@amitdeshmukh/ax-crew 8.0.3 → 8.1.0

package/.claude/settings.local.json

@@ -1,7 +1,8 @@
 {
   "permissions": {
     "allow": [
-      "Bash(npm test:*)"
+      "Bash(npm test:*)",
+      "Bash(git push:*)"
     ]
   }
 }

package/CHANGELOG.md

@@ -1,5 +1,29 @@
 # Changelog
 
+## [8.1.0] - 2026-03-04
+
+### Added
+- Per-agent execution engine selection via `executionMode: "axagent" | "axgen"` with internal routing.
+- `axAgentOptions` support to configure modern AxAgent RLM capabilities from AxCrew.
+- RLM examples:
+  - `examples/rlm-long-task.ts` (long task + context management + cost output)
+  - `examples/rlm-shared-fields.ts` (shared fields + sub-agent propagation + cost output)
+- New test coverage for execution-mode metrics parity: `tests/execution-mode-metrics.test.ts`.
+
+### Changed
+- Preserved external invocation API (`forward` / `streamingForward`) while routing internally by `executionMode`.
+- Updated peer dependency minimums to `@ax-llm/ax@^19.0.11` and `@ax-llm/ax-tools@^19.0.11`.
+- Updated README and docs to explain:
+  - why AxAgent changed (RLM split architecture),
+  - mandatory RLM options (`axAgentOptions.runtime`, `axAgentOptions.contextFields`),
+  - metrics-first cost reporting terminology.
+
+### Fixed
+- `examples/graphjin-database-agent.ts` now:
+  - initializes agents inside `main()` for proper error handling,
+  - always calls `crew.destroy()` in `finally`,
+  - uses `answer` output consistently for `ManagerAgent`.
+
 ## [8.0.0] - 2025-01-17
 
 ### Added
@@ -320,4 +344,3 @@ v2.0.0.html).
 
 ### Fixed
 - Updates to [dateTime.ts](src/dateTime.ts) to enable use in Gemini models
-

package/README.md

@@ -50,6 +50,7 @@ const config = {
   crew: [{
     name: "Planner",
     description: "Creates a plan to complete a task",
+    executionMode: "axagent", // "axagent" | "axgen"
     signature: "task:string \"a task to be completed\" -> plan:string \"a plan to execute the task\"",
     provider: "google-gemini",
     providerKeyName: "GEMINI_API_KEY",
@@ -126,6 +127,7 @@ Key TypeScript features:
 - **Functions (tools)**: Register callable functions via a registry and reference by name in agent `functions`.
 - **State**: `crew.state.set/get/getAll()` shared across all agents.
 - **Persona**: Use `definition` (preferred) or `prompt` to set the system program. If both are present, `definition` wins.
+- **Execution mode**: Set `executionMode` to `axagent` (default) or `axgen` per agent.
 - **Streaming**: Use `streamingForward()` for token streams.
 - **Metrics**: Per‑agent `getMetrics()` + crew‑level `getCrewMetrics()` snapshots.
 
