
|
||||
}
|
||||
|
||||
const originalFetch = global.fetch;
|
||||
const captured: { value: CapturedRequest | null } = { value: null };
|
||||
type FetchInput = Parameters<typeof fetch>[0];
|
||||
global.fetch = (async (input: FetchInput, init?: RequestInit) => {
|
||||
const url = typeof input === "string" ? input : input instanceof URL ? input.toString() : input.url;
|
||||
captured.value = { url, body: typeof init?.body === "string" ? init.body : null };
|
||||
return originalFetch(input as Parameters<typeof fetch>[0], init);
|
||||
}) as typeof global.fetch;
|
||||
|
||||
const context: Context = {
|
||||
systemPrompt: ["Reply with exactly two words."],
|
||||
messages: [{ role: "user", content: "Say hi.", timestamp: Date.now() }],
|
||||
};
|
||||
|
||||
const stream = streamOpenAICompletions(model as Model<"openai-completions">, context, { apiKey });
|
||||
let text = "";
|
||||
let stopReason: string | undefined;
|
||||
let cost = 0;
|
||||
let firstError: unknown;
|
||||
let inputTokens = 0;
|
||||
let outputTokens = 0;
|
||||
for await (const ev of stream) {
|
||||
if (ev.type === "text_delta") text += ev.delta;
|
||||
else if (ev.type === "done") {
|
||||
stopReason = ev.reason;
|
||||
const usage = ev.message.usage;
|
||||
cost = usage?.cost?.total ?? 0;
|
||||
inputTokens = usage?.input ?? 0;
|
||||
outputTokens = usage?.output ?? 0;
|
||||
} else if (ev.type === "error") {
|
||||
firstError = ev.error.errorMessage ?? ev.error;
|
||||
stopReason = ev.reason;
|
||||
}
|
||||
}
|
||||
|
||||
const snapshot = (captured as { value: CapturedRequest | null }).value;
|
||||
const parsedBody = snapshot?.body ? (JSON.parse(snapshot.body) as { model?: unknown }) : null;
|
||||
console.log("wire url:", snapshot?.url);
|
||||
console.log("wire model:", parsedBody?.model);
|
||||
console.log("text:", JSON.stringify(text.slice(0, 200)));
|
||||
console.log("stopReason:", stopReason);
|
||||
console.log("usage:", { input: inputTokens, output: outputTokens, costUSD: cost });
|
||||
|
||||
if (firstError) {
|
||||
console.error("\nLIVE FAIL — Wafer rejected the request:", firstError);
|
||||
process.exit(1);
|
||||
}
|
||||
if (snapshot?.url !== "https://pass.wafer.ai/v1/chat/completions") {
|
||||
console.error("\nLIVE FAIL — wire url was not the documented endpoint");
|
||||
process.exit(1);
|
||||
}
|
||||
if (parsedBody?.model !== "GLM-5.1") {
|
||||
console.error("\nLIVE FAIL — wire model id was not preserved verbatim:", parsedBody?.model);
|
||||
process.exit(1);
|
||||
}
|
||||
if (text.trim().length === 0) {
|
||||
console.error("\nLIVE FAIL — assistant returned empty text");
|
||||
process.exit(1);
|
||||
}
|
||||
|
||||
console.log(
|
||||
`\nLIVE OK — Wafer ${providerId} round-trip: GLM-5.1 endpoint preserved, ` +
|
||||
`${inputTokens}→${outputTokens} tokens, stopReason=${stopReason}.`,
|
||||
);
|
||||
@@ -0,0 +1,113 @@
|
||||
/**
|
||||
* Wafer Pass + Wafer Serverless provider wiring.
|
||||
*
|
||||
* Wafer exposes a single OpenAI-compatible base URL (`https://pass.wafer.ai/v1`)
|
||||
* for two SKUs whose entitlement differs server-side:
|
||||
* - `wafer-pass` (flat-rate)
|
||||
* - `wafer-serverless` (pay-as-you-go)
|
||||
*
|
||||
* Both providers route through `openai-completions` and the catalog id matches
|
||||
* the wire id (no rewrite). These tests defend the bundled catalog contract and
|
||||
* the case-sensitive id pass-through against the wire.
