Hay dos enfoques generales para desarrollar aplicaciones de agentes con IA.
-
Flujos de trabajo deterministas : el código define el flujo de control. Puede escribir la secuencia de pasos, bifurcación, paralelismo y control de errores mediante construcciones de programación estándar. El LLM realiza trabajo dentro de cada paso, pero no controla el flujo general.
-
Flujos de trabajo dirigidos por agente (bucles de agente): LLM controla el flujo de control. El agente decide qué herramientas llamar, en qué orden y cuándo se completa la tarea. Proporciona herramientas e instrucciones, pero el agente determina la ruta de ejecución en tiempo de ejecución.
Ambos enfoques se benefician de la ejecución duradera y se pueden implementar mediante el modelo de programación durable Task. En este artículo se muestra cómo compilar cada patrón mediante ejemplos de código.
Sugerencia
Estos patrones se alinean con los diseños de flujo de trabajo agente descritos en Building Effective Agents de Anthropic. El modelo de programación Durable Task se asigna naturalmente a estos patrones: las orquestaciones definen el control del flujo de trabajo y se guardan automáticamente, mientras que las actividades encapsulan operaciones no determinísticas como las llamadas LLM, invocaciones de herramientas y solicitudes de API.
Elección de un enfoque
La tabla siguiente le ayuda a decidir cuándo usar cada enfoque.
| Use flujos de trabajo deterministas cuando... |
Use los bucles de agente cuando... |
| La secuencia de pasos se conoce con antelación. |
La tarea está abierta y no se pueden predecir los pasos. |
| Necesita límites de protección explícitos sobre el comportamiento del agente. |
Quiere que el LLM decida qué herramientas usar y cuándo. |
| El cumplimiento o la auditoría requieren un flujo de control revisable. |
El agente debe adaptar su enfoque en función de los resultados intermedios. |
| Quiere combinar varios marcos de inteligencia artificial en un único flujo de trabajo. |
Va a crear un agente conversacional con funcionalidades de llamada a herramientas. |
Ambos enfoques proporcionan puntos de comprobación automáticos, políticas de reintento, escalado distribuido y soporte humano en el bucle a través de una ejecución duradera.
Patrones de flujo de trabajo deterministas
En un flujo de trabajo determinista, el código controla la ruta de ejecución. El LLM es llamado como un paso dentro del flujo de trabajo, pero no decide lo que sucede a continuación. El modelo de programación Durable Task se asigna de forma natural a este enfoque.
-
Las orquestaciones definen el control del flujo de trabajo (secuencia, bifurcación, paralelismo, manejo de errores) y se registran automáticamente.
-
Las actividades encapsulan operaciones no deterministas, como llamadas LLM, invocaciones de herramientas y solicitudes de API. Las actividades se pueden ejecutar en cualquier instancia de proceso disponible.
En los ejemplos siguientes se usa Durable Functions, que se ejecuta en Azure Functions con hospedaje sin servidor.
En los ejemplos siguientes se usan los SDK portátiles de Durable Task, que se ejecutan en cualquier proceso host, incluidos Azure Container Apps, Kubernetes, máquinas virtuales o localmente.
Encadenamiento de mensajes
El encadenamiento de mensajes es el patrón agencial más sencillo. Se divide una tarea compleja en una serie de interacciones LLM secuenciales, donde la salida de cada paso se convierte en la entrada del siguiente paso. Dado que cada llamada de actividad se controla automáticamente, un bloqueo a mitad de la canalización no le obliga a reiniciar desde cero y volver a consumir tokens LLM costosos: la ejecución se reanuda desde el último paso completado.
También puede insertar puertas de validación mediante programación entre pasos. Por ejemplo, después de generar un esquema, puede comprobar que cumple una restricción de longitud o tema antes de pasarla al paso de borrador.
Este patrón se asigna directamente al patrón de encadenamiento de funciones en el modelo de programación Durable Task.
Cuándo usar: Canalizaciones de generación de contenido, procesamiento de documentos de varios pasos, enriquecimiento secuencial de datos, flujos de trabajo que requieren puertas de validación intermedias.
[Function(nameof(PromptChainingOrchestration))]
public async Task<string> PromptChainingOrchestration(
[OrchestrationTrigger] TaskOrchestrationContext context)
{
var topic = context.GetInput<string>();
// Step 1: Generate research outline
string outline = await context.CallActivityAsync<string>(
nameof(GenerateOutlineAgent), topic);
// Step 2: Write first draft from outline
string draft = await context.CallActivityAsync<string>(
nameof(WriteDraftAgent), outline);
// Step 3: Refine and polish the draft
string finalContent = await context.CallActivityAsync<string>(
nameof(RefineDraftAgent), draft);
return finalContent;
}
Nota:
El estado de la orquestación se controla automáticamente en cada await instrucción. Si el proceso de host se bloquea o la máquina virtual se reinicia, la orquestación se reanudará automáticamente desde el último paso completado en lugar de empezar de nuevo.
@app.orchestration_trigger(context_name="context")
def prompt_chaining_orchestration(context: df.DurableOrchestrationContext):
topic = context.get_input()
# Step 1: Generate research outline
outline = yield context.call_activity("generate_outline_agent", topic)
# Step 2: Write first draft from outline
draft = yield context.call_activity("write_draft_agent", outline)
# Step 3: Refine and polish the draft
final_content = yield context.call_activity("refine_draft_agent", draft)
return final_content
Nota:
El estado de la orquestación se controla automáticamente en cada yield instrucción. Si el proceso de host se bloquea o la máquina virtual se recicla, la orquestación se reanudará automáticamente desde el último paso completado en lugar de iniciarse desde el principio.
const df = require("durable-functions");
df.app.orchestration("promptChainingOrchestration", function* (context) {
const topic = context.df.getInput();
// Step 1: Generate research outline
const outline = yield context.df.callActivity("generateOutlineAgent", topic);
// Step 2: Write first draft from outline
const draft = yield context.df.callActivity("writeDraftAgent", outline);
// Step 3: Refine and polish the draft
const finalContent = yield context.df.callActivity("refineDraftAgent", draft);
return finalContent;
});
Nota:
El estado de la orquestación se controla automáticamente en cada yield instrucción. Si el proceso anfitrión se bloquea o la máquina virtual se recicla, la orquestación se reanudará automáticamente desde el último paso completado en lugar de comenzar de nuevo.
@FunctionName("PromptChainingOrchestration")
public String promptChainingOrchestration(
@DurableOrchestrationTrigger(name = "ctx") TaskOrchestrationContext ctx) {
String topic = ctx.getInput(String.class);
// Step 1: Generate research outline
String outline = ctx.callActivity(
"GenerateOutlineAgent", topic, String.class).await();
// Step 2: Write first draft from outline
String draft = ctx.callActivity(
"WriteDraftAgent", outline, String.class).await();
// Step 3: Refine and polish the draft
String finalContent = ctx.callActivity(
"RefineDraftAgent", draft, String.class).await();
return finalContent;
}
Nota:
El estado de la orquestación se controla automáticamente en cada await() invocación. Si el proceso anfitrión se bloquea o la máquina virtual se recicla, la orquestación se reanudará automáticamente desde el último paso completado en lugar de comenzar de nuevo.