@@ -184,6 +186,103 @@ Add either field to any agent config. The chosen value becomes the Ax agent's un
 }
 ```
 
+### Execution Modes (`axagent` | `axgen`)
+
+Each agent can run in one of two execution modes supported by AxLLM:
+
+- `axagent` (default): Uses AxAgent capabilities.
+- `axgen`: Uses AxGen capabilities.
+
+Set mode in config:
+
+```json
+{
+  "name": "Planner",
+  "description": "Creates plans",
+  "executionMode": "axgen",
+  "signature": "task:string -> plan:string",
+  "provider": "google-gemini",
+  "providerKeyName": "GEMINI_API_KEY",
+  "ai": { "model": "gemini-1.5-flash", "temperature": 0 }
+}
+```
+
+Execution is routed internally by `executionMode` while you continue to call the same public APIs (`forward` / `streamingForward`).
+There is no external API split and no caller-side branching for `axagent` vs `axgen`.
+
+### AxAgent RLM Support
+
+`AxAgent` in newer Ax versions moved to an RLM split architecture (Actor/Responder + runtime loop).
+AxCrew supports this directly so you can use modern AxAgent capabilities without changing the AxCrew call surface.
+
+To configure those capabilities in AxCrew, use `axAgentOptions` on an agent (effective when `executionMode` is `axagent`):
+
+```typescript
+const config = {
+  crew: [{
+    name: "Researcher",
+    description: "Deep research agent",
+    executionMode: "axagent",
+    signature: "query:string, context:string? -> answer:string",
+    provider: "google-gemini",
+    providerKeyName: "GEMINI_API_KEY",
+    ai: { model: "gemini-1.5-pro", temperature: 0 },
+    axAgentOptions: {
+      runtime,
+      contextFields: ["context"],
+      mode: "simple",
+      maxTurns: 12
+    }
+  }]
+};
+```
+
+RLM mode requirements:
+- `axAgentOptions.runtime`: Provide an `AxJSRuntime` instance for runtime execution.
+- `axAgentOptions.contextFields`: Required in RLM mode (can be an empty array when no context fields are needed).
+
+Example runtime wiring:
+
+```typescript
+import { AxJSRuntime, AxJSRuntimePermission } from '@ax-llm/ax';
+
+const runtime = new AxJSRuntime({
+  permissions: [AxJSRuntimePermission.TIMING],
+});
+
+const config = {
+  crew: [{
+    name: "Analyzer",
+    executionMode: "axagent",
+    // ...
+    axAgentOptions: {
+      runtime,
+      contextFields: ["context"]
+    }
+  }]
+};
+```
+
+Why this exists:
+- Upstream AxAgent changed to support richer RLM behaviors (runtime-backed decomposition, actor/responder separation, context management).
+- AxCrew needs to preserve compatibility while exposing these capabilities.
+- `executionMode` keeps this explicit per-agent and lets AxCrew route internally with zero external API change.
+
+`forward()` and `streamingForward()` remain unchanged; AxCrew routes internally based on `executionMode`.
+
+### RLM Examples (with cost and tracing verification)
+
+The following examples in this repo demonstrate equivalent RLM behavior in AxCrew and print accumulated costs at the end:
+
+- [`examples/rlm-long-task.ts`](examples/rlm-long-task.ts): Long-task analysis with context management and RLM runtime.
+- [`examples/rlm-shared-fields.ts`](examples/rlm-shared-fields.ts): Shared-field propagation with parent/child AxAgent setup.
+
+Both examples:
+- run through the same public `forward()` API
+- use AxAgent RLM options through `axAgentOptions`
+- print per-agent and crew metrics (`estimatedCostUSD`, token usage)
+- are compatible with telemetry/tracing instrumentation
+
 ### Agent Examples
 You can provide examples to guide the behavior of your agents using the `examples` field in the agent configuration. Examples help the agent understand the expected input/output format and improve its responses.
 
@@ -236,7 +335,7 @@ Here's an example of how to set up the `FunctionRegistry` with built-in function
 
 ```javascript
 import { AxCrewFunctions } from '@amitdeshmukh/ax-crew';
-const crew = new AxCrew(configFilePath, AxCrewFunctions);
+const crew = new AxCrew(config, AxCrewFunctions);
 ```
 
 if you want to bring your own functions, you can do so by creating a new instance of `FunctionRegistry` and passing it to the `AxCrew` constructor.
@@ -248,7 +347,7 @@ const myFunctions: FunctionRegistryType = {
   GoogleSearch: googleSearchInstance.toFunction()
 };
 
-const crew = new AxCrew(configFilePath, myFunctions);
+const crew = new AxCrew(config, myFunctions);
 ```
 
