gott/Services/AnthropicService.cs

using System;
using System.Collections.Generic;
using System.IO;
using System.IO.Compression;
using System.Net.Http;
using System.Net.Http.Headers;
using System.Text;
using System.Text.Json;
using System.Threading;
using System.Threading.Tasks;
using Gott.Models;
using Microsoft.AspNetCore.Http;

namespace Gott.Services;

public class AnthropicService
{
    private readonly IHttpClientFactory _httpFactory;
    private readonly string _token;
    private readonly ToolService _toolService;
    private readonly DatabaseService _db;
    private readonly MemoryService _memoryService;
    private readonly bool _payloadLogging;
    private readonly bool _payloadLoggingGzip;

    public AnthropicService(IHttpClientFactory httpFactory, string token, ToolService toolService, DatabaseService db, MemoryService memoryService = null, bool payloadLogging = false, bool payloadLoggingGzip = false)
    {
        _httpFactory = httpFactory;
        _token = token;
        _toolService = toolService;
        _db = db;
        _memoryService = memoryService;
        _payloadLogging = payloadLogging;
        _payloadLoggingGzip = payloadLoggingGzip;
    }

    public async Task StreamChat(
        HttpResponse response,
        string modelId,
        List<object> systemPrompt,
        List<object> messages,
        string chainId,
        string workspace,
        string sessionId = null,
        string agentId = null,
        bool useMemory = true,
        string webSearchVersion = null,
        int maxIterations = 100,
        string toolChoice = "any",
        SubAgentContext review = null,
        SubAgentContext finalizer = null,
        bool allowUserChoice = true,
        bool planMode = false,
        CancellationToken workCt = default,
        CancellationToken hardCt = default,
        CancellationToken sseCt = default)
    {
        var sw = System.Diagnostics.Stopwatch.StartNew();
        var usage = new TokenUsage();
        int iterations = 0;
        int reviewRound = 0;
        string lastStopReason = "";
        int promptTrimRetries = 0;

        var chain = new ChainRecord
        {
            ChainId = chainId,
            SessionId = sessionId,
            Model = modelId
        };

        try
        {
            while (iterations < maxIterations)
            {
                // Hard-Stop: Sofortiger Abbruch aus while-Loop (Doppel-Klick Stop)
                hardCt.ThrowIfCancellationRequested();
                workCt.ThrowIfCancellationRequested();
                iterations++;

                var payload = new Dictionary<string, object>
                {
                    ["model"] = modelId,
                    ["max_tokens"] = 16384,
                    ["cache_control"] = new { type = "ephemeral" },
                    ["messages"] = messages,
                    ["stream"] = true
                };

                if (systemPrompt != null && systemPrompt.Count > 0)
                    payload["system"] = systemPrompt;

                AddToolsToPayload(payload, workspace, agentId, useMemory, webSearchVersion, toolChoice, iterations, allowUserChoice, planMode);

                var client = CreateClient();
                var json = JsonSerializer.Serialize(payload, new JsonSerializerOptions { WriteIndented = true });
                LogPayload(json, sessionId, chainId, iterations);
                var httpContent = new StringContent(json, Encoding.UTF8, "application/json");

                using var httpResp = await client.SendAsync(
                    new HttpRequestMessage(HttpMethod.Post, "https://api.anthropic.com/v1/messages") { Content = httpContent },
                    HttpCompletionOption.ResponseHeadersRead,
                    workCt
                );

                if (!httpResp.IsSuccessStatusCode)
                {
                    var errBody = await httpResp.Content.ReadAsStringAsync(workCt);
                    var errMsg = $"API {(int)httpResp.StatusCode}: {errBody}";

                    // Auto-Retry bei "prompt too long" / "exceed context limit": History kürzen und erneut versuchen
                    if ((int)httpResp.StatusCode == 400 &&
                        (errBody.Contains("prompt is too long") || errBody.Contains("exceed context limit")) &&
                        messages.Count > 4 && promptTrimRetries < 5)
                    {
                        promptTrimRetries++;
                        int trimCount = Math.Max(2, messages.Count / 5); // 20% der Messages entfernen
                        trimCount = TrimMessagesCleanly(messages, trimCount);
                        if (trimCount > 0)
                        {
                            await TryWriteSse(response, "info", new { message = $"History zu lang – {trimCount} ältere Messages entfernt, Retry ({promptTrimRetries}/5)..." }, sseCt);
                            iterations--; // Iteration nicht zählen, war ja kein echter Versuch
                            continue; // → nächste while-Iteration mit kürzerer History
                        }
                    }

                    await _db.SaveMessage(new MessageRecord
                    {
                        ChainId = chainId, Iteration = iterations,
                        Role = "system", Type = "error",
                        Content = errMsg, IsError = true
                    });
                    await TryWriteSse(response, "error", new { error = errMsg }, sseCt);
                    await FinalizeChain(chain, usage, iterations, "error", sw);
                    await TryWriteSse(response, "done", new { chainId = chain.ChainId }, sseCt);
                    return;
                }

                var (stopReason, assistantContent, executedTools, streamErrorInfo) = await ProcessStream(response, httpResp, usage, workspace, agentId, workCt, sseCt);
                lastStopReason = stopReason;
                promptTrimRetries = 0; // Erfolgreicher API-Call → Retry-Counter resetten

                messages.Add(new { role = "assistant", content = assistantContent });

                // Assistant-Blöcke persistieren
                foreach (var block in assistantContent)
                {
                    var blockJson = JsonSerializer.Serialize(block);
                    var el = JsonSerializer.Deserialize<JsonElement>(blockJson);
                    var blockType = el.GetProperty("type").GetString() ?? "";

                    if (blockType == "text")
                    {
                        await _db.SaveMessage(new MessageRecord
                        {
                            ChainId = chainId, Iteration = iterations,
                            Role = "assistant", Type = "text",
                            Content = el.GetProperty("text").GetString()
                        });
                    }
                    else if (blockType == "tool_use")
                    {
                        await _db.SaveMessage(new MessageRecord
                        {
                            ChainId = chainId, Iteration = iterations,
                            Role = "assistant", Type = "tool_use",
                            ToolName = el.GetProperty("name").GetString(),
                            ToolCallId = el.GetProperty("id").GetString(),
                            Content = el.GetProperty("input").ToString()
                        });
                    }
                    else if (blockType == "server_tool_use")
                    {
                        await _db.SaveMessage(new MessageRecord
                        {
                            ChainId = chainId, Iteration = iterations,
                            Role = "assistant", Type = "server_tool_use",
                            Content = el.GetRawText()
                        });
                    }
                    else if (blockType == "web_search_tool_result")
                    {
                        // Raw-JSON speichern für History-Rebuild (API braucht den vollständigen Block)
                        await _db.SaveMessage(new MessageRecord
                        {
                            ChainId = chainId, Iteration = iterations,
                            Role = "assistant", Type = "web_search_result",
                            Content = el.GetRawText()
                        });
                    }
                    else if (blockType.EndsWith("_tool_result"))
                    {
                        // Generische Server-Tool-Results (bash_code_execution_tool_result, code_execution_tool_result etc.)
                        await _db.SaveMessage(new MessageRecord
                        {
                            ChainId = chainId, Iteration = iterations,
                            Role = "assistant", Type = "server_tool_result",
                            Content = el.GetRawText()
                        });
                    }
                }

