fix: revert remove routeway provider

.env.example

@@ -179,6 +179,12 @@ LLM_NANO_GPT_API_KEY=your_nanogpt_key_here
 # Z.ai
 LLM_Z_AI_API_KEY=your_zai_key_here
 
+# Routeway
+LLM_ROUTEWAY_API_KEY=your_routeway_key_here
+
+# Routeway Discount
+LLM_ROUTEWAY_DISCOUNT_API_KEY=your_routeway_discount_key_here
+
 # AWS Bedrock
 LLM_AWS_BEDROCK_API_KEY=your_aws_bedrock_key_here
 

.env.unified.example

@@ -94,6 +94,12 @@ LLM_NANO_GPT_API_KEY=your_nanogpt_key_here
 # Z.ai
 LLM_Z_AI_API_KEY=your_zai_key_here
 
+# Routeway
+LLM_ROUTEWAY_API_KEY=your_routeway_key_here
+
+# Routeway Discount
+LLM_ROUTEWAY_DISCOUNT_API_KEY=your_routeway_discount_key_here
+
 # AWS Bedrock
 LLM_AWS_BEDROCK_API_KEY=your_aws_bedrock_key_here
 

.github/workflows/e2e.yml

@@ -70,6 +70,9 @@ jobs:
           LLM_NEBIUS_API_KEY: ${{ secrets.NEBIUS_API_KEY }}
           LLM_Z_AI_API_KEY: ${{ secrets.Z_AI_API_KEY }}
           LLM_NANO_GPT_API_KEY: ${{ secrets.NANO_GPT_API_KEY }}
+          LLM_ROUTEWAY_API_KEY: ${{ secrets.ROUTEWAY_API_KEY }}
+          LLM_ROUTEWAY_DISCOUNT_API_KEY: ${{ secrets.ROUTEWAY_DISCOUNT_API_KEY }}
+          LLM_ROUTEWAY_DISCOUNT_BASE_URL: ${{ secrets.ROUTEWAY_DISCOUNT_BASE_URL }}
           LLM_AWS_BEDROCK_API_KEY: ${{ secrets.AWS_BEDROCK_API_KEY }}
           LLM_AZURE_API_KEY: ${{ secrets.AZURE_API_KEY }}
           LLM_AZURE_RESOURCE: ${{ secrets.AZURE_RESOURCE }}

.run/dev.run.xml

@@ -7,6 +7,7 @@
     </scripts>
     <node-interpreter value="project" />
     <envs>
+      <env name="ROUTEWAY_PAID_DISCOUNT" value="0.5" />
       <env name="AUTH_SECRET" value="asdf" />
     </envs>
     <method v="2">

apps/api/src/routes/keys-provider.e2e.ts

@@ -126,8 +126,12 @@ describe(
 		test.each(testProviders)(
 			"POST /keys/provider with $name key",
 			async ({ providerId }) => {
-				// TODO temporarily skip nanogpt
-				if (providerId === "inference.net" || providerId === "nanogpt") {
+				// TODO temporarily skip routeway and nanogpt
+				if (
+					providerId === "routeway" ||
+					providerId === "inference.net" ||
+					providerId === "nanogpt"
+				) {
 					return;
 				}
 				const envVarName = getProviderEnvVar(providerId);

apps/docs/content/guides/cursor.mdx

@@ -122,7 +122,8 @@ Cursor allows you to configure different models for different features:
 2. **Autocomplete Model**: Use a faster, cost-effective model like `gpt-4o-mini`
 3. **Provider Specific Model**: Use a provider specific model like `openai/gpt-5`, `anthropic/claude-sonnet-4-5`, `google-ai-studio/gemini-2.0-flash-exp`
 4. **Custom Model**: Use a custom model like `custom/my-model`
-5. **Reasoning Model**: Use a reasoning model like `canopywave/kimi-k2-thinking` [with 75% off discount](https://llmgateway.io/changelog/canopywave-kimi-k2-thinking-discount)
+5. **Free Model**: Use a free model like `routeway/deepseek-r1t2-chimera-free`
+6. **Reasoning Model**: Use a reasoning model like `canopywave/kimi-k2-thinking` [with 75% off discount](https://llmgateway.io/changelog/canopywave-kimi-k2-thinking-discount)
 
 This gives you the best balance of performance and cost.
 

apps/gateway/src/chat/chat.ts

@@ -4670,6 +4670,10 @@ chat.openapi(completions, async (c) => {
 						finalPluginResults, // Plugin results including healing (if enabled)
 					);
 