|
||||
*/
|
||||
import { afterEach, describe, expect, it } from "bun:test";
|
||||
import { getBundledModel } from "../src/models";
|
||||
import { streamOpenAICompletions } from "../src/providers/openai-completions";
|
||||
import type { Context, Model } from "../src/types";
|
||||
|
||||
const originalFetch = global.fetch;
|
||||
|
||||
afterEach(() => {
|
||||
global.fetch = originalFetch;
|
||||
});
|
||||
|
||||
function sseResponse(events: unknown[]): Response {
|
||||
const payload = `${events.map(e => `data: ${typeof e === "string" ? e : JSON.stringify(e)}`).join("\n\n")}\n\n`;
|
||||
return new Response(payload, {
|
||||
status: 200,
|
||||
headers: { "content-type": "text/event-stream" },
|
||||
});
|
||||
}
|
||||
|
||||
describe("Wafer Pass provider", () => {
|
||||
it("ships a bundled GLM-5.1 entry with zai-family thinking compat", () => {
|
||||
const model = getBundledModel<"openai-completions">("wafer-pass", "GLM-5.1");
|
||||
expect(model).toBeDefined();
|
||||
expect(model.id).toBe("GLM-5.1");
|
||||
expect(model.provider).toBe("wafer-pass");
|
||||
expect(model.api).toBe("openai-completions");
|
||||
expect(model.baseUrl).toBe("https://pass.wafer.ai/v1");
|
||||
expect(model.reasoning).toBe(true);
|
||||
expect(model.input).toEqual(["text"]);
|
||||
expect(model.compat?.thinkingFormat).toBe("zai");
|
||||
expect(model.compat?.reasoningContentField).toBe("reasoning_content");
|
||||
expect(model.compat?.supportsDeveloperRole).toBe(false);
|
||||
});
|
||||
|
||||
it("ships a bundled Qwen3.5-397B-A17B entry with vision input and no reasoning", () => {
|
||||
const model = getBundledModel<"openai-completions">("wafer-pass", "Qwen3.5-397B-A17B");
|
||||
expect(model).toBeDefined();
|
||||
expect(model.id).toBe("Qwen3.5-397B-A17B");
|
||||
expect(model.provider).toBe("wafer-pass");
|
||||
expect(model.reasoning).toBe(false);
|
||||
expect(model.input).toEqual(["text", "image"]);
|
||||
});
|
||||
|
||||
it("preserves the catalog id verbatim on the wire (no rewrite, case-sensitive)", async () => {
|
||||
const model = getBundledModel<"openai-completions">("wafer-pass", "GLM-5.1");
|
||||
const captured: { url: string | null; body: string | null } = { url: null, body: null };
|
||||
global.fetch = (async (input: unknown, init?: RequestInit) => {
|
||||
captured.url = typeof input === "string" ? input : input instanceof URL ? input.toString() : String(input);
|
||||
captured.body = typeof init?.body === "string" ? init.body : null;
|
||||
return sseResponse(["[DONE]"]);
|
||||
}) as typeof global.fetch;
|
||||
|
||||
const context: Context = {
|
||||
systemPrompt: ["t"],
|
||||
messages: [{ role: "user", content: "hi", timestamp: Date.now() }],
|
||||
};
|
||||
const stream = streamOpenAICompletions(model as Model<"openai-completions">, context, {
|
||||
apiKey: "wfr_test",
|
||||
});
|
||||
for await (const _event of stream) {
|
||||
/* drain */
|
||||
}
|
||||
|
||||
expect(captured.url).toBe("https://pass.wafer.ai/v1/chat/completions");
|
||||
expect(captured.body).not.toBeNull();
|
||||
const parsed = JSON.parse(captured.body ?? "{}") as { model?: unknown };
|
||||
// Wafer's docs note model names are case-insensitive on input, but the
|
||||
// canonical id has mixed case; we must round-trip it unchanged so users
|
||||
// who pin `GLM-5.1` don't end up with usage rows under `glm-5.1` or
|
||||
// hitting the upstream 404 path.
|
||||
expect(parsed.model).toBe("GLM-5.1");
|
||||
});
|
||||
});
|
||||
|
||||
describe("Wafer Serverless provider", () => {
|
||||
it("ships the documented Serverless catalog (GLM-5.1, Qwen3.5, Kimi-K2.6, Qwen3.6)", () => {
|
||||
const glm = getBundledModel<"openai-completions">("wafer-serverless", "GLM-5.1");
|
||||
expect(glm).toBeDefined();
|
||||
expect(glm.provider).toBe("wafer-serverless");
|
||||
expect(glm.baseUrl).toBe("https://pass.wafer.ai/v1");
|
||||
expect(glm.compat?.thinkingFormat).toBe("zai");
|
||||
|
||||
const qwen35 = getBundledModel<"openai-completions">("wafer-serverless", "Qwen3.5-397B-A17B");
|
||||
expect(qwen35).toBeDefined();
|
||||
expect(qwen35.provider).toBe("wafer-serverless");
|
||||
|
||||
const kimi = getBundledModel<"openai-completions">("wafer-serverless", "Kimi-K2.6");
|
||||
expect(kimi).toBeDefined();
|
||||
expect(kimi.contextWindow).toBe(262144);
|
||||
|
||||
const qwen36 = getBundledModel<"openai-completions">("wafer-serverless", "Qwen3.6-35B-A3B");
|
||||
expect(qwen36).toBeDefined();
|
||||
// Serverless-only Qwen3.6 — context capped at 32k per docs.
|
||||
expect(qwen36.contextWindow).toBe(32768);
|
||||
});
|
||||
|
||||
it("does not expose Serverless-only ids on the Wafer Pass catalog", () => {
|
||||
expect(getBundledModel("wafer-pass", "Kimi-K2.6")).toBeUndefined();
|
||||
expect(getBundledModel("wafer-pass", "Qwen3.6-35B-A3B")).toBeUndefined();
|
||||
});
|
||||
});
|
||||
@@ -247,6 +247,8 @@ export function getExtraHelpText(): string {
|
||||
OPENCODE_API_KEY - OpenCode Zen/OpenCode Go models
|
||||
CURSOR_ACCESS_TOKEN - Cursor AI models
|
||||
AI_GATEWAY_API_KEY - Vercel AI Gateway
|
||||
WAFER_PASS_API_KEY - Wafer Pass (flat-rate subscription; GLM-5.1, Qwen3.5)
|
||||
WAFER_SERVERLESS_API_KEY - Wafer Serverless (pay-as-you-go)
|
||||
|
||||
${chalk.dim("# Cloud Providers")}
|
||||
AWS_PROFILE - AWS Bedrock (or AWS_ACCESS_KEY_ID + AWS_SECRET_ACCESS_KEY)
|
||||
|
||||
Reference in New Issue
Block a user