bios-sdk 0.1.1-rc.26 → 0.1.1-rc.35

package/README.md

@@ -107,6 +107,39 @@ for await (const chunk of client.inference.streamChatCompletions({
 // controller.abort() cancels an unfinished generation.
 ```
 
+#### Serverless catalog models
+
+Call any catalog model by id on the unified `/v1` endpoint with a workspace
+platform key that carries the serverless scope — no per-deployment key. The
+gateway routes by `model`; dedicated deployments and serverless models share the
+same endpoint. `reasoningEffort` is forwarded to the endpoint, and streaming
+surfaces `content` and `reasoning_content` deltas incrementally.
+
+```typescript
+const client = new BiOS({
+  inferenceKey: 'bios-platform-key-with-serverless-scope',
+});
+
+for await (const chunk of client.inference.streamChatCompletions({
+  model: 'meta-llama/Llama-3.1-8B-Instruct', // serverless catalog id
+  messages: [{ role: 'user', content: 'Explain tensor parallelism briefly.' }],
+  reasoningEffort: 'low',
+})) {
+  const delta = (chunk.choices as any)?.[0]?.delta ?? {};
+  if (delta.reasoning_content) process.stdout.write(delta.reasoning_content);
+  if (delta.content) process.stdout.write(delta.content);
+}
+
+// Non-streaming; the final chunk carries `usage` when the model reports it.
+const completion = await client.inference.chatCompletions({
+  model: 'meta-llama/Llama-3.1-8B-Instruct',
+  messages: [{ role: 'user', content: 'One sentence on GPUs.' }],
+});
+```
+
+Streaming billing is charged server-side on completed usage; the SDK only needs
+to request `usage` where the endpoint exposes it (no client change).
+
 ### Models
 
 The catalog lists only models hosted on BiOS (the platform's own verified

package/dist/resources/inference.d.ts

@@ -24,6 +24,12 @@ export interface ChatCompletionParams extends Record<string, unknown> {
             name: string;
         };
     };
+    /**
+     * Standardized reasoning effort. Forwarded to `/v1/chat/completions` as
+     * `reasoning_effort`. Use `'none'` to disable reasoning where the model
+     * allows it.
+     */
+    reasoningEffort?: 'none' | 'minimal' | 'low' | 'medium' | 'high' | 'max';
     inferenceKey?: string;
     idempotencyKey?: string;
     requestId?: string;

package/dist/resources/inference.js

@@ -580,7 +580,7 @@ export class Inference {
     // OpenAI-compatible key-scoped inference — `/v1/chat/completions`
     // --------------------------------------------------------------------------
     prepare(params, stream) {
-        const { messages, model, tools, toolChoice, inferenceKey, idempotencyKey, requestId, signal, ...extra } = params;
+        const { messages, model, tools, toolChoice, reasoningEffort, inferenceKey, idempotencyKey, requestId, signal, ...extra } = params;
         const key = inferenceKey || this.key;
         if (!key)
             throw new Error('an inferenceKey is required');
@@ -591,6 +591,8 @@ export class Inference {
             body.tools = tools;
         if (toolChoice !== undefined)
             body.tool_choice = toolChoice;
+        if (reasoningEffort !== undefined)
+            body.reasoning_effort = reasoningEffort;
         validateChatRequest(body);
         const headers = {
             Authorization: `Bearer ${key}`,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "bios-sdk",
-  "version": "0.1.1-rc.26",
+  "version": "0.1.1-rc.35",
   "description": "Official TypeScript SDK for the BIOS training and deployment platform API",
   "type": "module",
   "main": "./dist/index.js",