[DurableTask]
public class PromptChainingOrchestration : TaskOrchestrator<string, string>
{
public override async Task<string> RunAsync(
TaskOrchestrationContext context, string topic)
{
// Step 1: Generate research outline
string outline = await context.CallActivityAsync<string>(
nameof(GenerateOutlineAgent), topic);
// Step 2: Write first draft from outline
string draft = await context.CallActivityAsync<string>(
nameof(WriteDraftAgent), outline);
// Step 3: Refine and polish the draft
string finalContent = await context.CallActivityAsync<string>(
nameof(RefineDraftAgent), draft);
return finalContent;
}
}
Nota:
El estado de la orquestación se controla automáticamente en cada await instrucción. Si el proceso anfitrión se bloquea o la máquina virtual se recicla, la orquestación se reanudará automáticamente desde el último paso completado en lugar de comenzar de nuevo.
def prompt_chaining_orchestration(ctx: task.OrchestrationContext, topic: str) -> str:
# Step 1: Generate research outline
outline = yield ctx.call_activity(generate_outline_agent, input=topic)
# Step 2: Write first draft from outline
draft = yield ctx.call_activity(write_draft_agent, input=outline)
# Step 3: Refine and polish the draft
final_content = yield ctx.call_activity(refine_draft_agent, input=draft)
return final_content
Nota:
El estado de la orquestación se controla automáticamente en cada yield instrucción. Si el proceso anfitrión se bloquea o la máquina virtual se recicla, la orquestación se reanudará automáticamente desde el último paso completado en lugar de comenzar de nuevo.
const promptChainingOrchestration: TOrchestrator = async function* (
ctx: OrchestrationContext, topic: string): any {
// Step 1: Generate research outline
const outline: string = yield ctx.callActivity(generateOutlineAgent, topic);
// Step 2: Write first draft from outline
const draft: string = yield ctx.callActivity(writeDraftAgent, outline);
// Step 3: Refine and polish the draft
const finalContent: string = yield ctx.callActivity(refineDraftAgent, draft);
return finalContent;
};
Nota:
El estado de la orquestación se controla automáticamente en cada yield instrucción. Si el proceso anfitrión se bloquea o la máquina virtual se recicla, la orquestación se reanudará automáticamente desde el último paso completado en lugar de comenzar de nuevo.
ctx -> {
String topic = ctx.getInput(String.class);
// Step 1: Generate research outline
String outline = ctx.callActivity(
"GenerateOutlineAgent", topic, String.class).await();
// Step 2: Write first draft from outline
String draft = ctx.callActivity(
"WriteDraftAgent", outline, String.class).await();
// Step 3: Refine and polish the draft
String finalContent = ctx.callActivity(
"RefineDraftAgent", draft, String.class).await();
ctx.complete(finalContent);
}
Nota:
El estado de la orquestación se controla automáticamente en cada await() invocación. Si el proceso anfitrión se bloquea o la máquina virtual se recicla, la orquestación se reanudará automáticamente desde el último paso completado en lugar de comenzar de nuevo.
Enrutamiento
El enrutamiento utiliza un paso de clasificación para determinar qué agente o modelo posterior debe gestionar una solicitud. La orquestación llama primero a una actividad clasificadora y, a continuación, se bifurca en el controlador adecuado en función del resultado. Esta enfoque le permite adaptar el mensaje, modelo, y conjunto de herramientas de cada controlador de forma independiente, por ejemplo, dirigiendo las preguntas de facturación a un agente especializado con acceso a las API de pago mientras se envían preguntas generales a un modelo más ligero.
Cuándo usar: Evaluación de prioridades de atención al cliente, clasificación de intenciones a agentes especializados, selección dinámica de modelos en función de la complejidad de la tarea.
[Function(nameof(RoutingOrchestration))]
public async Task<string> RoutingOrchestration(
[OrchestrationTrigger] TaskOrchestrationContext context)
{
var request = context.GetInput<SupportRequest>();
// Classify the request type
string category = await context.CallActivityAsync<string>(
nameof(ClassifyRequestAgent), request.Message);
// Route to the appropriate specialized agent
return category switch
{
"billing" => await context.CallActivityAsync<string>(
nameof(BillingAgent), request),
"technical" => await context.CallActivityAsync<string>(
nameof(TechnicalSupportAgent), request),
"general" => await context.CallActivityAsync<string>(
nameof(GeneralInquiryAgent), request),
_ => await context.CallActivityAsync<string>(
nameof(GeneralInquiryAgent), request),
};
}
@app.orchestration_trigger(context_name="context")
def routing_orchestration(context: df.DurableOrchestrationContext):
request = context.get_input()
# Classify the request type
category = yield context.call_activity("classify_request_agent", request["message"])
# Route to the appropriate specialized agent
if category == "billing":
return (yield context.call_activity("billing_agent", request))
elif category == "technical":
return (yield context.call_activity("technical_support_agent", request))
else:
return (yield context.call_activity("general_inquiry_agent", request))
const df = require("durable-functions");
df.app.orchestration("routingOrchestration", function* (context) {
const request = context.df.getInput();
// Classify the request type
const category = yield context.df.callActivity("classifyRequestAgent", request.message);
// Route to the appropriate specialized agent
switch (category) {
case "billing":
return yield context.df.callActivity("billingAgent", request);
case "technical":
return yield context.df.callActivity("technicalSupportAgent", request);
default:
return yield context.df.callActivity("generalInquiryAgent", request);
}
});
@FunctionName("RoutingOrchestration")
public String routingOrchestration(
@DurableOrchestrationTrigger(name = "ctx") TaskOrchestrationContext ctx) {
SupportRequest request = ctx.getInput(SupportRequest.class);
// Classify the request type
String category = ctx.callActivity(
"ClassifyRequestAgent", request.getMessage(), String.class).await();
// Route to the appropriate specialized agent
return switch (category) {
case "billing" -> ctx.callActivity(
"BillingAgent", request, String.class).await();
case "technical" -> ctx.callActivity(
"TechnicalSupportAgent", request, String.class).await();
default -> ctx.callActivity(
"GeneralInquiryAgent", request, String.class).await();
};
}
[DurableTask]
public class RoutingOrchestration : TaskOrchestrator<SupportRequest, string>
{
public override async Task<string> RunAsync(
TaskOrchestrationContext context, SupportRequest request)
{
// Classify the request type
string category = await context.CallActivityAsync<string>(
nameof(ClassifyRequestAgent), request.Message);
// Route to the appropriate specialized agent
return category switch
{
"billing" => await context.CallActivityAsync<string>(
nameof(BillingAgent), request),
"technical" => await context.CallActivityAsync<string>(
nameof(TechnicalSupportAgent), request),
_ => await context.CallActivityAsync<string>(
nameof(GeneralInquiryAgent), request),
};
}
}
def routing_orchestration(ctx: task.OrchestrationContext, request: dict) -> str:
# Classify the request type
category = yield ctx.call_activity(classify_request_agent, input=request["message"])
# Route to the appropriate specialized agent
if category == "billing":
return (yield ctx.call_activity(billing_agent, input=request))
elif category == "technical":
return (yield ctx.call_activity(technical_support_agent, input=request))
