Merge pull request #1175 from inf3rnus/main

Integration: Bytez Chat Model Provider

Author: VisargD

src/globals.ts

@@ -95,6 +95,7 @@ export const LEPTON: string = 'lepton';
 export const KLUSTER_AI: string = 'kluster-ai';
 export const NSCALE: string = 'nscale';
 export const HYPERBOLIC: string = 'hyperbolic';
+export const BYTEZ: string = 'bytez';
 export const FEATHERLESS_AI: string = 'featherless-ai';
 export const KRUTRIM: string = 'krutrim';
 export const QDRANT: string = 'qdrant';
@@ -158,6 +159,7 @@ export const VALID_PROVIDERS = [
   KLUSTER_AI,
   NSCALE,
   HYPERBOLIC,
+  BYTEZ,
   FEATHERLESS_AI,
   KRUTRIM,
   QDRANT,

src/providers/bytez/api.ts

@@ -0,0 +1,20 @@
+import { ProviderAPIConfig } from '../types';
+import { version } from '../../../package.json';
+
+const BytezInferenceAPI: ProviderAPIConfig = {
+  getBaseURL: () => 'https://api.bytez.com',
+  headers: async ({ providerOptions }) => {
+    const { apiKey } = providerOptions;
+
+    const headers: Record<string, string> = {};
+
+    headers['Authorization'] = `Key ${apiKey}`;
+    headers['user-agent'] = `portkey/${version}`;
+
+    return headers;
+  },
+  getEndpoint: ({ gatewayRequestBodyJSON: { version = 2, model } }) =>
+    `/models/v${version}/${model}`,
+};
+
+export default BytezInferenceAPI;

src/providers/bytez/chatComplete.ts

@@ -0,0 +1,77 @@
+import { BYTEZ } from '../../globals';
+import { ProviderConfig } from '../types';
+import { BytezResponse } from './types';
+import { generateErrorResponse } from '../utils';
+
+const BytezInferenceChatCompleteConfig: ProviderConfig = {
+  messages: {
+    param: 'messages',
+    required: true,
+  },
+  max_tokens: {
+    param: 'params.max_new_tokens',
+    default: 100,
+    min: 0,
+  },
+  temperature: {
+    param: 'params.temperature',
+    default: 1,
+    min: 0,
+    max: 2,
+  },
+  top_p: {
+    param: 'params.top_p',
+    default: 1,
+    min: 0,
+    max: 1,
+  },
+  stream: {
+    param: 'stream',
+    default: false,
+  },
+};
+
+function chatComplete(
+  response: BytezResponse,
+  responseStatus: number,
+  responseHeaders: any,
+  strictOpenAiCompliance: boolean,
+  endpoint: string,
+  requestBody: any
+) {
+  const { error, output } = response;
+
+  if (error) {
+    return generateErrorResponse(
+      {
+        message: error,
+        type: String(responseStatus),
+        param: null,
+        code: null,
+      },
+      BYTEZ
+    );
+  }
+
+  return {
+    id: crypto.randomUUID(),
+    object: 'chat.completion',
+    created: Date.now(),
+    model: requestBody.model,
+    choices: [
+      {
+        index: 0,
+        message: output,
+        logprobs: null,
+        finish_reason: 'stop',
+      },
+    ],
+    usage: {
+      completion_tokens: -1,
+      prompt_tokens: -1,
+      total_tokens: -1,
+    },
+  };
+}
+
+export { BytezInferenceChatCompleteConfig, chatComplete };

src/providers/bytez/index.ts

@@ -0,0 +1,13 @@
+import { ProviderConfigs } from '../types';
+import BytezInferenceAPI from './api';
+import { BytezInferenceChatCompleteConfig, chatComplete } from './chatComplete';
+
+const BytezInferenceAPIConfig: ProviderConfigs = {
+  api: BytezInferenceAPI,
+  chatComplete: BytezInferenceChatCompleteConfig,
+  responseTransforms: {
+    chatComplete,
+  },
+};
+
+export default BytezInferenceAPIConfig;

src/providers/bytez/types.ts

@@ -0,0 +1,10 @@
+interface Model {
+  task: string;
+}
+
+interface BytezResponse {
+  error: string;
+  output: Model[];
+}
+
+export { Model, BytezResponse };

src/providers/index.ts

@@ -1,3 +1,4 @@
+import BytezConfig from './bytez';
 import AI21Config from './ai21';
 import AnthropicConfig from './anthropic';
 import AnyscaleConfig from './anyscale';
@@ -120,6 +121,7 @@ const Providers: { [key: string]: ProviderConfigs } = {
   'kluster-ai': KlusterAIConfig,
   nscale: NscaleConfig,
   hyperbolic: HyperbolicConfig,
+  bytez: BytezConfig,
   'featherless-ai': FeatherlessAIConfig,
   krutrim: KrutrimConfig,
 };

src/public/index.html

@@ -1115,6 +1115,7 @@ <h3>Select Provider</h3>
             <option value="together-ai">Together AI</option>
             <option value="perplexity-ai">Perplexity AI</option>
             <option value="mistral-ai">Mistral AI</option>
+            <option value="bytez">Bytez</option>
             <option value="others">Others</option>
           </select>
         </div>
@@ -1465,6 +1466,7 @@ <h3>Enter API Key</h3>
         "together-ai": "llama-3.1-8b-instruct",
         "perplexity-ai": "pplx-7b-online",
         "mistral-ai": "mistral-small-latest",
+        "bytez": "google/gemma-3-1b-it",
         "others": "gpt-4o-mini"
       }
 

src/utils.ts

@@ -7,6 +7,7 @@ import {
   DEEPINFRA,
   SAMBANOVA,
   BEDROCK,
+  BYTEZ,
 } from './globals';
 import { Params } from './types/requestBody';
 
@@ -48,9 +49,13 @@ export const getStreamModeSplitPattern = (
     splitPattern = '\n';
   }
 
+  if (proxyProvider === BYTEZ) {
+    splitPattern = ' ';
+  }
+
   return splitPattern;
 };
-export type SplitPatternType = '\n\n' | '\r\n\r\n' | '\n' | '\r\n';
+export type SplitPatternType = '\n\n' | '\r\n\r\n' | '\n' | '\r\n' | ' ';
 
 export const getStreamingMode = (
   reqBody: Params,