                // Tool Results persistieren
                foreach (var tr in executedTools)
                {
                    await _db.SaveMessage(new MessageRecord
                    {
                        ChainId = chainId, Iteration = iterations,
                        Role = "user", Type = "tool_result",
                        ToolCallId = tr.Id,
                        Content = tr.Result,
                        IsError = tr.IsError
                    });
                }

                if (stopReason != "tool_use")
                {
                    // ── Review-Agent Phase ──
                    if (review != null && stopReason == "end_turn" && reviewRound < review.MaxReviewIterations)
                    {
                        var reviewResult = await RunReview(response, review, messages, chainId, workspace, sessionId, usage, iterations, reviewRound, workCt, sseCt);
                        iterations += reviewResult.ReviewIterationCount;
                        reviewRound++;

                        if (reviewResult.HasFeedback)
                        {
                            // Feedback als User-Message an Worker zurückgeben → nächste Iteration
                            messages.Add(new Dictionary<string, object>
                            {
                                ["role"] = "user",
                                ["content"] = new List<object> { new { type = "text", text = reviewResult.Feedback } }
                            });
                            await _db.SaveMessage(new MessageRecord
                            {
                                ChainId = chainId, Iteration = iterations,
                                Role = "user", Type = "text",
                                Content = reviewResult.Feedback
                            });
                            continue; // → nächste Worker-Iteration
                        }
                    }

                    // ── Finalizer-Agent Phase ──
                    if (finalizer != null && stopReason == "end_turn")
                    {
                        var finalizerIterations = await RunFinalizer(response, finalizer, messages, chainId, workspace, sessionId, usage, iterations, workCt, sseCt);
                        iterations += finalizerIterations;
                    }

                    var finalStopReason = string.IsNullOrEmpty(stopReason) ? "unexpected_end" : stopReason;
                    await FinalizeChain(chain, usage, iterations, finalStopReason, sw, streamErrorInfo);
                    await TryWriteSse(response, "done", new
                    {
                        model = modelId,
                        inputTokens = usage.InputTokens,
                        outputTokens = usage.OutputTokens,
                        cacheCreationTokens = usage.CacheCreationTokens,
                        cacheReadTokens = usage.CacheReadTokens,
                        iterations,
                        chainId,
                        durationMs = sw.ElapsedMilliseconds,
                        stopReason = finalStopReason,
                        streamErrorInfo,
                        planMode
                    }, sseCt);
                    return;
                }

                var toolResults = new List<object>();
                foreach (var tr in executedTools)
                {
                    toolResults.Add(new
                    {
                        type = "tool_result",
                        tool_use_id = tr.Id,
                        content = TruncateToolResult(tr.Result),
                        is_error = tr.IsError
                    });
                }

                messages.Add(new { role = "user", content = toolResults });

                // Hard-Stop Check nach Tool-Ausführung: Sofort raus statt nächste Iteration
                hardCt.ThrowIfCancellationRequested();
            }

            await FinalizeChain(chain, usage, iterations, "max_iterations", sw);
            await TryWriteSse(response, "error", new { error = "Max iterations erreicht" }, sseCt);
            await TryWriteSse(response, "done", new { chainId = chain.ChainId }, sseCt);
        }
        catch (OperationCanceledException) when (workCt.IsCancellationRequested || hardCt.IsCancellationRequested)
        {
            // Stop-Button gedrückt (workCt oder hardCt wurde gecancelled)
            await FinalizeChain(chain, usage, iterations, "stopped", sw);
            // sseCt ist hier bereits cancelled (gelinkt mit workCt), daher CancellationToken.None
            // Der SSE-Stream zum Client ist noch offen – nur der Stop-Button hat ausgelöst
            await TryWriteSse(response, "done", new
            {
                model = modelId,
                inputTokens = usage.InputTokens,
                outputTokens = usage.OutputTokens,
                cacheCreationTokens = usage.CacheCreationTokens,
                cacheReadTokens = usage.CacheReadTokens,
                iterations,
                chainId,
                durationMs = sw.ElapsedMilliseconds,
                stopReason = "stopped",
                planMode
            }, CancellationToken.None);
        }
        catch (Exception) when (workCt.IsCancellationRequested || hardCt.IsCancellationRequested)
        {
            // Stream-Dispose durch CancellationToken-Registration löst
            // ObjectDisposedException/IOException aus statt OperationCanceledException.
            // Wenn workCt/hardCt gecancelled ist, ist das ein regulärer Stop.
            await FinalizeChain(chain, usage, iterations, "stopped", sw);
            // sseCt ist hier bereits cancelled (gelinkt mit workCt), daher CancellationToken.None
            await TryWriteSse(response, "done", new
            {
                model = modelId,
                inputTokens = usage.InputTokens,
                outputTokens = usage.OutputTokens,
                cacheCreationTokens = usage.CacheCreationTokens,
                cacheReadTokens = usage.CacheReadTokens,
                iterations,
                chainId,
                durationMs = sw.ElapsedMilliseconds,
                stopReason = "stopped",
                planMode
            }, CancellationToken.None);
        }
        catch (OperationCanceledException)
        {
            // Client-Disconnect (cancellationToken gecancelled, aber NICHT workCt/hardCt)
            // Agent-Arbeit ist leider schon abgebrochen – Chain als "client_disconnected" finalisieren
            // (sollte durch TryWriteSse eigentlich nicht mehr passieren)
            await FinalizeChain(chain, usage, iterations, "client_disconnected", sw);
        }
        catch (Exception ex)
        {
            // Unerwarteter Fehler – in DB speichern und versuchen an Client zu senden
            var errMsg = $"Stream-Fehler: {ex.Message}";
            await _db.SaveMessage(new MessageRecord
            {
                ChainId = chainId, Iteration = iterations,
                Role = "system", Type = "error",
                Content = errMsg, IsError = true
            });
            await FinalizeChain(chain, usage, iterations, "error", sw);
            await TryWriteSse(response, "error", new { error = errMsg }, sseCt);
            await TryWriteSse(response, "done", new { chainId = chain.ChainId }, sseCt);
        }
    }

    // ── Preparer-Agent Logik ──

    public class PreparerResult
    {
        public bool ShouldForward { get; set; }
        public string ForwardedPrompt { get; set; }
        public string ResponseText { get; set; }
        public int IterationCount { get; set; }
    }

    /// <summary>
    /// Führt den Preparer-Agent aus. Der Preparer bekommt die User-Nachricht und die bisherige Konversation.
    /// Antwortet er mit [FORWARD] → der Text nach [FORWARD] wird an den Session-Agent weitergegeben.
    /// Sonst → seine Antwort wird als Rückfrage an den User gestreamt.
    /// </summary>
    public async Task<PreparerResult> RunPreparer(
        HttpResponse response,
        SubAgentContext preparer,
        List<object> historyMessages,
        string chainId,
        string workspace,
        string sessionId,
        CancellationToken cancellationToken)
    {
        // Preparer läuft VOR dem Worker und nutzt den Request-CancellationToken direkt.
        // Bei Client-Disconnect wird der Preparer abgebrochen (kein Sinn weiterzumachen,
        // da der Worker noch nicht gestartet hat).
        cancellationToken.ThrowIfCancellationRequested();

        await TryWriteSse(response, "preparer_start", new
        {
            agentId = preparer.AgentId,
            agentName = preparer.Agent?.Name ?? "Vorbereiter-Agent"
        }, cancellationToken);