 ### Adding Agents to the Crew
@@ -368,7 +467,7 @@ import { AxCrew, AxCrewFunctions } from '@amitdeshmukh/ax-crew';
 
 // Create a new instance of AxCrew
 const crew = new AxCrew(config, AxCrewFunctions);
-crew.addAgentsToCrew(['Planner', 'Calculator', 'Manager']);
+await crew.addAgentsToCrew(['Planner', 'Calculator', 'Manager']);
 
 // Get agent instances
 const Planner = crew.agents.get("Planner");
@@ -714,7 +813,7 @@ crew.resetCosts();
 ```
 
 Notes:
-- Legacy cost APIs (`getLastUsageCost`, `getAccumulatedCosts`, `getAggregatedCosts`) are superseded by metrics methods.
+- Legacy cost APIs (`getLastUsageCost`, `getAccumulatedCosts`, `getAggregatedCosts`) are superseded by metrics APIs.
 - Estimated cost values are numbers rounded to 5 decimal places.
 
 ### Telemetry Support (OpenTelemetry)
@@ -941,14 +1040,14 @@ console.log(playbook);
 |----------|-------------|
 | `"all"` | Apply feedback to all agents involved in the task |
 | `"primary"` | Apply only to the primary (entry) agent |
-| `"leaf"` | Apply only to leaf agents (no sub-agents) |
+| `"weighted"` | Apply to all involved agents (reserved for weighted scoring flows) |
 
 #### Examples
 
 See the ACE examples for complete demonstrations:
 
 - [`ace-customer-support.ts`](examples/ace-customer-support.ts) - Learn edge-case handling beyond standard policies
-- [`ace-feedback-routing.ts`](examples/ace-feedback-routing.ts) - Flight assistant with preference learning
+- [`ace-flight-finder.ts`](examples/ace-flight-finder.ts) - Flight assistant with preference learning
 
 ```bash
 # Run the customer support demo
@@ -957,4 +1056,4 @@ npx tsx examples/ace-customer-support.ts
 
 ## Changelog
 
-See [CHANGELOG.md](CHANGELOG.md) for a list of changes and version updates.
+See [CHANGELOG.md](CHANGELOG.md) for a list of changes and version updates.

package/dist/agents/agentConfig.d.ts

@@ -1,6 +1,6 @@
 import { AxDefaultCostTracker } from '@ax-llm/ax';
 import type { AxFunction } from '@ax-llm/ax';
-import type { AgentConfig, AxCrewConfig, AxCrewOptions, FunctionRegistryType, MCPTransportConfig, MCPStdioTransportConfig, MCPHTTPSSETransportConfig, MCPStreamableHTTPTransportConfig } from '../types.js';
+import type { AgentConfig, AgentExecutionMode, AxCrewAxAgentOptions, AxCrewConfig, AxCrewOptions, FunctionRegistryType, MCPTransportConfig, MCPStdioTransportConfig, MCPHTTPSSETransportConfig, MCPStreamableHTTPTransportConfig } from '../types.js';
 export declare function isStdioTransport(config: MCPTransportConfig): config is MCPStdioTransportConfig;
 export declare function isHTTPSSETransport(config: MCPTransportConfig): config is MCPHTTPSSETransportConfig;
 export declare function isStreambleHTTPTransport(config: MCPTransportConfig): config is MCPStreamableHTTPTransportConfig;
@@ -28,6 +28,8 @@ declare const parseCrewConfig: (input: AxCrewConfig) => {
 declare const parseAgentConfig: (agentName: string, crewConfig: AxCrewConfig, functions: FunctionRegistryType, state: Record<string, any>, options?: AxCrewOptions) => Promise<{
     ai: import("@ax-llm/ax").AxAI<string>;
     name: string;
+    executionMode: AgentExecutionMode;
+    axAgentOptions: AxCrewAxAgentOptions;
     description: string;
     definition: any;
     signature: string | import("@ax-llm/ax").AxSignature<Record<string, any>, Record<string, any>>;