else:
return (yield ctx.call_activity(general_inquiry_agent, input=request))
const routingOrchestration: TOrchestrator = async function* (
ctx: OrchestrationContext, request: SupportRequest): any {
// Classify the request type
const category: string = yield ctx.callActivity(classifyRequestAgent, request.message);
// Route to the appropriate specialized agent
switch (category) {
case "billing":
return yield ctx.callActivity(billingAgent, request);
case "technical":
return yield ctx.callActivity(technicalSupportAgent, request);
default:
return yield ctx.callActivity(generalInquiryAgent, request);
}
};
ctx -> {
SupportRequest request = ctx.getInput(SupportRequest.class);
// Classify the request type
String category = ctx.callActivity(
"ClassifyRequestAgent", request.getMessage(), String.class).await();
// Route to the appropriate specialized agent
String result = switch (category) {
case "billing" -> ctx.callActivity(
"BillingAgent", request, String.class).await();
case "technical" -> ctx.callActivity(
"TechnicalSupportAgent", request, String.class).await();
default -> ctx.callActivity(
"GeneralInquiryAgent", request, String.class).await();
};
ctx.complete(result);
}
Paralelización
Cuando tenga varias subtareas independientes, puede enviarlas como llamadas de actividad paralelas y esperar todos los resultados antes de continuar. Durable Task Scheduler distribuye estas actividades automáticamente en todas las instancias de proceso disponibles, lo que significa que agregar más trabajos reduce directamente el tiempo total del reloj.
Una variante común es la votación multimodelo: se envía el mismo mensaje a varios modelos (o el mismo modelo con temperaturas diferentes) en paralelo y, a continuación, agrega o selecciona entre las respuestas. Dado que cada rama paralela está en puntos de control de forma independiente, un error transitorio en una rama no afecta a los demás.
Este patrón se asigna directamente al patrón fan-out/fan-in en Durable Task.
Cuándo usar: Análisis por lotes de documentos, llamadas a herramientas paralelas, evaluación multimodelo, moderación de contenido con varios revisores.
[Function(nameof(ParallelResearchOrchestration))]
public async Task<string> ParallelResearchOrchestration(
[OrchestrationTrigger] TaskOrchestrationContext context)
{
var request = context.GetInput<ResearchRequest>();
// Fan-out: research multiple subtopics in parallel
var researchTasks = request.Subtopics
.Select(subtopic => context.CallActivityAsync<string>(
nameof(ResearchSubtopicAgent), subtopic))
.ToList();
string[] researchResults = await Task.WhenAll(researchTasks);
// Aggregate: synthesize all research into a single summary
string summary = await context.CallActivityAsync<string>(
nameof(SynthesizeAgent),
new { request.Topic, Research = researchResults });
return summary;
}
@app.orchestration_trigger(context_name="context")
def parallel_research_orchestration(context: df.DurableOrchestrationContext):
request = context.get_input()
# Fan-out: research multiple subtopics in parallel
research_tasks = []
for subtopic in request["subtopics"]:
research_tasks.append(
context.call_activity("research_subtopic_agent", subtopic)
)
research_results = yield context.task_all(research_tasks)
# Aggregate: synthesize all research into a single summary
summary = yield context.call_activity("synthesize_agent", {
"topic": request["topic"],
"research": research_results
})
return summary
const df = require("durable-functions");
df.app.orchestration("parallelResearchOrchestration", function* (context) {
const request = context.df.getInput();
// Fan-out: research multiple subtopics in parallel
const tasks = request.subtopics.map((subtopic) =>
context.df.callActivity("researchSubtopicAgent", subtopic)
);
const researchResults = yield context.df.Task.all(tasks);
// Aggregate: synthesize all research into a single summary
const summary = yield context.df.callActivity("synthesizeAgent", {
topic: request.topic,
research: researchResults,
});
return summary;
});
@FunctionName("ParallelResearchOrchestration")
public String parallelResearchOrchestration(
@DurableOrchestrationTrigger(name = "ctx") TaskOrchestrationContext ctx) {
ResearchRequest request = ctx.getInput(ResearchRequest.class);
// Fan-out: research multiple subtopics in parallel
List<Task<String>> tasks = request.getSubtopics().stream()
.map(subtopic -> ctx.callActivity(
"ResearchSubtopicAgent", subtopic, String.class))
.collect(Collectors.toList());
List<String> researchResults = ctx.allOf(tasks).await();
// Aggregate: synthesize all research into a single summary
String summary = ctx.callActivity(
"SynthesizeAgent", researchResults, String.class).await();
return summary;
}
[DurableTask]
public class ParallelResearchOrchestration : TaskOrchestrator<ResearchRequest, string>
{
public override async Task<string> RunAsync(
TaskOrchestrationContext context, ResearchRequest request)
{
// Fan-out: research multiple subtopics in parallel
var researchTasks = request.Subtopics
.Select(subtopic => context.CallActivityAsync<string>(
nameof(ResearchSubtopicAgent), subtopic))
.ToList();
string[] researchResults = await Task.WhenAll(researchTasks);
// Aggregate: synthesize all research into a single summary
string summary = await context.CallActivityAsync<string>(
nameof(SynthesizeAgent),
new { request.Topic, Research = researchResults });
return summary;
}
}
def parallel_research_orchestration(ctx: task.OrchestrationContext, request: dict) -> str:
# Fan-out: research multiple subtopics in parallel
research_tasks = []
for subtopic in request["subtopics"]:
research_tasks.append(
ctx.call_activity(research_subtopic_agent, input=subtopic)
)
research_results = yield task.when_all(research_tasks)
# Aggregate: synthesize all research into a single summary
summary = yield ctx.call_activity(synthesize_agent, input={
"topic": request["topic"],
"research": research_results
})
return summary
const parallelResearchOrchestration: TOrchestrator = async function* (
ctx: OrchestrationContext,
request: { topic: string; subtopics: string[] }): any {
// Fan-out: research multiple subtopics in parallel
const tasks = request.subtopics.map((subtopic) =>
ctx.callActivity(researchSubtopicAgent, subtopic)
);
const researchResults: string[] = yield whenAll(tasks);
// Aggregate: synthesize all research into a single summary
const summary: string = yield ctx.callActivity(synthesizeAgent, {
topic: request.topic,
research: researchResults,
});
return summary;
};
ctx -> {
ResearchRequest request = ctx.getInput(ResearchRequest.class);
// Fan-out: research multiple subtopics in parallel
List<Task<String>> tasks = request.getSubtopics().stream()
.map(subtopic -> ctx.callActivity(
"ResearchSubtopicAgent", subtopic, String.class))
.collect(Collectors.toList());
List<String> researchResults = ctx.allOf(tasks).await();
// Aggregate: synthesize all research into a single summary
String summary = ctx.callActivity(
"SynthesizeAgent", researchResults, String.class).await();
ctx.complete(summary);
}
Trabajadores del Orquestador
En este patrón, un orquestador central llama primero a un LLM (a través de una actividad) para planificar el trabajo. En función de la salida de LLM, el orquestador determina a continuación qué subtareas se necesitan. A continuación, el orquestador envía esas subtareas a orquestaciones de trabajo especializadas. La diferencia clave de paralelización es que el conjunto de subtareas no se fija en tiempo de diseño; el orquestador los determina dinámicamente en tiempo de ejecución.