        // Preparer bekommt die History als Kontext
        var preparerMessages = new List<object>(historyMessages);

        var usage = new TokenUsage();
        var (text, iterations) = await RunAgentChat(
            response, preparer.ModelId, preparer.SystemPrompt, preparer.AgentId,
            preparer.Workspace, preparer.UseMemory, preparer.WebSearchVersion,
            preparer.ToolChoice, preparerMessages, chainId, sessionId, usage,
            0, "preparer", preparer.Agent?.Name ?? "Vorbereiter-Agent",
            cancellationToken, cancellationToken);

        var result = new PreparerResult { IterationCount = iterations, ResponseText = text };

        // Logik: Standardmäßig wird IMMER weitergeleitet.
        // Nur wenn der Preparer explizit [QUESTION] nutzt, wird gestoppt (= Rückfrage an User).
        // [FORWARD] wird weiterhin erkannt und entfernt, ist aber nicht mehr zwingend nötig.
        if (!string.IsNullOrEmpty(text) && text.Contains("[QUESTION]"))
        {
            // Explizite Rückfrage an User
            var questionText = text.Replace("[QUESTION]", "").Trim();
            result.ShouldForward = false;
            result.ResponseText = questionText;

            await TryWriteSse(response, "preparer_done", new
            {
                agentName = preparer.Agent?.Name ?? "Vorbereiter-Agent",
                forwarded = false,
                responseText = questionText
            }, cancellationToken);
        }
        else
        {
            // Weiterleitung (Standard): [FORWARD]-Tag entfernen falls vorhanden
            var forwardedText = text ?? "";
            if (forwardedText.Contains("[FORWARD]"))
            {
                var idx = forwardedText.IndexOf("[FORWARD]");
                forwardedText = forwardedText.Substring(idx + "[FORWARD]".Length).Trim();
            }

            result.ShouldForward = true;
            result.ForwardedPrompt = forwardedText;

            await TryWriteSse(response, "preparer_done", new
            {
                agentName = preparer.Agent?.Name ?? "Vorbereiter-Agent",
                forwarded = true
            }, cancellationToken);
        }

        return result;
    }

    // ── Review-Agent Logik ──

    private class ReviewResult
    {
        public bool HasFeedback { get; set; }
        public string Feedback { get; set; }
        public int ReviewIterationCount { get; set; }
    }

    /// <summary>
    /// Führt den Review-Agent aus. Der Review-Agent bekommt die Text-Konversation
    /// und kann Tools nutzen (z.B. read_file um die Arbeit zu prüfen).
    /// Default = Approved. Nur bei explizitem [FEEDBACK]-Tag wird Feedback an den Worker zurückgegeben.
    /// reviewRound = aktuelle Runde (0-basiert), wird vom Caller hochgezählt.
    /// </summary>
    private async Task<ReviewResult> RunReview(
        HttpResponse response,
        SubAgentContext review,
        List<object> workerMessages,
        string chainId,
        string workspace,
        string sessionId,
        TokenUsage usage,
        int baseIterations,
        int reviewRound,
        CancellationToken workCt,
        CancellationToken sseCt)
    {
        workCt.ThrowIfCancellationRequested();

        // SSE: Review-Phase Start
        await TryWriteSse(response, "review_start", new
        {
            round = reviewRound + 1,
            maxRounds = review.MaxReviewIterations,
            agentId = review.AgentId,
            agentName = review.Agent?.Name ?? "Review-Agent"
        }, sseCt);

        // Review-Agent bekommt eigene Message-History
        var reviewMessages = BuildReviewMessages(workerMessages);

        // Review-Agent iteriert mit eigenen Tools
        var (reviewText, reviewIterations) = await RunAgentChat(
            response, review.ModelId, review.SystemPrompt, review.AgentId,
            workspace, review.UseMemory, review.WebSearchVersion,
            review.ToolChoice, reviewMessages, chainId, sessionId, usage,
            baseIterations, "review", review.Agent?.Name ?? "Review-Agent",
            workCt, sseCt);

        var result = new ReviewResult { ReviewIterationCount = reviewIterations };

        // Default = Approved (analog zum Preparer, wo Default = Forward ist).
        // Nur bei explizitem [FEEDBACK]-Tag wird Feedback an den Worker zurückgespielt.
        // [APPROVED] wird weiterhin erkannt, ist aber nicht mehr zwingend nötig.
        if (!string.IsNullOrEmpty(reviewText) && reviewText.Contains("[FEEDBACK]"))
        {
            // Explizites Feedback – an Worker zurückspielen
            var feedbackText = reviewText.Replace("[FEEDBACK]", "").Replace("[APPROVED]", "").Trim();
            result.HasFeedback = true;
            result.Feedback = $"[Review-Feedback von {review.Agent?.Name ?? "Review-Agent"} (Runde {reviewRound + 1}/{review.MaxReviewIterations})]:\n{feedbackText}";

            await TryWriteSse(response, "review_feedback", new
            {
                round = reviewRound + 1,
                maxRounds = review.MaxReviewIterations,
                feedback = feedbackText,
                agentName = review.Agent?.Name ?? "Review-Agent"
            }, sseCt);
        }
        else
        {
            // Default: Approved (kein [FEEDBACK]-Tag → alles OK)
            await TryWriteSse(response, "review_approved", new
            {
                round = reviewRound + 1,
                agentName = review.Agent?.Name ?? "Review-Agent"
            }, sseCt);
            result.HasFeedback = false;
        }

        return result;
    }

    // ── Finalizer-Agent Logik ──

    /// <summary>
    /// Führt den Finalizer-Agent aus. Wird einmal nach der Review-Phase (oder direkt nach dem Worker) aufgerufen.
    /// Der Finalizer kann Tools nutzen, gibt aber kein Feedback an den Worker zurück.
    /// </summary>
    private async Task<int> RunFinalizer(
        HttpResponse response,
        SubAgentContext finalizer,
        List<object> workerMessages,
        string chainId,
        string workspace,
        string sessionId,
        TokenUsage usage,
        int baseIterations,
        CancellationToken workCt,
        CancellationToken sseCt)
    {
        await TryWriteSse(response, "finalizer_start", new
        {
            agentId = finalizer.AgentId,
            agentName = finalizer.Agent?.Name ?? "Finalizer-Agent"
        }, sseCt);

        // Finalizer bekommt die Konversation als Kontext (gleiche Methode wie Review)
        var finalizerMessages = BuildReviewMessages(workerMessages);

        var (text, iterations) = await RunAgentChat(
            response, finalizer.ModelId, finalizer.SystemPrompt, finalizer.AgentId,
            finalizer.Workspace, finalizer.UseMemory, finalizer.WebSearchVersion,
            finalizer.ToolChoice, finalizerMessages, chainId, sessionId, usage,
            baseIterations, "finalizer", finalizer.Agent?.Name ?? "Finalizer-Agent",
            workCt, sseCt);

        await TryWriteSse(response, "finalizer_done", new
        {
            agentName = finalizer.Agent?.Name ?? "Finalizer-Agent"
        }, sseCt);

        return iterations;
    }

    /// <summary>
    /// Baut die Messages für den Review-Agent auf.
    /// Der Review-Agent bekommt die komplette Worker-Konversation als eine User-Message.
    /// </summary>
    private List<object> BuildReviewMessages(List<object> workerMessages)
    {
        // Nur die Text-Konversation extrahieren – Tool-Calls und Tool-Results weglassen.
        // Der Review/Finalizer-Agent hat eigene Tools und kann bei Bedarf selbst Dateien lesen.
        // Aufeinanderfolgende Messages gleicher Rolle werden zusammengeführt,
        // da die Anthropic API abwechselnde user/assistant-Rollen erfordert.
        var messages = new List<object>();
        string lastRole = null;
        var currentParts = new List<string>();

