w4c-workflows-api/Services/Execution/AgentScriptExecutor.cs

using System.Diagnostics;
using System.Net.Http.Json;
using System.Text.Json;

namespace w4c_workflows.Services.Execution;

/// <summary>
/// Executes a workflow <c>agent</c> step (W9) by invoking an LLM agent over HTTP
/// (D3): POSTs the step's prompt to chatapi <c>/api/agents/{id}/run</c> and returns
/// the assistant's content as the task output, so the run lifecycle continues. The
/// workflows engine stays decoupled from BotSharp — it only knows chatapi's HTTP
/// contract. The agent id is carried in the task's <c>entry.file</c> and the prompt
/// in the task's <c>input</c> (the previous task's output).
///
/// An LLM call may take tens of seconds; this executor blocks synchronously (the
/// worker awaits it), matching the engine's existing synchronous step model. If the
/// engine later needs non-blocking long-running steps, this executor is the seam to
/// hook async/callback semantics into.
/// </summary>
public sealed class AgentScriptExecutor : IScriptExecutor
{
    private readonly IHttpClientFactory _httpFactory;
    private readonly string _chatApiUrl;
    private readonly string _chatApiToken;
    private readonly int _timeoutSeconds;
    private readonly ILogger<AgentScriptExecutor> _logger;

    private static readonly JsonSerializerOptions Json = new(JsonSerializerDefaults.Web);

    public AgentScriptExecutor(IHttpClientFactory httpFactory, IConfiguration config, ILogger<AgentScriptExecutor> logger)
    {
        _httpFactory = httpFactory;
        _chatApiUrl = (config["Workflows:ChatApiUrl"] ?? string.Empty).TrimEnd('/');
        _chatApiToken = config["Workflows:ChatApiToken"] ?? string.Empty;
        _timeoutSeconds = ParseInt(config["Workflows:AgentStepTimeoutSeconds"], 120);
        _logger = logger;
    }

    public string Language => "agent";

    public bool IsAvailable() => !string.IsNullOrWhiteSpace(_chatApiUrl);

    public async Task<ExecutionResult> ExecuteAsync(TaskInvocation invocation, CancellationToken ct)
    {
        var sw = Stopwatch.StartNew();
        try
        {
            if (!IsAvailable())
                return Fail("agent step is not configured (Workflows:ChatApiUrl is empty)", sw.Elapsed);

            var agentId = invocation.EntryFile;
            if (string.IsNullOrWhiteSpace(agentId))
                return Fail("agent step requires entry.file to carry the agent id", sw.Elapsed);

            var request = new AgentRunRequestDto { Prompt = invocation.Input ?? string.Empty };
            var body = JsonSerializer.Serialize(request, Json);

            using var client = _httpFactory.CreateClient();
            using var httpReq = new HttpRequestMessage(HttpMethod.Post, $"{_chatApiUrl}/api/agents/{Uri.EscapeDataString(agentId)}/run")
            {
                Content = new StringContent(body, System.Text.Encoding.UTF8, "application/json"),
            };
            httpReq.Headers.TryAddWithoutValidation("X-Tenant-Id", invocation.TenantId);
            if (!string.IsNullOrWhiteSpace(_chatApiToken))
                httpReq.Headers.TryAddWithoutValidation("Authorization", $"Bearer {_chatApiToken}");

            using var cts = CancellationTokenSource.CreateLinkedTokenSource(ct);
            cts.CancelAfter(TimeSpan.FromSeconds(_timeoutSeconds + 15));
            using var response = await client.SendAsync(httpReq, HttpCompletionOption.ResponseContentRead, cts.Token);

            if (!response.IsSuccessStatusCode)
            {
                var errText = await response.Content.ReadAsStringAsync(ct);
                return Fail($"agent '{agentId}' run failed ({(int)response.StatusCode}): {Truncate(errText)}", sw.Elapsed);
            }

            var result = await response.Content.ReadFromJsonAsync<AgentRunResultDto>(Json, ct)
                ?? new AgentRunResultDto { Content = "" };

            // Return the assistant's content as task output. It must be valid, storable
            // JSON, so wrap the plain text as a JSON string (same rule as local executors).
            var output = JsonSerializer.Serialize(result.Content ?? string.Empty, Json);
            return new ExecutionResult(true, 0, result.Content ?? string.Empty, string.Empty, output, null, sw.Elapsed);
        }
        catch (Exception ex) when (ex is not OperationCanceledException)
        {
            sw.Stop();
            return Fail(ex.Message, sw.Elapsed);
        }
        finally
        {
            sw.Stop();
        }
    }

    private ExecutionResult Fail(string message, TimeSpan elapsed)
    {
        _logger.LogWarning("Agent step execution failed: {Message}", message);
        return new ExecutionResult(false, -1, string.Empty, string.Empty, null, message, elapsed);
    }

    private static string Truncate(string text)
        => text.Length > 400 ? text[..400] + "…" : text;

    private static int ParseInt(string? text, int fallback)
        => int.TryParse(text, out var value) && value > 0 ? value : fallback;

    private sealed class AgentRunRequestDto
    {
        public string Prompt { get; set; } = string.Empty;
    }

    private sealed class AgentRunResultDto
    {
        public string? Content { get; set; }
    }
}