Este patrón usa suborquestaciones, que son subflujos de trabajo con puntos de control independientes. Cada orquestación de trabajo puede contener varios pasos, reintentos y paralelismo anidado.
Cuándo usar: Procesos de investigación profunda, flujos de trabajo de agentes de codificación que modifican varios archivos, colaboración entre múltiples agentes donde cada uno tiene un rol distinto.
[Function(nameof(OrchestratorWorkersOrchestration))]
public async Task<string> OrchestratorWorkersOrchestration(
[OrchestrationTrigger] TaskOrchestrationContext context)
{
var request = context.GetInput<ResearchRequest>();
// Central orchestrator: determine what research is needed
string[] subtasks = await context.CallActivityAsync<string[]>(
nameof(PlanResearchAgent), request.Topic);
// Delegate to worker orchestrations in parallel
var workerTasks = subtasks
.Select(subtask => context.CallSubOrchestratorAsync<string>(
nameof(ResearchWorkerOrchestration), subtask))
.ToList();
string[] results = await Task.WhenAll(workerTasks);
// Synthesize results
string finalReport = await context.CallActivityAsync<string>(
nameof(SynthesizeAgent),
new { request.Topic, Research = results });
return finalReport;
}
@app.orchestration_trigger(context_name="context")
def orchestrator_workers_orchestration(context: df.DurableOrchestrationContext):
request = context.get_input()
# Central orchestrator: determine what research is needed
subtasks = yield context.call_activity("plan_research_agent", request["topic"])
# Delegate to worker orchestrations in parallel
worker_tasks = []
for subtask in subtasks:
worker_tasks.append(
context.call_sub_orchestrator("research_worker_orchestration", subtask)
)
results = yield context.task_all(worker_tasks)
# Synthesize results
final_report = yield context.call_activity("synthesize_agent", {
"topic": request["topic"],
"research": results
})
return final_report
const df = require("durable-functions");
df.app.orchestration("orchestratorWorkersOrchestration", function* (context) {
const request = context.df.getInput();
// Central orchestrator: determine what research is needed
const subtasks = yield context.df.callActivity("planResearchAgent", request.topic);
// Delegate to worker orchestrations in parallel
const workerTasks = subtasks.map((subtask) =>
context.df.callSubOrchestrator("researchWorkerOrchestration", subtask)
);
const results = yield context.df.Task.all(workerTasks);
// Synthesize results
const finalReport = yield context.df.callActivity("synthesizeAgent", {
topic: request.topic,
research: results,
});
return finalReport;
});
@FunctionName("OrchestratorWorkersOrchestration")
public String orchestratorWorkersOrchestration(
@DurableOrchestrationTrigger(name = "ctx") TaskOrchestrationContext ctx) {
ResearchRequest request = ctx.getInput(ResearchRequest.class);
// Central orchestrator: determine what research is needed
List<String> subtasks = ctx.callActivity(
"PlanResearchAgent", request.getTopic(), List.class).await();
// Delegate to worker orchestrations in parallel
List<Task<String>> workerTasks = subtasks.stream()
.map(subtask -> ctx.callSubOrchestrator(
"ResearchWorkerOrchestration", subtask, String.class))
.collect(Collectors.toList());
List<String> results = ctx.allOf(workerTasks).await();
// Synthesize results
String finalReport = ctx.callActivity(
"SynthesizeAgent", results, String.class).await();
return finalReport;
}
[DurableTask]
public class OrchestratorWorkersOrchestration : TaskOrchestrator<ResearchRequest, string>
{
public override async Task<string> RunAsync(
TaskOrchestrationContext context, ResearchRequest request)
{
// Central orchestrator: determine what research is needed
string[] subtasks = await context.CallActivityAsync<string[]>(
nameof(PlanResearchAgent), request.Topic);
// Delegate to worker orchestrations in parallel
var workerTasks = subtasks
.Select(subtask => context.CallSubOrchestratorAsync<string>(
nameof(ResearchWorkerOrchestration), subtask))
.ToList();
string[] results = await Task.WhenAll(workerTasks);
// Synthesize results
string finalReport = await context.CallActivityAsync<string>(
nameof(SynthesizeAgent),
new { request.Topic, Research = results });
return finalReport;
}
}
def orchestrator_workers_orchestration(ctx: task.OrchestrationContext, request: dict) -> str:
# Central orchestrator: determine what research is needed
subtasks = yield ctx.call_activity(plan_research_agent, input=request["topic"])
# Delegate to worker orchestrations in parallel
worker_tasks = []
for subtask in subtasks:
worker_tasks.append(
ctx.call_sub_orchestrator(research_worker_orchestration, input=subtask)
)
results = yield task.when_all(worker_tasks)
# Synthesize results
final_report = yield ctx.call_activity(synthesize_agent, input={
"topic": request["topic"],
"research": results
})
return final_report
const orchestratorWorkersOrchestration: TOrchestrator = async function* (
ctx: OrchestrationContext, request: ResearchRequest): any {
// Central orchestrator: determine what research is needed
const subtasks: string[] = yield ctx.callActivity(planResearchAgent, request.topic);
// Delegate to worker orchestrations in parallel
const workerTasks = subtasks.map((subtask) =>
ctx.callSubOrchestrator(researchWorkerOrchestration, subtask)
);
const results: string[] = yield whenAll(workerTasks);
// Synthesize results
const finalReport: string = yield ctx.callActivity(synthesizeAgent, {
topic: request.topic,
research: results,
});
return finalReport;
};
ctx -> {
ResearchRequest request = ctx.getInput(ResearchRequest.class);
// Central orchestrator: determine what research is needed
List<String> subtasks = ctx.callActivity(
"PlanResearchAgent", request.getTopic(), List.class).await();
// Delegate to worker orchestrations in parallel
List<Task<String>> workerTasks = subtasks.stream()
.map(subtask -> ctx.callSubOrchestrator(
"ResearchWorkerOrchestration", subtask, String.class))
.collect(Collectors.toList());
List<String> results = ctx.allOf(workerTasks).await();
// Synthesize results
String finalReport = ctx.callActivity(
"SynthesizeAgent", results, String.class).await();
ctx.complete(finalReport);
}
Evaluador-optimizador
El patrón de optimizador-optimizador empareja un agente de generador con un agente de evaluador en un bucle de refinamiento. El generador produce la salida, el evaluador la puntúa según los criterios de calidad y proporciona comentarios, y el bucle se repite hasta que la salida cumpla los criterios o se alcance un número máximo de iteraciones. Dado que cada iteración del bucle se guarda en un punto de control, un bloqueo después de tres rondas exitosas de refinamiento no perderá ese progreso.
