Esistono due approcci generali per la creazione di applicazioni agenti con l'intelligenza artificiale:
-
Flussi di lavoro deterministici : il codice definisce il flusso di controllo. Si scrive la sequenza di passaggi, diramazione, parallelismo e gestione degli errori usando costrutti di programmazione standard. L'LLM esegue operazioni all'interno di ogni passaggio, ma non controlla il processo complessivo.
-
Flussi di lavoro diretti dall'agente (cicli dell'agente) — LLM determina il flusso di controllo. L'agente decide quali strumenti chiamare, in quale ordine e quando l'attività è stata completata. Sono disponibili strumenti e istruzioni, ma l'agente determina il percorso di esecuzione in fase di esecuzione.
Entrambi gli approcci traggono vantaggio dall'esecuzione durevole e possono essere implementati usando il modello di programmazione Durable Task. Questo articolo illustra come compilare ogni modello usando esempi di codice.
Suggerimento
Questi modelli sono allineati alle progettazioni del flusso di lavoro agentico descritte in Building Effective Agents di Anthropic. Il modello di programmazione Durable Task esegue naturalmente il mapping a questi modelli: le orchestrazioni definiscono il flusso di controllo del flusso di lavoro e vengono automaticamente controllate, mentre le attività eseguono il wrapping di operazioni non deterministiche come chiamate LLM, chiamate degli strumenti e richieste API.
Scegliere un approccio
La tabella seguente consente di decidere quando usare ogni approccio.
| Usare flussi di lavoro deterministici quando... |
Usa cicli dell'agente quando... |
| La sequenza di passaggi è nota in anticipo. |
L'attività è aperta e i passaggi non possono essere stimati. |
| Sono necessarie protezioni esplicite sul comportamento dell'agente. |
Si vuole che l'LLM decida quali strumenti usare e quando. |
| La conformità o la verificabilità richiede un flusso di controllo verificabile. |
L'agente deve adattare il proprio approccio in base ai risultati intermedi. |
| Si vogliono combinare più framework di intelligenza artificiale in un singolo flusso di lavoro. |
Si sta creando un agente di conversazione con funzionalità di chiamata agli strumenti. |
Entrambi gli approcci offrono i checkpoint automatici, le politiche di ripetizione, il ridimensionamento distribuito e il supporto umano nel ciclo tramite esecuzione durevole.
Modelli di flusso di lavoro deterministici
In un flusso di lavoro deterministico, il codice controlla il percorso di esecuzione. LLM viene chiamato come passaggio all'interno del flusso di lavoro, ma non decide cosa accade successivamente. Il modello di programmazione Durable Task si adatta naturalmente a questo approccio.
-
Le orchestrazioni definiscono il flusso di controllo del flusso di lavoro (sequenza, diramazione, parallelismo, gestione degli errori) e vengono automaticamente salvate con un checkpoint.
-
Le attività avvolgono operazioni non deterministiche come chiamate LLM, invocazioni di strumenti e richieste API. Le attività possono essere eseguite in qualsiasi istanza di calcolo disponibile.
Gli esempi seguenti usano Durable Functions, che viene eseguito in Funzioni di Azure con hosting serverless.
Gli esempi seguenti usano gli SDK delle attività portabili Durable, che vengono eseguiti in qualsiasi ambiente di calcolo host, tra cui App contenitore di Azure, Kubernetes, macchine virtuali o localmente.
Concatenamento istruzioni
Il concatenamento dei prompt è il modello agentico più semplice. Un'attività complessa viene suddivisa in una serie di interazioni LLM sequenziali, in cui l'output di ogni passaggio viene inserito nell'input del passaggio successivo. Poiché ogni chiamata di attività viene automaticamente sottoposta a checkpoint, un arresto anomalo a metà della pipeline non forza il riavvio da zero e il riutilizzo di token LLM costosi, l'esecuzione riprende dall'ultimo passaggio completato.
È anche possibile inserire controlli di convalida a livello di codice tra i passaggi. Ad esempio, dopo aver generato una struttura, è possibile verificare che soddisfi un vincolo di lunghezza o argomento prima di passarlo al passaggio di stesura.
Questo modello corrisponde direttamente al modello di concatenamento delle funzioni nel modello di programmazione Durable Task.
Quando usare: Pipeline di generazione del contenuto, elaborazione di documenti in più passaggi, arricchimento sequenziale dei dati, flussi di lavoro che richiedono controlli di convalida intermedi.
[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;
}
Annotazioni
Lo stato dell'orchestrazione viene automaticamente sottoposto a checkpoint in ogni await istruzione. Se il processo host si arresta in modo anomalo o la macchina virtuale viene riciclata, l'orchestrazione riprenderà automaticamente dall'ultimo passaggio completato anziché ricominciare.
@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
Annotazioni
Lo stato dell'orchestrazione viene automaticamente checkpointato in ogni yield istruzione. Se il processo host si arresta in modo anomalo o la macchina virtuale viene riciclata, l'orchestrazione riprenderà automaticamente dall'ultimo passaggio completato anziché ricominciare.