+					if (!finishReason && !streamingError && usedProvider === "routeway") {
+						finishReason = "stop";
+					}
+
 					// Enhanced logging for Google models streaming to debug missing responses
 					if (
 						usedProvider === "google-ai-studio" ||

apps/gateway/src/chat/tools/transform-streaming-to-openai.ts

@@ -1095,6 +1095,7 @@ export function transformStreamingToOpenai(
 
 		case "mistral":
 		case "novita":
+		case "routeway":
 		case "zai": {
 			// Transform standard OpenAI streaming format with finish reason mapping
 			transformedData = transformOpenaiStreaming(data, usedModel);

apps/gateway/src/lib/costs.spec.ts

@@ -146,6 +146,10 @@ describe("calculateCosts", () => {
 		);
 		expect(resultWithoutDiscount.discount).toBeUndefined(); // No discount field when discount is 1
 
+		// The actual test for routeway discount would require the models to be re-imported
+		// after setting the env var, which is complex in a test environment.
+		// Instead, let's verify the logic works by testing the cost calculation directly.
+
 		// Test that the discount field appears when a discount is applied (using the actual logic from costs.ts)
 		const testDiscount = 0.8; // 80% off
 		const discountMultiplier = 1 - testDiscount; // Pay 20% of original price
@@ -156,7 +160,8 @@ describe("calculateCosts", () => {
 		const expectedOutputCost = 50 * outputPrice * discountMultiplier;
 		const expectedTotalCost = expectedInputCost + expectedOutputCost;
 
-		// Verify our calculation logic is sound
+		// Since we can't easily test the routeway model due to env var timing,
+		// let's verify our calculation logic is sound
 		expect(expectedInputCost).toBeCloseTo(0.000016); // 100 * 0.8e-6 * 0.2
 		expect(expectedOutputCost).toBeCloseTo(0.00004); // 50 * 4.0e-6 * 0.2
 		expect(expectedTotalCost).toBeCloseTo(0.000056);

apps/ui/src/components/landing/hero.tsx

@@ -52,6 +52,7 @@ const PROVIDER_LOGOS: { name: string; providerId: ProviderId }[] = [
 	{ name: "Novita", providerId: "novita" },
 	{ name: "Nebius", providerId: "nebius" },
 	{ name: "Zai", providerId: "zai" },
+	{ name: "Routeway", providerId: "routeway" },
 	{ name: "NanoGPT", providerId: "nanogpt" },
 	{ name: "Canopywave", providerId: "canopywave" },
 	{ name: "AWS Bedrock", providerId: "aws-bedrock" },

apps/ui/src/components/provider-keys/provider-logo.ts

@@ -22,6 +22,7 @@ export const providerLogoUrls: Partial<
 	alibaba: ProviderIcons.alibaba,
 	nebius: ProviderIcons.nebius,
 	zai: ProviderIcons.zai,
+	routeway: ProviderIcons.routeway,
 	nanogpt: ProviderIcons.nanogpt,
 	"aws-bedrock": ProviderIcons["aws-bedrock"],
 	azure: ProviderIcons.azure,