        foreach (var msg in workerMessages)
        {
            var json = JsonSerializer.Serialize(msg);
            var el = JsonSerializer.Deserialize<JsonElement>(json);

            if (!el.TryGetProperty("role", out var role)) continue;
            var roleStr = role.GetString() ?? "";

            var textParts = new List<string>();

            if (el.TryGetProperty("content", out var content))
            {
                if (content.ValueKind == JsonValueKind.String)
                {
                    var s = content.GetString() ?? "";
                    if (!string.IsNullOrWhiteSpace(s))
                        textParts.Add(s);
                }
                else if (content.ValueKind == JsonValueKind.Array)
                {
                    foreach (var block in content.EnumerateArray())
                    {
                        var blockType = block.TryGetProperty("type", out var bt) ? bt.GetString() ?? "" : "";
                        if (blockType == "text")
                        {
                            var text = block.TryGetProperty("text", out var t) ? t.GetString() ?? "" : "";
                            if (!string.IsNullOrWhiteSpace(text))
                                textParts.Add(text);
                        }
                    }
                }
            }

            if (textParts.Count == 0) continue;

            if (roleStr == lastRole)
            {
                // Gleiche Rolle wie vorher → zusammenführen
                currentParts.AddRange(textParts);
            }
            else
            {
                // Neue Rolle → vorherige Message abschließen
                if (lastRole != null && currentParts.Count > 0)
                {
                    messages.Add(new Dictionary<string, object>
                    {
                        ["role"] = lastRole,
                        ["content"] = new List<object> { new { type = "text", text = string.Join("\n\n", currentParts) } }
                    });
                }
                lastRole = roleStr;
                currentParts = new List<string>(textParts);
            }
        }

        // Letzte angesammelte Message abschließen
        if (lastRole != null && currentParts.Count > 0)
        {
            messages.Add(new Dictionary<string, object>
            {
                ["role"] = lastRole,
                ["content"] = new List<object> { new { type = "text", text = string.Join("\n\n", currentParts) } }
            });
        }

        // Anthropic erfordert: erste Message = user, letzte Message = user,
        // und abwechselnde Rollen (durch die Zusammenführung oben bereits sichergestellt).
        if (messages.Count > 0)
        {
            var first = (Dictionary<string, object>)messages[0];
            if ((string)first["role"] == "assistant")
            {
                messages.Insert(0, new Dictionary<string, object>
                {
                    ["role"] = "user",
                    ["content"] = new List<object> { new { type = "text", text = "Hier ist die bisherige Konversation. Bitte überprüfe den aktuellen Zustand." } }
                });
            }

            var last = (Dictionary<string, object>)messages[messages.Count - 1];
            if ((string)last["role"] == "assistant")
            {
                messages.Add(new Dictionary<string, object>
                {
                    ["role"] = "user",
                    ["content"] = new List<object> { new { type = "text", text = "Bitte überprüfe ob die Aufgabe korrekt erledigt wurde." } }
                });
            }
        }

        // Falls keine Text-Messages vorhanden, Fallback mit Hinweis
        if (messages.Count == 0)
        {
            messages.Add(new Dictionary<string, object>
            {
                ["role"] = "user",
                ["content"] = new List<object> { new { type = "text", text = "Die bisherige Konversation enthielt nur Tool-Aufrufe ohne Text. Bitte überprüfe den aktuellen Zustand des Workspaces mit deinen Tools." } }
            });
        }

        return messages;
    }

    /// <summary>
    /// Generischer Agent-Chat (non-streaming für den Agent, aber SSE-Events für den Client).
    /// Wird für Review-Agent und Finalizer-Agent verwendet.
    /// agentType steuert SSE-Event-Prefixe und DB-Message-Tags ("review" oder "finalizer").
    /// </summary>
    private async Task<(string text, int iterations)> RunAgentChat(
        HttpResponse response,
        string modelId,
        List<object> systemPrompt,
        string agentId,
        string agentWorkspace,
        bool useMemory,
        string webSearchVersion,
        string agentToolChoice,
        List<object> messages,
        string chainId,
        string sessionId,
        TokenUsage usage,
        int baseIterations,
        string agentType,
        string agentName,
        CancellationToken workCt,
        CancellationToken sseCt = default)
    {
        int iterations = 0;
        var finalText = new StringBuilder();
        int maxToolIterations = 20;
        var tag = agentType == "preparer" ? "Preparer" : agentType == "finalizer" ? "Finalizer" : "Review";

        while (iterations < maxToolIterations)
        {
            workCt.ThrowIfCancellationRequested();
            iterations++;

            var payload = new Dictionary<string, object>
            {
                ["model"] = modelId,
                ["max_tokens"] = 8192,
                ["cache_control"] = new { type = "ephemeral" },
                ["messages"] = messages,
                ["stream"] = false
            };

            if (systemPrompt != null && systemPrompt.Count > 0)
                payload["system"] = systemPrompt;

            AddToolsToPayload(payload, agentWorkspace, agentId, useMemory, webSearchVersion, agentToolChoice, iterations);

            var client = CreateClient();
            var json = JsonSerializer.Serialize(payload, new JsonSerializerOptions { WriteIndented = true });
            LogPayload(json, sessionId, chainId, baseIterations + iterations);
            var httpContent = new StringContent(json, Encoding.UTF8, "application/json");

            using var httpResp = await client.SendAsync(
                new HttpRequestMessage(HttpMethod.Post, "https://api.anthropic.com/v1/messages") { Content = httpContent },
                cancellationToken: workCt);

            var respBody = await httpResp.Content.ReadAsStringAsync(workCt);

            if (!httpResp.IsSuccessStatusCode)
            {
                // Auto-Retry bei "prompt too long" in Sub-Agents
                if ((int)httpResp.StatusCode == 400 &&
                    (respBody.Contains("prompt is too long") || respBody.Contains("exceed context limit")) &&
                    messages.Count > 4)
                {
                    int trimCount = Math.Max(2, messages.Count / 5);
                    trimCount = TrimMessagesCleanly(messages, trimCount);
                    if (trimCount > 0)
                    {
                        await TryWriteSse(response, "info", new { message = $"{tag}: History zu lang – {trimCount} Messages entfernt, Retry..." }, sseCt);
                        iterations--;
                        continue;
                    }
                }
                return ($"{tag}-Agent API Fehler: {(int)httpResp.StatusCode}: {respBody}", iterations);
            }

            var resp = JsonSerializer.Deserialize<JsonElement>(respBody);

            if (resp.TryGetProperty("usage", out var u))
                TrackUsage(usage, u);

            var stopReason = resp.TryGetProperty("stop_reason", out var sr) ? sr.GetString() ?? "" : "";
            var content = resp.GetProperty("content");

            var assistantContent = new List<object>();
            var toolResults = new List<object>();

            foreach (var block in content.EnumerateArray())
            {
                var blockType = block.GetProperty("type").GetString() ?? "";

                if (blockType == "text")
                {
                    var text = block.GetProperty("text").GetString() ?? "";
                    assistantContent.Add(new { type = "text", text });
                    finalText.Append(text);

                    await TryWriteSse(response, $"{agentType}_text", new
                    {
                        text,
                        agentName,
                        agentId
                    }, sseCt);