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;
});
Annotazioni
Lo stato dell'orchestrazione viene automaticamente sottoposto a checkpoint a ciascuna istruzione yield. Se il processo host si arresta in modo anomalo o la macchina virtuale viene riciclata, l'orchestrazione riprenderà automaticamente dall'ultimo passaggio completato anziché ricominciare.
@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;
}
Annotazioni
Lo stato dell'orchestrazione viene automaticamente salvato a ogni await() chiamata. Se il processo host si arresta in modo anomalo o la macchina virtuale viene riciclata, l'orchestrazione riprenderà automaticamente dall'ultimo passaggio completato anziché ricominciare.
[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;
}
}
Annotazioni
Lo stato dell'orchestrazione viene automaticamente sottoposto a checkpoint a ciascuna istruzione await. Se il processo host si arresta in modo anomalo o la macchina virtuale viene riciclata, l'orchestrazione riprenderà automaticamente dall'ultimo passaggio completato anziché ricominciare.
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
Annotazioni
Lo stato dell'orchestrazione viene automaticamente sottoposto a checkpoint a ciascuna istruzione yield. Se il processo host si arresta in modo anomalo o la macchina virtuale viene riciclata, l'orchestrazione riprenderà automaticamente dall'ultimo passaggio completato anziché ricominciare.
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;
};
Annotazioni
Lo stato dell'orchestrazione viene automaticamente sottoposto a checkpoint a ciascuna istruzione yield. Se il processo host si arresta in modo anomalo o la macchina virtuale viene riciclata, l'orchestrazione riprenderà automaticamente dall'ultimo passaggio completato anziché ricominciare.
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);
}
Annotazioni
Lo stato dell'orchestrazione viene automaticamente registrato durante ogni invocazione await(). Se il processo host si arresta in modo anomalo o la macchina virtuale viene riciclata, l'orchestrazione riprenderà automaticamente dall'ultimo passaggio completato anziché ricominciare.
Instradamento
Il routing usa un passaggio di classificazione per determinare quale agente downstream o modello deve gestire una richiesta. L'orchestrazione chiama prima un'attività di classificazione, quindi passa al gestore appropriato in base al risultato. Questo approccio consente di personalizzare in modo indipendente il prompt, il modello e il set di strumenti di ogni gestore, ad esempio indirizzando domande di fatturazione a un agente specializzato con accesso alle API di pagamento inviando domande generali a un modello più leggero.
Quando usare: Valutazione del supporto clienti, classificazione delle finalità per agenti specializzati, selezione dinamica del modello in base alla complessità delle attività.
[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);
}
Parallelizzazione
Quando si dispone di più sottoattività indipendenti, è possibile inviarle come chiamate di attività parallele e attendere tutti i risultati prima di procedere. L'utilità di pianificazione delle attività durevoli distribuisce automaticamente queste attività in tutte le istanze di calcolo disponibili, ciò implica che l'aggiunta di più worker riduce direttamente il tempo totale.
Una variante comune è il voto a più modelli: si invia la stessa richiesta a più modelli (o allo stesso modello con temperature diverse) in parallelo, quindi aggregare o selezionare le risposte. Poiché ogni ramo parallelo è sottoposto a checkpoint indipendentemente, un errore temporaneo in un ramo non influisce sugli altri.
Questo modello si mappa direttamente al modello fan-out/fan-in nel Durable Task.
Quando usare: Analisi batch di documenti, chiamate di strumenti paralleli, valutazione multimodelli, moderazione del contenuto con più revisori.
[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);
}
Agenti di orchestrazione
In questo modello, un orchestratore centrale chiama prima un LLM (tramite un'attività) per pianificare il lavoro. In base all'output dell'LLM, l'agente di orchestrazione determina quindi quali sottoattività sono necessarie. L'orchestratore invia quindi tali sottoattività alle orchestrazioni di lavoratori specializzati. La differenza principale rispetto alla parallelizzazione è che il set di sottoattività non è fisso in fase di progettazione; l'agente di orchestrazione li determina in modo dinamico in fase di esecuzione.
Questo modello usa sotto orchestrazioni, che sono flussi di lavoro figlio con checkpoint indipendente. Ogni orchestrazione del worker può contenere più passaggi, tentativi e parallelismo annidato.
Quando usare: Pipeline di ricerca approfondita, codifica dei flussi di lavoro dell'agente che modificano più file, collaborazione multi-agente in cui ogni agente ha un ruolo 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);
}
Analizzatore-ottimizzatore
Il criterio di ottimizzazione dell'analizzatore associa un agente generatore a un agente di valutazione in un ciclo di perfezionamento. Il generatore produce output, l'analizzatore lo valuta rispetto ai criteri di qualità e fornisce feedback, e il ciclo si ripete fino a quando l'output non supera il test o viene raggiunto il numero massimo di iterazioni. Poiché ogni iterazione del ciclo viene registrata, un arresto anomalo dopo tre giri di perfezionamento riusciti non farà perdere i progressi.