apps/ui/src/content/changelog/2025-09-02-routeway-free-models.md

@@ -0,0 +1,94 @@
+---
+id: "17"
+slug: "routeway-free-models"
+date: "2025-09-02"
+title: "Enhanced Auto-Routing & 5 New Free Models via RouteWay"
+summary: "Expanding auto-routing options & access DeepSeek R1T2 Chimera, GLM-4.5 Air, Kimi K2, GPT-OSS 20B, and GPT-4.1 completely free through our RouteWay integration."
+image:
+  src: "/changelog/routeway-free-models.png"
+  alt: "Enhanced Auto-Routing & 5 New Free Models via RouteWay"
+  width: 1768
+  height: 677
+---
+
+We're excited to expand our **auto-routing capabilities** with **5 new free models** through our [RouteWay provider](/providers/routeway) integration! These models enhance our intelligent routing system by providing more cost-effective alternatives.
+
+## 🤖 Auto-Routing with Free Models
+
+**Expanded Free Options**: Our intelligent routing system now has more free alternatives to choose from.
+
+**Example Usage**: Use auto-routing with the `free_models_only` flag to prioritize these cost-effective models:
+
+```bash
+curl -X POST "https://api.llmgateway.io/v1/chat/completions" \
+  -H "Authorization: Bearer $LLM_GATEWAY_API_KEY" \
+  -H "Content-Type: application/json" \
+  -d '{
+    "model": "auto",
+    "free_models_only": true,
+    "messages": [
+      {
+        "role": "user",
+        "content": "Explain quantum computing in simple terms"
+      }
+    ]
+  }'
+```
+
+The system will automatically select the best available free model from the expanded pool, including these new RouteWay options.
+
+## 🆓 New Free Models
+
+All models are available with generous usage limits and zero per-token costs:
+
+### DeepSeek R1T2 Chimera (Free)
+
+```bash
+deepseek-r1t2-chimera-free
+```
+
+- **Cost**: $0.00 input/output tokens
+- **Context Window**: 128,000 tokens
+- **Capabilities**: Streaming, JSON output
+
+### GLM-4.5 Air (Free)
+
+```bash
+glm-4.5-air-free
+```
+
+- **Cost**: $0.00 input/output tokens
+- **Context Window**: 128,000 tokens
+- **Capabilities**: Streaming, JSON output
+
+### Kimi K2 (Free)
+
+```bash
+kimi-k2-free
+```
+
+- **Cost**: $0.00 input/output tokens
+- **Context Window**: 128,000 tokens
+- **Capabilities**: Streaming, JSON output
+
+### GPT-OSS 20B (Free)
+
+```bash
+gpt-oss-20b-free
+```
+
+- **Cost**: $0.00 input/output tokens
+- **Context Window**: 128,000 tokens
+- **Capabilities**: Streaming, JSON output
+
+## 🚀 Getting Started
+
+**Auto-Routing**: These models are automatically available in our routing system - no configuration needed.
+
+**Direct Access**: Use the model identifiers above (with `-free` suffix) to access specific models directly.
+
+**Browse Models**: Visit our [Models Directory](/models) to explore and compare all available options.
+
+---
+
+This expansion significantly enhances our auto-routing system's ability to provide cost-effective AI solutions while maintaining high quality and availability.

apps/ui/src/content/changelog/2025-09-19-claude-code-50-percent-off.md

@@ -0,0 +1,42 @@
+---
+id: "21"
+slug: "claude-code-50-percent-off"
+date: "2025-09-19"
+title: "50% Off Claude Code - Limited Time Offer"
+summary: "Get 50% off Claude Code for a limited time. Use any model via LLMGateway's unified API."
+image:
+  src: "/changelog/claude-code-50-percent-off.png"
+  alt: "Claude Code 50% off limited time offer"
+  width: 1768
+  height: 677
+---
+
+🎉 **Limited Time: 50% Off Claude Code via RouteWay!**
+
+For a limited time, get **50% off Claude Code** through our RouteWay provider integration.
+
+## Setup Instructions
+
+To access Claude Code with the 50% discount, configure these environment variables:
+
+```shell
+export ANTHROPIC_BASE_URL=https://api.llmgateway.io
+export ANTHROPIC_AUTH_TOKEN=llmgtwy_.... # your llmgateway.io api key here
+
+# now run claude!
+claude
+```
+
+### Optional Variable
+
+```shell
+# optionally, choose your model, otherwise it will use the default Claude model via LLMGateway
+export ANTHROPIC_MODEL=gpt-5
+
+# now run claude!
+claude
+```
+
+---
+
+Don't miss this limited-time opportunity to access Claude Code at 50% off via RouteWay!

packages/actions/src/get-provider-endpoint.ts

@@ -116,6 +116,9 @@ export function getProviderEndpoint(
 			case "zai":
 				url = "https://api.z.ai";
 				break;
+			case "routeway":
+				url = "https://api.routeway.ai";
+				break;
 			case "nanogpt":
 				url = "https://nano-gpt.com/api";
 				break;
@@ -361,6 +364,7 @@ export function getProviderEndpoint(
 		case "deepseek":
 		case "moonshot":
 		case "nebius":
+		case "routeway":
 		case "nanogpt":
 		case "canopywave":
 		case "minimax":

packages/models/src/models.ts

@@ -12,6 +12,7 @@ import { moonshotModels } from "./models/moonshot.js";
 import { nousresearchModels } from "./models/nousresearch.js";
 import { openaiModels } from "./models/openai.js";
 import { perplexityModels } from "./models/perplexity.js";
+import { routewayModels } from "./models/routeway.js";
 import { xaiModels } from "./models/xai.js";
 import { zaiModels } from "./models/zai.js";
 
@@ -257,5 +258,6 @@ export const models = [
 	...alibabaModels,
 	...bytedanceModels,
 	...nousresearchModels,
+	...routewayModels,
 	...zaiModels,
 ] as const satisfies ModelDefinition[];