package/dist/agents/agentConfig.js

@@ -174,10 +174,13 @@ const parseAgentConfig = async (agentName, crewConfig, functions, state, options
             // Add MCP functions to functions
             ...mcpFunctions
         ];
+        const executionMode = agentConfigData.executionMode === 'axgen' ? 'axgen' : 'axagent';
         // Return AI instance and Agent parameters
         return {
             ai: aiInstance,
             name: agentName,
+            executionMode,
+            axAgentOptions: (agentConfigData.axAgentOptions ?? {}),
             description: agentConfigData.description,
             definition: agentConfigData.definition ?? agentConfigData.prompt,
             signature: agentConfigData.signature,

package/dist/agents/index.d.ts

@@ -1,22 +1,26 @@
 import { AxAgent, AxAI } from "@ax-llm/ax";
 import type { AxSignature, AxAgentic, AxFunction, AxProgramForwardOptions, AxProgramStreamingForwardOptions, AxGenStreamingOut } from "@ax-llm/ax";
-import type { StateInstance, FunctionRegistryType, UsageCost, AxCrewConfig, AxCrewOptions, MCPTransportConfig, ACEConfig } from "../types.js";
+import type { StateInstance, FunctionRegistryType, UsageCost, AxCrewConfig, AxCrewOptions, MCPTransportConfig, ACEConfig, AgentExecutionMode, AxCrewAxAgentOptions } from "../types.js";
 declare class StatefulAxAgent extends AxAgent<any, any> {
     state: StateInstance;
     axai: any;
     private agentName;
+    private agentDefinition;
+    private executionMode;
+    private axGenProgram;
     private costTracker?;
-    private lastRecordedCostUSD;
     private debugEnabled;
+    private static readonly modernAxAgentRuntime;
     private aceConfig?;
     private aceOptimizer?;
     private acePlaybook?;
     private aceBaseInstruction?;
     private isAxAIService;
-    private isAxAIInstance;
     constructor(ai: AxAI, options: Readonly<{
         name: string;
         description: string;
+        executionMode?: AgentExecutionMode;
+        axAgentOptions?: AxCrewAxAgentOptions;
         definition?: string;
         signature: string | AxSignature;
         agents?: AxAgentic<any, any>[] | undefined;
@@ -25,8 +29,28 @@ declare class StatefulAxAgent extends AxAgent<any, any> {
         mcpServers?: Record<string, MCPTransportConfig> | undefined;
         debug?: boolean;
     }>, state: StateInstance);
+    /**
+     * @deprecated Use setExamplesCompat() to avoid Ax runtime version coupling.
+     */
+    setExamples(examples: Readonly<Array<Record<string, any>>>): void;
+    setExamplesCompat(examples: Readonly<Array<Record<string, any>>>): void;
+    /**
+     * @deprecated Use setDescriptionCompat() to avoid Ax runtime version coupling.
+     */
+    setDescription(description: string): void;
+    setDescriptionCompat(description: string): void;
+    getUsage(): (import("@ax-llm/ax").AxModelUsage & {
+        ai: string;
+        model: string;
+    })[];
+    resetUsage(): void;
+    private resolveInvocationArgs;
+    private executeForwardByMode;
+    private recordUsageMetrics;
+    private runForwardInvocation;
     forward(values: Record<string, any>, options?: Readonly<AxProgramForwardOptions<any>>): Promise<Record<string, any>>;
     forward(ai: AxAI, values: Record<string, any>, options?: Readonly<AxProgramForwardOptions<any>>): Promise<Record<string, any>>;
+    private runStreamingInvocation;
     streamingForward(values: Record<string, any>, options?: Readonly<AxProgramStreamingForwardOptions<any>>): AxGenStreamingOut<any>;
     streamingForward(ai: AxAI, values: Record<string, any>, options?: Readonly<AxProgramStreamingForwardOptions<any>>): AxGenStreamingOut<any>;
     getLastUsageCost(): UsageCost | null;