Este patrón es especialmente útil cuando la calidad se puede medir mediante programación, por ejemplo, validar que el código generado se compila o que una traducción conserva las entidades con nombre.
Cuándo usar: Generación de código con revisión automatizada, traducción literaria, refinamiento de contenido iterativo, tareas de búsqueda complejas que requieren varias rondas de análisis.
[Function(nameof(EvaluatorOptimizerOrchestration))]
public async Task<string> EvaluatorOptimizerOrchestration(
[OrchestrationTrigger] TaskOrchestrationContext context)
{
var request = context.GetInput<ContentRequest>();
int maxIterations = 5;
string content = "";
string feedback = "";
for (int i = 0; i < maxIterations; i++)
{
// Generate or refine content
content = await context.CallActivityAsync<string>(
nameof(GenerateContentAgent),
new { request.Prompt, PreviousContent = content, Feedback = feedback });
// Evaluate quality
var evaluation = await context.CallActivityAsync<EvaluationResult>(
nameof(EvaluateContentAgent), content);
if (evaluation.MeetsQualityBar)
return content;
feedback = evaluation.Feedback;
}
return content; // Return best effort after max iterations
}
@app.orchestration_trigger(context_name="context")
def evaluator_optimizer_orchestration(context: df.DurableOrchestrationContext):
request = context.get_input()
max_iterations = 5
content = ""
feedback = ""
for i in range(max_iterations):
# Generate or refine content
content = yield context.call_activity("generate_content_agent", {
"prompt": request["prompt"],
"previous_content": content,
"feedback": feedback
})
# Evaluate quality
evaluation = yield context.call_activity("evaluate_content_agent", content)
if evaluation["meets_quality_bar"]:
return content
feedback = evaluation["feedback"]
return content # Return best effort after max iterations
const df = require("durable-functions");
df.app.orchestration("evaluatorOptimizerOrchestration", function* (context) {
const request = context.df.getInput();
const maxIterations = 5;
let content = "";
let feedback = "";
for (let i = 0; i < maxIterations; i++) {
// Generate or refine content
content = yield context.df.callActivity("generateContentAgent", {
prompt: request.prompt,
previousContent: content,
feedback: feedback,
});
// Evaluate quality
const evaluation = yield context.df.callActivity("evaluateContentAgent", content);
if (evaluation.meetsQualityBar) {
return content;
}
feedback = evaluation.feedback;
}
return content; // Return best effort after max iterations
});
@FunctionName("EvaluatorOptimizerOrchestration")
public String evaluatorOptimizerOrchestration(
@DurableOrchestrationTrigger(name = "ctx") TaskOrchestrationContext ctx) {
ContentRequest request = ctx.getInput(ContentRequest.class);
int maxIterations = 5;
String content = "";
String feedback = "";
for (int i = 0; i < maxIterations; i++) {
// Generate or refine content
content = ctx.callActivity("GenerateContentAgent",
new GenerateInput(request.getPrompt(), content, feedback),
String.class).await();
// Evaluate quality
EvaluationResult evaluation = ctx.callActivity(
"EvaluateContentAgent", content, EvaluationResult.class).await();
if (evaluation.meetsQualityBar()) {
return content;
}
feedback = evaluation.getFeedback();
}
return content; // Return best effort after max iterations
}
[DurableTask]
public class EvaluatorOptimizerOrchestration : TaskOrchestrator<ContentRequest, string>
{
public override async Task<string> RunAsync(
TaskOrchestrationContext context, ContentRequest request)
{
int maxIterations = 5;
string content = "";
string feedback = "";
for (int i = 0; i < maxIterations; i++)
{
// Generate or refine content
content = await context.CallActivityAsync<string>(
nameof(GenerateContentAgent),
new { request.Prompt, PreviousContent = content, Feedback = feedback });
// Evaluate quality
var evaluation = await context.CallActivityAsync<EvaluationResult>(
nameof(EvaluateContentAgent), content);
if (evaluation.MeetsQualityBar)
return content;
feedback = evaluation.Feedback;
}
return content; // Return best effort after max iterations
}
}
def evaluator_optimizer_orchestration(ctx: task.OrchestrationContext, request: dict) -> str:
max_iterations = 5
content = ""
feedback = ""
for i in range(max_iterations):
# Generate or refine content
content = yield ctx.call_activity(generate_content_agent, input={
"prompt": request["prompt"],
"previous_content": content,
"feedback": feedback
})
# Evaluate quality
evaluation = yield ctx.call_activity(evaluate_content_agent, input=content)
if evaluation["meets_quality_bar"]:
return content
feedback = evaluation["feedback"]
return content # Return best effort after max iterations
const evaluatorOptimizerOrchestration: TOrchestrator = async function* (
ctx: OrchestrationContext, request: ContentRequest): any {
const maxIterations = 5;
let content = "";
let feedback = "";
for (let i = 0; i < maxIterations; i++) {
// Generate or refine content
content = yield ctx.callActivity(generateContentAgent, {
prompt: request.prompt,
previousContent: content,
feedback: feedback,
});
// Evaluate quality
const evaluation = yield ctx.callActivity(evaluateContentAgent, content);
if (evaluation.meetsQualityBar) {
return content;
}
feedback = evaluation.feedback;
}
return content; // Return best effort after max iterations
};
ctx -> {
ContentRequest request = ctx.getInput(ContentRequest.class);
int maxIterations = 5;
String content = "";
String feedback = "";
for (int i = 0; i < maxIterations; i++) {
// Generate or refine content
content = ctx.callActivity("GenerateContentAgent",
new GenerateInput(request.getPrompt(), content, feedback),
String.class).await();
// Evaluate quality
EvaluationResult evaluation = ctx.callActivity(
"EvaluateContentAgent", content, EvaluationResult.class).await();
if (evaluation.meetsQualityBar()) {
ctx.complete(content);
return;
}
feedback = evaluation.getFeedback();
}
ctx.complete(content); // Return best effort after max iterations
}
Bucles de agentes
En una implementación típica de un agente de IA, se invoca un LLM en un bucle, llamando a herramientas y tomando decisiones hasta que la tarea se completa o se alcanza una condición de parada. A diferencia de los flujos de trabajo deterministas, la trayectoria de ejecución no está predefinida. El agente determina qué hacer en cada paso en función de los resultados de los pasos anteriores.