                    await _db.SaveMessage(new MessageRecord
                    {
                        ChainId = chainId,
                        Iteration = baseIterations + iterations,
                        Role = "assistant",
                        Type = "text",
                        Content = $"[{tag}: {agentName}] {text}",
                        Tag = agentType
                    });
                }
                else if (blockType == "tool_use")
                {
                    var toolId = block.GetProperty("id").GetString() ?? "";
                    var toolName = block.GetProperty("name").GetString() ?? "";
                    var toolInput = block.GetProperty("input");

                    assistantContent.Add(new { type = "tool_use", id = toolId, name = toolName, input = toolInput });

                    await TryWriteSse(response, $"{agentType}_tool_use", new
                    {
                        toolName,
                        toolId,
                        input = toolInput.ToString(),
                        agentName
                    }, sseCt);

                    await _db.SaveMessage(new MessageRecord
                    {
                        ChainId = chainId,
                        Iteration = baseIterations + iterations,
                        Role = "assistant",
                        Type = "tool_use",
                        Content = toolInput.GetRawText(),
                        ToolName = toolName,
                        ToolCallId = toolId,
                        Tag = agentType
                    });

                    string toolResult;
                    bool isError = false;
                    try { toolResult = await _toolService.ExecuteTool(toolName, toolInput, agentWorkspace, _memoryService, agentId, workCt); }
                    catch (OperationCanceledException) { throw; }
                    catch (Exception ex) { toolResult = $"Fehler: {ex.Message}"; isError = true; }

                    toolResults.Add(new
                    {
                        type = "tool_result",
                        tool_use_id = toolId,
                        content = TruncateToolResult(toolResult),
                        is_error = isError
                    });

                    await TryWriteSse(response, $"{agentType}_tool_end", new
                    {
                        toolName,
                        toolId,
                        input = toolInput.ToString(),
                        result = toolResult,
                        isError,
                        agentName
                    }, sseCt);

                    await _db.SaveMessage(new MessageRecord
                    {
                        ChainId = chainId,
                        Iteration = baseIterations + iterations,
                        Role = "user",
                        Type = "tool_result",
                        Content = toolResult,
                        ToolName = toolName,
                        ToolCallId = toolId,
                        IsError = isError,
                        Tag = agentType
                    });
                }
                else if (blockType == "web_search_tool_result")
                {
                    assistantContent.Add(JsonSerializer.Deserialize<JsonElement>(block.GetRawText()));
                    // Raw-JSON speichern für History-Rebuild (API braucht den vollständigen Block)
                    await _db.SaveMessage(new MessageRecord
                    {
                        ChainId = chainId,
                        Iteration = baseIterations + iterations,
                        Role = "assistant",
                        Type = "web_search_result",
                        Content = block.GetRawText(),
                        Tag = agentType
                    });
                }
                else if (blockType == "server_tool_use")
                {
                    var (toolId, toolName, inputEl) = ParseServerToolUse(block);
                    assistantContent.Add(new { type = "server_tool_use", id = toolId, name = toolName, input = inputEl });
                }
                else if (blockType.EndsWith("_tool_result"))
                {
                    // Generische Server-Tool-Results (bash_code_execution_tool_result, code_execution_tool_result etc.)
                    assistantContent.Add(JsonSerializer.Deserialize<JsonElement>(block.GetRawText()));
                    await _db.SaveMessage(new MessageRecord
                    {
                        ChainId = chainId,
                        Iteration = baseIterations + iterations,
                        Role = "assistant", Type = "server_tool_result",
                        Content = block.GetRawText(),
                        Tag = agentType
                    });
                }
            }

            messages.Add(new Dictionary<string, object> { ["role"] = "assistant", ["content"] = assistantContent });

            if (stopReason == "end_turn" || stopReason == "stop_sequence" || toolResults.Count == 0)
                break;

            messages.Add(new Dictionary<string, object> { ["role"] = "user", ["content"] = toolResults });
        }

        return (finalText.ToString(), iterations);
    }

    /// <summary>
    /// Non-Streaming Chat für Telegram. Gibt den finalen Antworttext zurück.
    /// onIteration wird nach jeder Iteration aufgerufen (z.B. für Typing-Indicator).
    /// </summary>
    public async Task<string> ChatNonStreaming(
        ChatContext ctx,
        List<object> messages,
        Func<int, Task> onIteration = null)
    {
        var sw = System.Diagnostics.Stopwatch.StartNew();
        var usage = new TokenUsage();
        int iterations = 0;
        var finalText = new StringBuilder();

        var chain = new ChainRecord
        {
            ChainId = ctx.ChainId,
            SessionId = ctx.SessionId,
            Model = ctx.ModelId
        };

        try
        {
            while (iterations < ctx.MaxIterations)
            {
                iterations++;
                onIteration?.Invoke(iterations);

                var payload = new Dictionary<string, object>
                {
                    ["model"] = ctx.ModelId,
                    ["max_tokens"] = 16384,
                    ["cache_control"] = new { type = "ephemeral" },
                    ["messages"] = messages,
                    ["stream"] = false
                };

                if (ctx.SystemPrompt != null && ctx.SystemPrompt.Count > 0)
                    payload["system"] = ctx.SystemPrompt;

                AddToolsToPayload(payload, ctx.Workspace, ctx.AgentId, ctx.UseMemory, ctx.WebSearchVersion, ctx.ToolChoice, iterations);

                var client = CreateClient();
                var json = JsonSerializer.Serialize(payload, new JsonSerializerOptions { WriteIndented = true });
                LogPayload(json, ctx.SessionId, ctx.ChainId, iterations);
                var httpContent = new StringContent(json, Encoding.UTF8, "application/json");

                using var httpResp = await client.SendAsync(
                    new HttpRequestMessage(HttpMethod.Post, "https://api.anthropic.com/v1/messages") { Content = httpContent });

                var respBody = await httpResp.Content.ReadAsStringAsync();

                if (!httpResp.IsSuccessStatusCode)
                {
                    // Auto-Retry bei "prompt too long" in Non-Streaming
                    if ((int)httpResp.StatusCode == 400 &&
                        (respBody.Contains("prompt is too long") || respBody.Contains("exceed context limit")) &&
                        messages.Count > 4)
                    {
                        int trimCount = Math.Max(2, messages.Count / 5);
                        trimCount = TrimMessagesCleanly(messages, trimCount);
                        if (trimCount > 0)
                        {
                            iterations--;
                            continue;
                        }
                    }

                    var errMsg = $"API {(int)httpResp.StatusCode}: {respBody}";
                    await _db.SaveMessage(new MessageRecord
                    {
                        ChainId = ctx.ChainId, Iteration = iterations,
                        Role = "system", Type = "error",
                        Content = errMsg, IsError = true
                    });
                    await FinalizeChain(chain, usage, iterations, "error", sw);
                    throw new Exception($"Claude API Fehler: {(int)httpResp.StatusCode}");
                }

                var resp = JsonSerializer.Deserialize<JsonElement>(respBody);

                // Usage tracken
                if (resp.TryGetProperty("usage", out var u))
                    TrackUsage(usage, u);

                var stopReason = resp.TryGetProperty("stop_reason", out var sr) ? sr.GetString() ?? "" : "";
                var content = resp.GetProperty("content");

                var assistantContent = new List<object>();
                var toolResults = new List<object>();

                foreach (var block in content.EnumerateArray())
                {
                    var blockType = block.GetProperty("type").GetString() ?? "";

                    if (blockType == "text")
                    {
                        var text = block.GetProperty("text").GetString() ?? "";
                        assistantContent.Add(new { type = "text", text });
                        finalText.Append(text);