Questo modello è particolarmente utile quando la qualità può essere misurata in modo programmatico — ad esempio, la validazione che il codice generato viene compilato correttamente o che una traduzione conserva le entità denominate.
Quando usare: Generazione di codice con revisione automatizzata, traduzione letteraria, perfezionamento del contenuto iterativo, attività di ricerca complesse che richiedono più round di analisi.
[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
}
Cicli dell'agente
In un'implementazione tipica dell'agente di intelligenza artificiale, un LLM viene richiamato in un ciclo, chiamando gli strumenti e prendendo decisioni fino al completamento dell'attività o fino a quando non viene raggiunta una condizione di arresto. A differenza dei flussi di lavoro deterministici, il percorso di esecuzione non è predefinito. L'agente determina le operazioni da eseguire in ogni passaggio in base ai risultati dei passaggi precedenti.
I cicli degli agenti sono ideali per le attività in cui non è possibile stimare il numero o l'ordine dei passaggi. Gli esempi comuni includono agenti di codifica aperti, ricerche autonome e bot di conversazione con funzionalità di chiamata agli strumenti.
Esistono due approcci consigliati per implementare cicli agente con il modello di programmazione Durable Task:
| Avvicinarsi |
Descrizione |
Quando utilizzare |
|
Basata sull'orchestrazione |
Scrivere il loop dell'agente come un'orchestrazione durevole. Le chiamate agli strumenti vengono implementate come attività e l'input umano usa eventi esterni. L'orchestrazione controlla la struttura del ciclo, mentre il modello linguistico di grandi dimensioni (LLM) gestisce le decisioni al suo interno. |
È necessario un controllo granulare sul ciclo, sui criteri di ripetizione dei tentativi per strumento, sul bilanciamento del carico distribuito delle chiamate agli strumenti o sulla possibilità di eseguire il debug del ciclo nell'IDE con punti di interruzione. |
|
Basato su entità |
Ogni istanza di agente è un'entità durevole. Il framework dell'agente gestisce internamente il ciclo e l'entità fornisce stato persistente e persistenza della sessione. |
Si usa un framework agente (ad esempio Microsoft Agent Framework) che implementa già il ciclo dell'agente e si vuole aggiungere durabilità con modifiche minime al codice. |
Cicli di agenti basati sull'orchestrazione
Un ciclo di agenti basato su orchestrazioni combina diverse funzionalità di Durable Task: orchestrazioni eterne (continue-as-new) per mantenere la memoria limitata, fan-out/fan-in per l'esecuzione parallela degli strumenti e eventi esterni per l'interazione umana nel ciclo. Ogni iterazione del ciclo:
- Invia il contesto di conversazione corrente all' LLM tramite un'attività o un'entità con stato.
- Riceve la risposta del modello LLM, che può includere richieste agli strumenti.
- Esegue qualsiasi chiamata di strumento come attività (distribuita tra le risorse di calcolo disponibili).
- Facoltativamente, attende l'input umano usando eventi esterni.
- Continua il ciclo con lo stato aggiornato o si completa quando l'agente segnala che è finito.
[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.");
}
Cicli di agenti basati su entità
Se si usa un framework agente che implementa già il proprio ciclo di agenti, è possibile eseguirne il wrapping in un'entità durevole per aggiungere durabilità senza riscrivere la logica del ciclo. Ogni istanza di entità rappresenta una singola sessione dell'agente. L'entità riceve messaggi, delega internamente al framework dell'agente e mantiene lo stato della conversazione attraverso le interazioni.
Il vantaggio principale di questo approccio è la semplicità: scrivere l'agente usando il framework preferito e aggiungere durabilità come problema di hosting anziché riprogettare il flusso di controllo dell'agente. L'entità funge da wrapper durevole, gestendo automaticamente la persistenza e il ripristino della sessione.
Gli esempi seguenti illustrano come incapsulare un SDK dell'agente esistente come entità durevole. L'entità espone un'operazione message che viene chiamata dai client per inviare l'input dell'utente. Internamente, l'entità delega al framework dell'agente, che gestisce il proprio ciclo di chiamate agli strumenti.
// 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);
Annotazioni
Le entità durevoli in Java richiedono la versione 1.9.0 o successiva dei pacchetti durabletask-azure-functions e 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();
}
}
// 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: [] };
}
}
Annotazioni
Le entità durevoli in Java richiedono la versione 1.9.0 o successiva del pacchetto 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();
}
}
L'estensione Durable Task per Microsoft Agent Framework usa questo approccio. Esegue l'incapsulamento degli agenti di Microsoft Agent Framework come entità durevoli, fornendo sessioni persistenti, checkpoint automatici ed endpoint API integrati con una singola riga di configurazione.
Passaggi successivi