Los bucles de agente son adecuados para las tareas en las que no se puede predecir el número o el orden de los pasos. Entre los ejemplos comunes se incluyen agentes de codificación abiertos, investigaciones autónomas y bots conversacionales con funcionalidades de llamada a herramientas.
Hay dos enfoques recomendados para implementar bucles de agentes con el modelo de programación de Durable Task:
| Enfoque |
Descripción |
Cuándo se deben usar |
|
Basado en orquestación |
Plantee el ciclo del agente como una orquestación duradera. Las llamadas a herramientas se implementan como actividades y la entrada del usuario se gestiona mediante eventos externos. La orquestación controla la estructura del bucle mientras LLM controla las decisiones dentro de este. |
Necesita un control detallado sobre el bucle, las directivas de reintento por herramienta, el balanceo de carga distribuido de las llamadas a herramientas o la capacidad de depurar el bucle en su IDE con puntos de interrupción. |
|
Basado en entidades |
Cada instancia del agente es una entidad duradera. El marco del agente gestiona el bucle internamente, y la entidad ofrece persistencia tanto del estado como de la sesión de manera duradera. |
Usa un marco de agente (como Microsoft Agent Framework) que ya implementa el bucle del agente y quiere agregar durabilidad con cambios mínimos de código. |
Bucles de agente basados en la orquestación
Un bucle de agente basado en orquestación combina varias capacidades de Durable Task: orquestaciones eternas (continue-as-new) para mantener la memoria limitada, fan-out/fan-in para la ejecución paralela de herramientas y eventos externos para interacciones humanas en el bucle. Cada iteración del bucle:
- Envía el contexto actual de la conversación al LLM a través de una actividad o entidad con estado.
- Recibe la respuesta del LLM, que puede incluir llamadas a herramientas.
- Ejecuta cualquier llamada de herramienta como actividades (distribuidas entre los recursos informáticos disponibles).
- Opcionalmente, espera la entrada humana mediante eventos externos.
- Continúa el bucle con el estado actualizado o se completa cuando el agente señala que ha terminado.
[Function(nameof(AgentLoopOrchestration))]
public async Task<string> AgentLoopOrchestration(
[OrchestrationTrigger] TaskOrchestrationContext context)
{
// Get state from input (supports continue-as-new)
var state = context.GetInput<AgentState>() ?? new AgentState();
int maxIterations = 100;
while (state.Iteration < maxIterations)
{
// Send conversation history to the LLM
var llmResponse = await context.CallActivityAsync<LlmResponse>(
nameof(CallLlmAgent), state.Messages);
state.Messages.Add(llmResponse.Message);
// If the LLM returned tool calls, execute them in parallel
if (llmResponse.ToolCalls is { Count: > 0 })
{
var toolTasks = llmResponse.ToolCalls
.Select(tc => context.CallActivityAsync<ToolResult>(
nameof(ExecuteTool), tc))
.ToList();
ToolResult[] toolResults = await Task.WhenAll(toolTasks);
foreach (var result in toolResults)
state.Messages.Add(result.ToMessage());
}
// If the LLM needs human input, wait for it
else if (llmResponse.NeedsHumanInput)
{
string humanInput = await context.WaitForExternalEvent<string>("HumanInput");
state.Messages.Add(new Message("user", humanInput));
}
// LLM is done
else
{
return llmResponse.FinalAnswer;
}
state.Iteration++;
// Periodically continue-as-new to keep the history bounded
if (state.Iteration % 10 == 0)
{
context.ContinueAsNew(state);
return null!; // Orchestration will restart with updated state
}
}
return "Max iterations reached.";
}
@app.orchestration_trigger(context_name="context")
def agent_loop_orchestration(context: df.DurableOrchestrationContext):
# Get state from input (supports continue-as-new)
state = context.get_input() or {"messages": [], "iteration": 0}
max_iterations = 100
while state["iteration"] < max_iterations:
# Send conversation history to the LLM
llm_response = yield context.call_activity("call_llm_agent", state["messages"])
state["messages"].append(llm_response["message"])
# If the LLM returned tool calls, execute them
if llm_response.get("tool_calls"):
tool_tasks = [
context.call_activity("execute_tool", tc)
for tc in llm_response["tool_calls"]
]
tool_results = yield context.task_all(tool_tasks)
for result in tool_results:
state["messages"].append(result)
# If the LLM needs human input, wait for it
elif llm_response.get("needs_human_input"):
human_input = yield context.wait_for_external_event("HumanInput")
state["messages"].append({"role": "user", "content": human_input})
# LLM is done
else:
return llm_response["final_answer"]
state["iteration"] += 1
# Periodically continue-as-new to keep the history bounded
if state["iteration"] % 10 == 0:
context.continue_as_new(state)
return
return "Max iterations reached."