                        await _db.SaveMessage(new MessageRecord
                        {
                            ChainId = ctx.ChainId, Iteration = iterations,
                            Role = "assistant", Type = "text",
                            Content = text
                        });
                    }
                    else if (blockType == "tool_use")
                    {
                        var toolId = block.GetProperty("id").GetString() ?? "";
                        var toolName = block.GetProperty("name").GetString() ?? "";
                        var toolInput = block.GetProperty("input");

                        assistantContent.Add(new { type = "tool_use", id = toolId, name = toolName, input = toolInput });

                        await _db.SaveMessage(new MessageRecord
                        {
                            ChainId = ctx.ChainId, Iteration = iterations,
                            Role = "assistant", Type = "tool_use",
                            Content = toolInput.GetRawText(),
                            ToolName = toolName, ToolCallId = toolId
                        });

                        string toolResult;
                        bool isError = false;
                        try { toolResult = await _toolService.ExecuteTool(toolName, toolInput, ctx.Workspace, _memoryService, ctx.AgentId); }
                        catch (OperationCanceledException) { throw; }
                        catch (Exception ex) { toolResult = $"Fehler: {ex.Message}"; isError = true; }

                        toolResults.Add(new
                        {
                            type = "tool_result",
                            tool_use_id = toolId,
                            content = TruncateToolResult(toolResult),
                            is_error = isError
                        });

                        await _db.SaveMessage(new MessageRecord
                        {
                            ChainId = ctx.ChainId, Iteration = iterations,
                            Role = "user", Type = "tool_result",
                            Content = toolResult,
                            ToolName = toolName, ToolCallId = toolId,
                            IsError = isError
                        });
                    }
                    else if (blockType == "web_search_tool_result")
                    {
                        assistantContent.Add(JsonSerializer.Deserialize<JsonElement>(block.GetRawText()));
                        // Raw-JSON speichern für History-Rebuild (API braucht den vollständigen Block)
                        await _db.SaveMessage(new MessageRecord
                        {
                            ChainId = ctx.ChainId, Iteration = iterations,
                            Role = "assistant", Type = "web_search_result",
                            Content = block.GetRawText()
                        });
                    }
                    else if (blockType == "server_tool_use")
                    {
                        var (toolId, toolName, inputEl) = ParseServerToolUse(block);
                        assistantContent.Add(new { type = "server_tool_use", id = toolId, name = toolName, input = inputEl });
                    }
                    else if (blockType.EndsWith("_tool_result"))
                    {
                        // Generische Server-Tool-Results (bash_code_execution_tool_result, code_execution_tool_result etc.)
                        assistantContent.Add(JsonSerializer.Deserialize<JsonElement>(block.GetRawText()));
                        await _db.SaveMessage(new MessageRecord
                        {
                            ChainId = ctx.ChainId, Iteration = iterations,
                            Role = "assistant", Type = "server_tool_result",
                            Content = block.GetRawText()
                        });
                    }
                }

                messages.Add(new Dictionary<string, object> { ["role"] = "assistant", ["content"] = assistantContent });

                if (stopReason == "end_turn" || stopReason == "stop_sequence" || toolResults.Count == 0)
                {
                    await FinalizeChain(chain, usage, iterations, stopReason, sw);
                    break;
                }

                messages.Add(new Dictionary<string, object> { ["role"] = "user", ["content"] = toolResults });
            }
        }
        catch
        {
            if (iterations > 0)
                await FinalizeChain(chain, usage, iterations, "error", sw);
            throw;
        }

        return finalText.Length > 0 ? finalText.ToString() : "(Keine Textantwort)";
    }

    private void LogPayload(string json, string sessionId, string chainId, int iteration)
    {
        if (!_payloadLogging) return;
        try
        {
            var logsDir = Path.Combine(AppContext.BaseDirectory, "data", "logs");
            Directory.CreateDirectory(logsDir);
            var timestamp = DateTimeOffset.UtcNow.ToUnixTimeSeconds();
            var baseName = $"{timestamp}-{sessionId}-{chainId}-iter{iteration}-payload";

            if (_payloadLoggingGzip)
            {
                var filePath = Path.Combine(logsDir, baseName + ".json.gz");
                using var fs = File.Create(filePath);
                using var gz = new GZipStream(fs, CompressionLevel.Fastest);
                using var writer = new StreamWriter(gz, Encoding.UTF8);
                writer.Write(json);
            }
            else
            {
                File.WriteAllText(Path.Combine(logsDir, baseName + ".json"), json);
            }
        }
        catch { /* Logging darf nie die App crashen */ }
    }

    private async Task FinalizeChain(ChainRecord chain, TokenUsage usage, int iterations, string stopReason, System.Diagnostics.Stopwatch sw, string streamErrorInfo = null)
    {
        chain.InputTokens = usage.InputTokens;
        chain.OutputTokens = usage.OutputTokens;
        chain.CacheCreationTokens = usage.CacheCreationTokens;
        chain.CacheReadTokens = usage.CacheReadTokens;
        chain.Iterations = iterations;
        chain.StopReason = stopReason;
        chain.StreamErrorInfo = streamErrorInfo;
        chain.DurationMs = sw.ElapsedMilliseconds;
        await _db.UpdateChain(chain);
    }

    private async Task<(string StopReason, List<object> Content, List<ToolResult> Tools, string StreamErrorInfo)> ProcessStream(
        HttpResponse response,
        HttpResponseMessage httpResp,
        TokenUsage usage,
        string workspace,
        string agentId = null,
        CancellationToken workCt = default,
        CancellationToken sseCt = default)
    {
        var stream = await httpResp.Content.ReadAsStreamAsync(workCt);
        // Bei Cancellation den Stream disposen, damit ReadLineAsync sofort abbricht.
        // StreamReader.ReadLineAsync propagiert den CancellationToken nicht zuverlässig
        // an den underlying NetworkStream – der Read blockiert sonst bis Anthropic
        // das nächste Chunk sendet (kann bei extended thinking Minuten dauern).
        using var ctReg = workCt.Register(() => { try { stream.Dispose(); } catch { } });
        using var reader = new StreamReader(stream);

        var stopReason = "";
        var assistantContent = new List<object>();
        var currentToolId = "";
        var currentToolName = "";
        var currentToolInputJson = new StringBuilder();
        var currentTextBlock = new StringBuilder();

        var blockTypes = new Dictionary<int, string>();
        var webSearchResultBlocks = new Dictionary<int, object>();
        var webSearchSources = new Dictionary<int, List<object>>();
        var serverToolResultBlocks = new Dictionary<int, object>(); // bash_code_execution_tool_result, code_execution_tool_result etc.
        var executedTools = new List<ToolResult>();
        var streamErrors = new List<object>();

        string line;
        while ((line = await reader.ReadLineAsync(workCt)) != null)
        {
            if (!line.StartsWith("data: ")) continue;
            var data = line["data: ".Length..];
            if (string.IsNullOrWhiteSpace(data)) continue;

