provider: SambaNova FastAPI implementation.

src/globals.ts

@@ -67,6 +67,7 @@ export const GITHUB: string = 'github';
 export const DEEPBRICKS: string = 'deepbricks';
 export const SILICONFLOW: string = 'siliconflow';
 export const CEREBRAS: string = 'cerebras';
+export const SAMBANOVA: string = 'sambanova';
 
 export const VALID_PROVIDERS = [
   ANTHROPIC,
@@ -108,6 +109,7 @@ export const VALID_PROVIDERS = [
   SILICONFLOW,
   HUGGING_FACE,
   CEREBRAS,
+  SAMBANOVA,
 ];
 
 export const CONTENT_TYPES = {

src/providers/index.ts

@@ -40,6 +40,7 @@ import DeepbricksConfig from './deepbricks';
 import SiliconFlowConfig from './siliconflow';
 import HuggingfaceConfig from './huggingface';
 import { cerebrasProviderAPIConfig } from './cerebras';
+import SambaNovaConfig from './sambanova';
 
 const Providers: { [key: string]: ProviderConfigs } = {
   openai: OpenAIConfig,
@@ -81,6 +82,7 @@ const Providers: { [key: string]: ProviderConfigs } = {
   deepbricks: DeepbricksConfig,
   siliconflow: SiliconFlowConfig,
   cerebras: cerebrasProviderAPIConfig,
+  sambanova: SambaNovaConfig,
 };
 
 export default Providers;

src/providers/sambanova/api.ts

@@ -0,0 +1,19 @@
+import { ProviderAPIConfig } from '../types';
+
+const SambaNovaAPIConfig: ProviderAPIConfig = {
+  getBaseURL: ({ providerOptions }) =>
+    providerOptions.urlToFetch || 'https://fast-api.snova.ai/v1',
+  headers: ({ providerOptions }) => {
+    return { Authorization: `Basic ${providerOptions.apiKey}` };
+  },
+  getEndpoint: ({ fn }) => {
+    switch (fn) {
+      case 'chatComplete':
+        return '/chat/completions';
+      default:
+        return '';
+    }
+  },
+};
+
+export default SambaNovaAPIConfig;

src/providers/sambanova/chatComplete.ts

@@ -0,0 +1,114 @@
+import { SAMBANOVA } from '../../globals';
+import {
+  ChatCompletionResponse,
+  ErrorResponse,
+  ProviderConfig,
+} from '../types';
+
+export const SamvaNovaChatCompleteConfig: ProviderConfig = {
+  model: {
+    param: 'model',
+    required: true,
+    default: 'Meta-Llama-3.1-8B-Instruct',
+  },
+  messages: {
+    param: 'messages',
+    default: '',
+  },
+  max_tokens: {
+    param: 'max_tokens',
+    default: 100,
+    min: 0,
+  },
+  // Currently FastAPI supports stream responses only.
+  stream: {
+    param: 'stream',
+    default: true,
+  },
+  stream_options: {
+    param: 'stream_options',
+  },
+  stop: {
+    param: 'stop',
+  },
+};
+
+export interface SamvaNovaChatCompleteResponse extends ChatCompletionResponse {}
+
+export interface SamvaNovaErrorResponse extends ErrorResponse {}
+
+export interface SamvaNovaStreamChunk {
+  id: string;
+  object: string;
+  created: number;
+  model: string;
+  system_fingerprint: string;
+  choices: {
+    delta: {
+      content?: string;
+    };
+    index: number;
+    finish_reason: string | null;
+    logprobs: object | null;
+  }[];
+  usage: {
+    is_last_response: boolean;
+    total_tokens: number;
+    prompt_tokens: number;
+    completion_tokens: number;
+    time_to_first_token: number;
+    end_time: number;
+    start_time: number;
+    total_latency: number;
+    total_tokens_per_sec: number;
+    completion_tokens_per_sec: number;
+    completion_tokens_after_first_per_sec: number;
+    completion_tokens_after_first_per_sec_first_ten: number;
+  } | null;
+}
+
+export const SamvaNovaChatCompleteStreamChunkTransform: (
+  response: string
+) => string = (responseChunk) => {
+  let chunk = responseChunk.trim();
+  chunk = chunk.replace(/^data: /, '');
+  chunk = chunk.trim();
+  if (chunk === '[DONE]') {
+    return `data: ${chunk}\n\n`;
+  }
+
+  const parsedChunk: SamvaNovaStreamChunk = JSON.parse(chunk);
+  if (parsedChunk.usage) {
+    return `data: ${JSON.stringify({
+      id: parsedChunk.id,
+      object: parsedChunk.object,
+      created: parsedChunk.created,
+      model: parsedChunk.model,
+      provider: SAMBANOVA,
+      choices: [],
+      usage: {
+        prompt_tokens: parsedChunk.usage.prompt_tokens || 0,
+        completion_tokens: parsedChunk.usage.completion_tokens || 0,
+        total_tokens: parsedChunk.usage.total_tokens || 0,
+      },
+    })}\n\n`;
+  }
+  return `data: ${JSON.stringify({
+    id: parsedChunk.id,
+    object: parsedChunk.object,
+    created: parsedChunk.created,
+    model: parsedChunk.model,
+    provider: SAMBANOVA,
+    choices: [
+      {
+        index: parsedChunk.choices[0].index || 0,
+        delta: {
+          role: 'assistant',
+          content: parsedChunk.choices[0].delta.content,
+        },
+        logprobs: null,
+        finish_reason: parsedChunk.choices[0].finish_reason || null,
+      },
+    ],
+  })}\n\n`;
+};

src/providers/sambanova/index.ts

@@ -0,0 +1,16 @@
+import { ProviderConfigs } from '../types';
+import SambaNovaAPIConfig from './api';
+import {
+  SamvaNovaChatCompleteConfig,
+  SamvaNovaChatCompleteStreamChunkTransform,
+} from './chatComplete';
+
+const SambaNovaConfig: ProviderConfigs = {
+  chatComplete: SamvaNovaChatCompleteConfig,
+  api: SambaNovaAPIConfig,
+  responseTransforms: {
+    'stream-chatComplete': SamvaNovaChatCompleteStreamChunkTransform,
+  },
+};
+
+export default SambaNovaConfig;