const df = require("durable-functions");
df.app.orchestration("agentLoopOrchestration", function* (context) {
// Get state from input (supports continue-as-new)
const state = context.df.getInput() || { messages: [], iteration: 0 };
const maxIterations = 100;
while (state.iteration < maxIterations) {
// Send conversation history to the LLM
const llmResponse = yield context.df.callActivity("callLlmAgent", state.messages);
state.messages.push(llmResponse.message);
// If the LLM returned tool calls, execute them
if (llmResponse.toolCalls && llmResponse.toolCalls.length > 0) {
const toolTasks = llmResponse.toolCalls.map((tc) =>
context.df.callActivity("executeTool", tc)
);
const toolResults = yield context.df.Task.all(toolTasks);
for (const result of toolResults) {
state.messages.push(result);
}
// If the LLM needs human input, wait for it
} else if (llmResponse.needsHumanInput) {
const humanInput = yield context.df.waitForExternalEvent("HumanInput");
state.messages.push({ role: "user", content: humanInput });
// LLM is done
} else {
return llmResponse.finalAnswer;
}
state.iteration++;
// Periodically continue-as-new to keep the history bounded
if (state.iteration % 10 === 0) {
context.df.continueAsNew(state);
return;
}
}
return "Max iterations reached.";
});
@FunctionName("AgentLoopOrchestration")
public String agentLoopOrchestration(
@DurableOrchestrationTrigger(name = "ctx") TaskOrchestrationContext ctx) {
// Get state from input (supports continue-as-new)
AgentState state = ctx.getInput(AgentState.class);
if (state == null) state = new AgentState();
int maxIterations = 100;
while (state.getIteration() < maxIterations) {
// Send conversation history to the LLM
LlmResponse llmResponse = ctx.callActivity(
"CallLlmAgent", state.getMessages(), LlmResponse.class).await();
state.getMessages().add(llmResponse.getMessage());
// If the LLM returned tool calls, execute them
if (llmResponse.getToolCalls() != null && !llmResponse.getToolCalls().isEmpty()) {
List<Task<ToolResult>> toolTasks = llmResponse.getToolCalls().stream()
.map(tc -> ctx.callActivity("ExecuteTool", tc, ToolResult.class))
.collect(Collectors.toList());
List<ToolResult> toolResults = ctx.allOf(toolTasks).await();
for (ToolResult result : toolResults) {
state.getMessages().add(result.toMessage());
}
// If the LLM needs human input, wait for it
} else if (llmResponse.needsHumanInput()) {
String humanInput = ctx.waitForExternalEvent("HumanInput", String.class).await();
state.getMessages().add(new Message("user", humanInput));
// LLM is done
} else {
return llmResponse.getFinalAnswer();
}
state.incrementIteration();
// Periodically continue-as-new to keep the history bounded
if (state.getIteration() % 10 == 0) {
ctx.continueAsNew(state);
return null;
}
}
return "Max iterations reached.";
}
[DurableTask]
public class AgentLoopOrchestration : TaskOrchestrator<AgentState, string>
{
public override async Task<string> RunAsync(
TaskOrchestrationContext context, AgentState? state)
{
state ??= new AgentState();
int maxIterations = 100;
while (state.Iteration < maxIterations)
{
// Send conversation history to the LLM
var llmResponse = await context.CallActivityAsync<LlmResponse>(
nameof(CallLlmAgent), state.Messages);
state.Messages.Add(llmResponse.Message);
// If the LLM returned tool calls, execute them
if (llmResponse.ToolCalls is { Count: > 0 })
{
var toolTasks = llmResponse.ToolCalls
.Select(tc => context.CallActivityAsync<ToolResult>(
nameof(ExecuteTool), tc))
.ToList();
ToolResult[] toolResults = await Task.WhenAll(toolTasks);
foreach (var result in toolResults)
state.Messages.Add(result.ToMessage());
}
// If the LLM needs human input, wait for it
else if (llmResponse.NeedsHumanInput)
{
string humanInput = await context.WaitForExternalEvent<string>("HumanInput");
state.Messages.Add(new Message("user", humanInput));
}
// LLM is done
else
{
return llmResponse.FinalAnswer;
}
state.Iteration++;
// Periodically continue-as-new to keep the history bounded
if (state.Iteration % 10 == 0)
{
context.ContinueAsNew(state);
return null!;
}
}
return "Max iterations reached.";
}
}
def agent_loop_orchestration(ctx: task.OrchestrationContext, state: dict | None) -> str:
if state is None:
state = {"messages": [], "iteration": 0}
max_iterations = 100
while state["iteration"] < max_iterations:
# Send conversation history to the LLM
llm_response = yield ctx.call_activity(call_llm_agent, input=state["messages"])
state["messages"].append(llm_response["message"])
# If the LLM returned tool calls, execute them
if llm_response.get("tool_calls"):
tool_tasks = [
ctx.call_activity(execute_tool, input=tc)
for tc in llm_response["tool_calls"]
]
tool_results = yield task.when_all(tool_tasks)
for result in tool_results:
state["messages"].append(result)
# If the LLM needs human input, wait for it
elif llm_response.get("needs_human_input"):
human_input = yield ctx.wait_for_external_event("HumanInput")
state["messages"].append({"role": "user", "content": human_input})
# LLM is done
else:
return llm_response["final_answer"]
state["iteration"] += 1
# Periodically continue-as-new to keep the history bounded
if state["iteration"] % 10 == 0:
ctx.continue_as_new(state)
return
return "Max iterations reached."
const agentLoopOrchestration: TOrchestrator = async function* (
ctx: OrchestrationContext, state: AgentState | null): any {
if (!state) state = { messages: [], iteration: 0 };
const maxIterations = 100;
while (state.iteration < maxIterations) {
// Send conversation history to the LLM
const llmResponse = yield ctx.callActivity(callLlmAgent, state.messages);
state.messages.push(llmResponse.message);
// If the LLM returned tool calls, execute them
if (llmResponse.toolCalls && llmResponse.toolCalls.length > 0) {
const toolTasks = llmResponse.toolCalls.map((tc: any) =>
ctx.callActivity(executeTool, tc)
);
const toolResults = yield whenAll(toolTasks);
for (const result of toolResults) {
state.messages.push(result);
}
// If the LLM needs human input, wait for it
} else if (llmResponse.needsHumanInput) {
const humanInput: string = yield ctx.waitForExternalEvent("HumanInput");
state.messages.push({ role: "user", content: humanInput });
// LLM is done
} else {
return llmResponse.finalAnswer;
}
state.iteration++;
// Periodically continue-as-new to keep the history bounded
if (state.iteration % 10 === 0) {
ctx.continueAsNew(state);
return;
}
}
return "Max iterations reached.";
};
ctx -> {
AgentState state = ctx.getInput(AgentState.class);
if (state == null) state = new AgentState();
int maxIterations = 100;
while (state.getIteration() < maxIterations) {
// Send conversation history to the LLM
LlmResponse llmResponse = ctx.callActivity(
"CallLlmAgent", state.getMessages(), LlmResponse.class).await();
state.getMessages().add(llmResponse.getMessage());
// If the LLM returned tool calls, execute them
if (llmResponse.getToolCalls() != null && !llmResponse.getToolCalls().isEmpty()) {
List<Task<ToolResult>> toolTasks = llmResponse.getToolCalls().stream()
.map(tc -> ctx.callActivity("ExecuteTool", tc, ToolResult.class))
.collect(Collectors.toList());
List<ToolResult> toolResults = ctx.allOf(toolTasks).await();
for (ToolResult result : toolResults) {
state.getMessages().add(result.toMessage());
}
// If the LLM needs human input, wait for it
} else if (llmResponse.needsHumanInput()) {
String humanInput = ctx.waitForExternalEvent("HumanInput", String.class).await();
state.getMessages().add(new Message("user", humanInput));
// LLM is done
} else {
ctx.complete(llmResponse.getFinalAnswer());
return;
}
state.incrementIteration();
// Periodically continue-as-new to keep the history bounded
if (state.getIteration() % 10 == 0) {
ctx.continueAsNew(state);
return;
}
}
ctx.complete("Max iterations reached.");
}
Bucles de agentes basados en entidades
Si usa un marco de agente que ya implementa su propio bucle de agente, puede encapsularlo en una entidad duradera para agregar durabilidad sin volver a escribir la lógica del bucle. Cada instancia de entidad representa una sesión de agente única. La entidad recibe mensajes, los delega internamente al marco de agentes y conserva el estado de la conversación a lo largo de las interacciones.