            JsonElement evt;
            try { evt = JsonSerializer.Deserialize<JsonElement>(data); }
            catch { continue; }

            var evtType = evt.TryGetProperty("type", out var t) ? t.GetString() ?? "" : "";

            switch (evtType)
            {
                case "message_start":
                {
                    if (evt.TryGetProperty("message", out var msg) &&
                        msg.TryGetProperty("usage", out var u))
                        TrackUsage(usage, u);
                    break;
                }

                case "content_block_start":
                {
                    var idx = evt.GetProperty("index").GetInt32();
                    var cb = evt.GetProperty("content_block");
                    var cbType = cb.GetProperty("type").GetString() ?? "";
                    blockTypes[idx] = cbType;

                    if (cbType == "text")
                    {
                        currentTextBlock.Clear();
                        await TryWriteSse(response, "text_start", new { index = idx }, sseCt);
                    }
                    else if (cbType == "tool_use")
                    {
                        if (currentTextBlock.Length > 0)
                        {
                            assistantContent.Add(new { type = "text", text = currentTextBlock.ToString() });
                            currentTextBlock.Clear();
                        }
                        currentToolId = cb.GetProperty("id").GetString() ?? "";
                        currentToolName = cb.GetProperty("name").GetString() ?? "";
                        currentToolInputJson.Clear();
                        await TryWriteSse(response, "tool_start", new { name = currentToolName, id = currentToolId }, sseCt);
                    }
                    else if (cbType == "server_tool_use")
                    {
                        if (currentTextBlock.Length > 0)
                        {
                            assistantContent.Add(new { type = "text", text = currentTextBlock.ToString() });
                            currentTextBlock.Clear();
                        }
                        currentToolId = cb.GetProperty("id").GetString() ?? "";
                        currentToolName = cb.GetProperty("name").GetString() ?? "";
                        currentToolInputJson.Clear();
                        await TryWriteSse(response, "web_search_start", new { name = currentToolName, id = currentToolId }, sseCt);
                    }
                    else if (cbType == "web_search_tool_result")
                    {
                        var rawBlock = JsonSerializer.Deserialize<JsonElement>(cb.GetRawText());
                        webSearchResultBlocks[idx] = rawBlock;
                        webSearchSources[idx] = ExtractWebSearchSources(cb);
                    }
                    else if (cbType.EndsWith("_tool_result"))
                    {
                        // Generische Server-Tool-Results (bash_code_execution_tool_result, code_execution_tool_result etc.)
                        serverToolResultBlocks[idx] = JsonSerializer.Deserialize<JsonElement>(cb.GetRawText());
                    }
                    break;
                }

                case "content_block_delta":
                {
                    var delta = evt.GetProperty("delta");
                    var deltaType = delta.GetProperty("type").GetString() ?? "";

                    if (deltaType == "text_delta")
                    {
                        var text = delta.GetProperty("text").GetString() ?? "";
                        currentTextBlock.Append(text);
                        await TryWriteSse(response, "text_delta", new { text }, sseCt);
                    }
                    else if (deltaType == "input_json_delta")
                    {
                        var partial = delta.GetProperty("partial_json").GetString() ?? "";
                        currentToolInputJson.Append(partial);
                    }
                    break;
                }

                case "content_block_stop":
                {
                    var idx = evt.GetProperty("index").GetInt32();
                    if (blockTypes.TryGetValue(idx, out var bt))
                    {
                        if (bt == "text")
                        {
                            var text = currentTextBlock.ToString();
                            if (text.Length > 0)
                            {
                                assistantContent.Add(new { type = "text", text });
                                await TryWriteSse(response, "text_end", new { }, sseCt);
                            }
                            currentTextBlock.Clear();
                        }
                        else if (bt == "tool_use")
                        {
                            JsonElement inputEl;
                            try { inputEl = JsonSerializer.Deserialize<JsonElement>(currentToolInputJson.ToString()); }
                            catch { inputEl = JsonSerializer.Deserialize<JsonElement>("{}"); }

                            assistantContent.Add(new
                            {
                                type = "tool_use",
                                id = currentToolId,
                                name = currentToolName,
                                input = inputEl
                            });

                            // ask_user_choice: NICHT ausführen, sondern als SSE-Event an Client senden
                            if (currentToolName == ToolService.AskUserChoiceToolName)
                            {
                                var question = inputEl.TryGetProperty("question", out var q) ? q.GetString() ?? "" : "";
                                var choices = new List<string>();
                                if (inputEl.TryGetProperty("choices", out var c))
                                {
                                    if (c.ValueKind == JsonValueKind.Array)
                                    {
                                        foreach (var ch in c.EnumerateArray())
                                            choices.Add(ch.GetString() ?? "");
                                    }
                                    else if (c.ValueKind == JsonValueKind.String)
                                    {
                                        // Fallback: Claude hat einen einzelnen String statt Array gesendet
                                        choices.Add(c.GetString() ?? "");
                                    }
                                }

                                await TryWriteSse(response, "user_choice", new
                                {
                                    toolCallId = currentToolId,
                                    question,
                                    choices
                                }, sseCt);

                                // Stream restlich konsumieren ohne weitere Verarbeitung
                                // (Claude sendet nach dem tool_use noch message_delta mit stop_reason)
                                stopReason = "awaiting_user_choice";
                                continue;
                            }

                            // Cancellation prüfen bevor Tool ausgeführt wird
                            workCt.ThrowIfCancellationRequested();

                            string toolResult;
                            bool isError = false;
                            try { toolResult = await _toolService.ExecuteTool(currentToolName, inputEl, workspace, _memoryService, agentId, workCt); }
                            catch (OperationCanceledException) { throw; }
                            catch (Exception ex) { toolResult = $"Fehler: {ex.Message}"; isError = true; }

                            workCt.ThrowIfCancellationRequested();
                            executedTools.Add(new ToolResult(currentToolId, toolResult, isError));

                            await TryWriteSse(response, "tool_end", new
                            {
                                name = currentToolName,
                                input = currentToolInputJson.ToString(),
                                result = toolResult,
                                isError
                            }, sseCt);
                        }
                        else if (bt == "server_tool_use")
                        {
                            // Server-side tool (Web Search) – Anthropic führt es aus, wir speichern nur die Anfrage
                            JsonElement inputEl;
                            try { inputEl = JsonSerializer.Deserialize<JsonElement>(currentToolInputJson.ToString()); }
                            catch { inputEl = JsonSerializer.Deserialize<JsonElement>("{}"); }

                            assistantContent.Add(new
                            {
                                type = "server_tool_use",
                                id = currentToolId,
                                name = currentToolName,
                                input = inputEl
                            });
                        }
                        else if (bt == "web_search_tool_result")
                        {
                            // Web-Search-Ergebnis von Anthropic – als Raw-JSON in assistantContent aufnehmen
                            assistantContent.Add(webSearchResultBlocks[idx]);
                            var sources = webSearchSources.TryGetValue(idx, out var s) ? s : new List<object>();
                            await TryWriteSse(response, "web_search_end", new { id = currentToolId, sources }, sseCt);
                        }
                        else if (bt.EndsWith("_tool_result") && serverToolResultBlocks.TryGetValue(idx, out var srvResultBlock))
                        {
                            // Generische Server-Tool-Results (bash_code_execution_tool_result, code_execution_tool_result etc.)
                            assistantContent.Add(srvResultBlock);
                        }
                    }
                    break;
                }

                case "message_delta":
                {
                    if (evt.TryGetProperty("delta", out var d))
                    {
                        if (d.TryGetProperty("stop_reason", out var sr))
                        {
                            // Nicht überschreiben wenn bereits auf awaiting_user_choice gesetzt
                            if (stopReason != "awaiting_user_choice")
                                stopReason = sr.GetString() ?? "";

