@ax-llm/ax 16.1.12 → 17.0.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ax-llm/ax",
-  "version": "16.1.12",
+  "version": "17.0.0",
   "type": "module",
   "description": "The best library to work with LLMs",
   "repository": {

package/skills/ax-agent.md

@@ -1,12 +1,12 @@
 ---
 name: ax-agent
 description: This skill helps with building AxAgent-based agents using @ax-llm/ax. Use when the user asks about agent(), AxAgent, child agents, tool functions, smart model routing, RLM mode, stopping agents, or composing multi-agent hierarchies.
-version: "16.1.12"
+version: "17.0.0"
 ---
 
 # AxAgent Guide (@ax-llm/ax)
 
-AxAgent is the agent framework in Ax. It wraps AxGen with support for child agents, tool use, smart model routing, and RLM (Recursive Language Model) mode for processing long contexts through a code interpreter.
+AxAgent is the agent framework in Ax. It wraps AxGen with support for child agents, tool use, smart model routing, and RLM (Recursive Language Model) mode for processing long contexts through runtime-backed code execution.
 
 ## Quick Reference
 
@@ -167,7 +167,7 @@ const result = await myAgent.forward(llm, values, {
 
 ### `stop()` method
 
-Call `stop()` from any context — a timer, event handler, or another async task — to halt the multi-step loop:
+Call `stop()` from any context — a timer, event handler, or another async task — to halt the multi-step loop. `stop()` aborts all in-flight calls started by the same `AxAgent` instance (including retry backoff waits):
 
 ```typescript
 const myAgent = agent('question:string -> answer:string', {
@@ -266,7 +266,7 @@ const weatherAgent = agent('query:string -> response:string', {
   functions: [getCurrentWeather]
 });
 
-const llm = ai({ name: 'openai', apiKey: process.env.OPENAI_API_KEY! });
+const llm = ai({ name: 'openai', apiKey: process.env.OPENAI_APIKEY! });
 const result = await weatherAgent.forward(llm, { query: 'Weather in Tokyo?' });
 ```
 
@@ -358,7 +358,7 @@ The LLM writes code to chunk, filter, and iterate over the document, using `llmQ
 
 ```typescript
 import { agent, ai } from '@ax-llm/ax';
-import { AxRLMJSInterpreter } from '@ax-llm/ax';
+import { AxJSRuntime } from '@ax-llm/ax';
 
 const analyzer = agent(
   'context:string, query:string -> answer:string, evidence:string[]',
@@ -367,8 +367,8 @@ const analyzer = agent(
     description: 'Analyzes long documents using code interpreter and sub-LM queries',
     maxSteps: 15,
     rlm: {
-      contextFields: ['context'],              // Fields to load into interpreter
-      interpreter: new AxRLMJSInterpreter(),   // Code interpreter implementation
+      contextFields: ['context'],              // Fields to load into runtime session
+      runtime: new AxJSRuntime(),          // Code runtime implementation
       maxLlmCalls: 30,                         // Cap on sub-LM calls (default: 50)
       subModel: 'gpt-4o-mini',                // Model for llmQuery (default: same as parent)
     },
@@ -378,15 +378,15 @@ const analyzer = agent(
 
 ### Sandbox Permissions
 
-By default, the `AxRLMJSInterpreter` sandbox blocks all dangerous Web APIs (network, storage, etc.). You can selectively grant access using the `AxRLMJSInterpreterPermission` enum:
+By default, the `AxJSRuntime` sandbox blocks all dangerous Web APIs (network, storage, etc.). You can selectively grant access using the `AxJSRuntimePermission` enum:
 
 ```typescript
-import { AxRLMJSInterpreter, AxRLMJSInterpreterPermission } from '@ax-llm/ax';
+import { AxJSRuntime, AxJSRuntimePermission } from '@ax-llm/ax';
 
-const interpreter = new AxRLMJSInterpreter({
+const runtime = new AxJSRuntime({
   permissions: [
-    AxRLMJSInterpreterPermission.NETWORK,
-    AxRLMJSInterpreterPermission.STORAGE,
+    AxJSRuntimePermission.NETWORK,
+    AxJSRuntimePermission.STORAGE,
   ],
 });
 ```
@@ -410,7 +410,7 @@ Context fields aren't limited to plain strings. You can pass structured data —
 
 ```typescript
 import { agent, f, s } from '@ax-llm/ax';
-import { AxRLMJSInterpreter } from '@ax-llm/ax';
+import { AxJSRuntime } from '@ax-llm/ax';
 
 const sig = s('query:string -> answer:string, evidence:string[]')
   .appendInputField('documents', f.object({
@@ -424,7 +424,7 @@ const analyzer = agent(sig, {
   description: 'Analyzes structured document collections using RLM',
   rlm: {
     contextFields: ['documents'],
-    interpreter: new AxRLMJSInterpreter(),
+    runtime: new AxJSRuntime(),
   },
 });
 ```
@@ -470,7 +470,7 @@ In RLM mode, the agent gets a `codeInterpreter` tool. The LLM's typical workflow
 
 ### Custom Interpreters
 
-The built-in `AxRLMJSInterpreter` uses Web Workers for sandboxed code execution. For other environments, implement the `AxCodeInterpreter` interface:
+The built-in `AxJSRuntime` uses Web Workers for sandboxed code execution. For other environments, implement the `AxCodeInterpreter` interface:
 
 ```typescript
 import type { AxCodeInterpreter, AxCodeSession } from '@ax-llm/ax';
@@ -507,7 +507,8 @@ RLM mode does not support true streaming. When using `streamingForward`, RLM run
 ```typescript
 interface AxRLMConfig {
   contextFields: string[];        // Input fields holding long context
-  interpreter: AxCodeInterpreter; // Code interpreter implementation
+  runtime?: AxCodeInterpreter;    // Preferred runtime key
+  interpreter?: AxCodeInterpreter; // Legacy alias (deprecated)
   maxLlmCalls?: number;           // Cap on sub-LM calls (default: 50)
   subModel?: string;              // Model for llmQuery sub-calls
 }

package/skills/ax-llm.md

@@ -1,7 +1,7 @@
 ---
 name: ax-llm
 description: This skill helps with using the @ax-llm/ax TypeScript library for building LLM applications. Use when the user asks about ax(), ai(), f(), s(), agent(), flow(), AxGen, AxAgent, AxFlow, signatures, streaming, or mentions @ax-llm/ax.
-version: "16.1.12"
+version: "17.0.0"
 ---
 
 # Ax Library (@ax-llm/ax) Usage Guide