La principal ventaja de este enfoque es la simplicidad: escribes tu agente usando tu entorno preferido y agregas durabilidad como un asunto de hospedaje en lugar de rediseñar el flujo de control del agente. La entidad actúa como un contenedor duradero, gestionando automáticamente la persistencia y recuperación de la sesión.
En los ejemplos siguientes se muestra cómo encapsular un SDK de agente existente como una entidad duradera. La entidad expone una operación message que los clientes invocan para enviar la entrada del usuario. Internamente, la entidad se delega en el marco del agente, que administra su propio bucle de llamada a herramientas.
// Define the entity that wraps an existing agent SDK
public class ChatAgentEntity : TaskEntity<ChatAgentState>
{
private readonly IChatClient _chatClient;
public ChatAgentEntity(IChatClient chatClient)
{
_chatClient = chatClient;
}
// Called by clients to send a message to the agent
public async Task<string> Message(string userMessage)
{
// Add the user message to the conversation history
State.Messages.Add(new ChatMessage(ChatRole.User, userMessage));
// Delegate to the agent SDK for the LLM call (with tool loop)
ChatResponse response = await _chatClient.GetResponseAsync(
State.Messages, State.Options);
// Persist the response in the entity state
State.Messages.AddRange(response.Messages);
return response.Text;
}
// Azure Functions entry point for the entity
[Function(nameof(ChatAgentEntity))]
public Task RunEntityAsync([EntityTrigger] TaskEntityDispatcher dispatcher)
{
return dispatcher.DispatchAsync<ChatAgentEntity>();
}
}
# Define the entity that wraps an existing agent SDK
@app.entity_trigger(context_name="context")
def chat_agent_entity(context):
# Load persisted conversation state
state = context.get_state(lambda: {"messages": []})
if context.operation_name == "message":
user_message = context.get_input()
# Add the user message to the conversation history
state["messages"].append({"role": "user", "content": user_message})
# Delegate to the agent SDK for the LLM call (with tool loop)
response = call_agent_sdk(state["messages"])
# Persist the response in the entity state
state["messages"].append({"role": "assistant", "content": response})
context.set_state(state)
context.set_result(response)
const df = require("durable-functions");
// Define the entity that wraps an existing agent SDK
const chatAgentEntity = async function (context) {
// Load persisted conversation state
let state = context.df.getState(() => ({ messages: [] }));
switch (context.df.operationName) {
case "message":
const userMessage = context.df.getInput();
// Add the user message to the conversation history
state.messages.push({ role: "user", content: userMessage });
// Delegate to the agent SDK for the LLM call (with tool loop)
const response = await callAgentSdk(state.messages);
// Persist the response in the entity state
state.messages.push({ role: "assistant", content: response });
context.df.setState(state);
context.df.return(response);
break;
}
};
df.app.entity("ChatAgent", chatAgentEntity);
Nota:
Las entidades duraderas de Java requieren la versión 1.9.0 o posterior de los paquetes /> /.
// Define the entity that wraps an existing agent SDK
public class ChatAgentEntity extends AbstractTaskEntity<ChatAgentState> {
// Called by clients to send a message to the agent
public String message(String userMessage) {
// Add the user message to the conversation history
this.state.getMessages().add(new ChatMessage("user", userMessage));
// Delegate to the agent SDK for the LLM call (with tool loop)
String response = callAgentSdk(this.state.getMessages());
// Persist the response in the entity state
this.state.getMessages().add(new ChatMessage("assistant", response));
return response;
}
@Override
protected ChatAgentState initializeState(TaskEntityOperation operation) {
return new ChatAgentState();
}
}
// Register the entity with Azure Functions
@FunctionName("ChatAgent")
public String chatAgentEntity(
@DurableEntityTrigger(name = "req") String req) {
return EntityRunner.loadAndRun(req, ChatAgentEntity::new);
}
// Define the entity that wraps an existing agent SDK
[DurableTask(Name = "ChatAgent")]
public class ChatAgentEntity : TaskEntity<ChatAgentState>
{
private readonly IChatClient _chatClient;
public ChatAgentEntity(IChatClient chatClient)
{
_chatClient = chatClient;
}
// Called by clients to send a message to the agent
public async Task<string> Message(string userMessage)
{
// Add the user message to the conversation history
State.Messages.Add(new ChatMessage(ChatRole.User, userMessage));
// Delegate to the agent SDK for the LLM call (with tool loop)
ChatResponse response = await _chatClient.GetResponseAsync(
State.Messages, State.Options);
// Persist the response in the entity state
State.Messages.AddRange(response.Messages);
return response.Text;
}
}
from durabletask.entities.durable_entity import DurableEntity
# Define the entity that wraps an existing agent SDK
class ChatAgentEntity(DurableEntity):
"""Durable entity wrapping an agent SDK."""
def message(self, user_message: str) -> str:
# Load persisted conversation state
state = self.get_state(default={"messages": []})
# Add the user message to the conversation history
state["messages"].append({"role": "user", "content": user_message})
# Delegate to the agent SDK for the LLM call (with tool loop)
response = call_agent_sdk(state["messages"])
# Persist the response in the entity state
state["messages"].append({"role": "assistant", "content": response})
self.set_state(state)
return response
import { TaskEntity } from "@microsoft/durabletask-js";
// Define the entity that wraps an existing agent SDK
class ChatAgentEntity extends TaskEntity<ChatAgentState> {
// Called by clients to send a message to the agent
async message(userMessage: string): Promise<string> {
// Add the user message to the conversation history
this.state.messages.push({ role: "user", content: userMessage });
// Delegate to the agent SDK for the LLM call (with tool loop)
const response = await callAgentSdk(this.state.messages);
// Persist the response in the entity state
this.state.messages.push({ role: "assistant", content: response });
return response;
}
initializeState(): ChatAgentState {
return { messages: [] };
}
}
Nota:
Las entidades duraderas de Java requieren la versión 1.9.0 o posterior del paquete durabletask-client.
// Define the entity that wraps an existing agent SDK
public class ChatAgentEntity extends AbstractTaskEntity<ChatAgentState> {
// Called by clients to send a message to the agent
public String message(String userMessage) {
// Add the user message to the conversation history
this.state.getMessages().add(new ChatMessage("user", userMessage));
// Delegate to the agent SDK for the LLM call (with tool loop)
String response = callAgentSdk(this.state.getMessages());
// Persist the response in the entity state
this.state.getMessages().add(new ChatMessage("assistant", response));
return response;
}
@Override
protected ChatAgentState initializeState(TaskEntityOperation operation) {
return new ChatAgentState();
}
}
La extensión Durable Task para Microsoft Agent Framework usa este enfoque. Encapsula agentes del Microsoft Agent Framework como entidades duraderas, proporcionando sesiones persistentes, puntos de control automáticos y puntos de conexión API integrados con una sola línea de configuración.
Pasos siguientes