                            // Bei problematischen stop_reasons Debug-Infos sammeln
                            if (stopReason is "unexpected_end" or "max_tokens")
                            {
                                var debugInfo = new Dictionary<string, object>
                                {
                                    ["event"] = "message_delta",
                                    ["stop_reason"] = stopReason,
                                    ["output_tokens"] = usage.OutputTokens,
                                    ["text_length"] = currentTextBlock.Length,
                                    ["timestamp"] = DateTime.UtcNow.ToString("o")
                                };
                                if (d.TryGetProperty("stop_sequence", out var ss) && ss.ValueKind != JsonValueKind.Null)
                                    debugInfo["stop_sequence"] = ss.GetString();
                                streamErrors.Add(debugInfo);
                            }
                        }
                    }
                    if (evt.TryGetProperty("usage", out var u2) &&
                        u2.TryGetProperty("output_tokens", out var ot))
                        usage.OutputTokens += ot.GetInt64();
                    break;
                }

                case "error":
                {
                    // API Error-Event (z.B. overloaded_error)
                    var errorInfo = new Dictionary<string, object>
                    {
                        ["event"] = "error",
                        ["timestamp"] = DateTime.UtcNow.ToString("o"),
                        ["raw"] = data
                    };
                    if (evt.TryGetProperty("error", out var err))
                    {
                        if (err.TryGetProperty("type", out var errType))
                            errorInfo["error_type"] = errType.GetString();
                        if (err.TryGetProperty("message", out var errMsg))
                            errorInfo["error_message"] = errMsg.GetString();
                    }
                    streamErrors.Add(errorInfo);
                    break;
                }
            }
        }

        if (currentTextBlock.Length > 0)
            assistantContent.Add(new { type = "text", text = currentTextBlock.ToString() });

        // Stream-Fehler als JSON serialisieren (null wenn keine Fehler)
        string streamErrorInfo = streamErrors.Count > 0
            ? JsonSerializer.Serialize(streamErrors)
            : null;

        return (stopReason, assistantContent, executedTools, streamErrorInfo);
    }

    // ── Shared Helpers ──

    internal const int MaxToolResultChars = 20_000;

    internal static string TruncateToolResult(string result)
    {
        return result.Length > MaxToolResultChars
            ? result[..MaxToolResultChars] + $"\n\n[... gekürzt, {result.Length - MaxToolResultChars} weitere Zeichen ...]"
            : result;
    }

    /// <summary>
    /// Extrahiert Web-Search-Sources aus einem web_search_tool_result-Block.
    /// </summary>
    private static List<object> ExtractWebSearchSources(JsonElement block)
    {
        var sources = new List<object>();
        if (block.TryGetProperty("content", out var searchContent) && searchContent.ValueKind == JsonValueKind.Array)
        {
            foreach (var item in searchContent.EnumerateArray())
            {
                var itemType = item.TryGetProperty("type", out var it) ? it.GetString() : "";
                if (itemType == "web_search_result")
                {
                    var url = item.TryGetProperty("url", out var u) ? u.GetString() ?? "" : "";
                    var title = item.TryGetProperty("title", out var tt) ? tt.GetString() ?? "" : "";
                    if (!string.IsNullOrEmpty(url))
                        sources.Add(new { url, title });
                }
            }
        }
        return sources;
    }

    /// <summary>
    /// Parst einen server_tool_use-Block und gibt (id, name, input) zurück.
    /// </summary>
    private static (string id, string name, JsonElement input) ParseServerToolUse(JsonElement block)
    {
        var toolId = block.GetProperty("id").GetString() ?? "";
        var toolName = block.GetProperty("name").GetString() ?? "";
        JsonElement inputEl;
        try { inputEl = block.GetProperty("input"); }
        catch { inputEl = JsonSerializer.Deserialize<JsonElement>("{}"); }
        return (toolId, toolName, inputEl);
    }

    /// <summary>
    /// Aktualisiert den TokenUsage-Tracker aus einem usage-JSON-Element.
    /// </summary>
    private static void TrackUsage(TokenUsage usage, JsonElement u)
    {
        if (u.TryGetProperty("input_tokens", out var it)) usage.InputTokens += it.GetInt64();
        if (u.TryGetProperty("output_tokens", out var ot)) usage.OutputTokens += ot.GetInt64();
        if (u.TryGetProperty("cache_creation_input_tokens", out var cc)) usage.CacheCreationTokens += cc.GetInt64();
        if (u.TryGetProperty("cache_read_input_tokens", out var cr)) usage.CacheReadTokens += cr.GetInt64();
    }

    /// <summary>
    /// Fügt Tool-Definitionen und WebSearch zum Payload hinzu.
    /// </summary>
    private void AddToolsToPayload(Dictionary<string, object> payload, string workspace, string agentId, bool useMemory, string webSearchVersion, string toolChoice, int iteration, bool allowUserChoice = false, bool planMode = false)
    {
        var hasWorkspace = !string.IsNullOrEmpty(workspace);
        var includeMemory = useMemory && _memoryService != null && !string.IsNullOrEmpty(agentId);
        var hasTools = hasWorkspace || includeMemory;
        var hasWebSearch = !string.IsNullOrEmpty(webSearchVersion);

        if (!hasTools && !hasWebSearch) return;

        var tools = new List<object>();
        if (hasTools)
        {
            var allTools = _toolService.GetToolDefinitions(includeFileTools: hasWorkspace, includeMemory: includeMemory, includeUserChoice: allowUserChoice);
            if (planMode)
                allTools = _toolService.FilterReadOnlyTools(allTools);
            tools.AddRange(allTools);
        }
        if (hasWebSearch)
            tools.Add(new { type = webSearchVersion, name = "web_search", max_uses = 5 });
        payload["tools"] = tools;
        var effectiveChoice = (!planMode && toolChoice == "any" && iteration == 1) ? "any" : "auto";
        payload["tool_choice"] = new { type = effectiveChoice };
    }

    /// <summary>
    /// Entfernt Messages vom Anfang der Liste, ohne tool_use/tool_result-Paare zu trennen.
    /// Gibt die tatsächliche Anzahl entfernter Messages zurück.
    /// </summary>
    private static int TrimMessagesCleanly(List<object> messages, int targetRemoveCount)
    {
        if (targetRemoveCount <= 0 || targetRemoveCount >= messages.Count)
            return 0;

        int removeCount = targetRemoveCount;

        // Sicherstellen, dass wir nicht mitten in einem tool_use/tool_result-Paar abschneiden
        while (removeCount < messages.Count - 1)
        {
            var firstRemaining = messages[removeCount];
            var firstRemainingJson = JsonSerializer.Serialize(firstRemaining);

            // Wenn die erste verbleibende Message ein tool_result ist → vorherige auch behalten
            if (firstRemainingJson.Contains("\"tool_result\""))
            {
                removeCount--;
                if (removeCount <= 0) return 0;
                continue;
            }

            // Wenn die letzte entfernte Message ein assistant mit tool_use ist → tool_result auch entfernen
            var lastRemovedJson = JsonSerializer.Serialize(messages[removeCount - 1]);
            if (lastRemovedJson.Contains("\"tool_use\"") && lastRemovedJson.Contains("\"assistant\""))
            {
                removeCount++;
                continue;
            }

            break;
        }

        if (removeCount > 0 && removeCount < messages.Count)
        {
            messages.RemoveRange(0, removeCount);
            return removeCount;
        }
        return 0;
    }

    private HttpClient CreateClient()
    {
        var client = _httpFactory.CreateClient();
        client.DefaultRequestHeaders.Authorization = new AuthenticationHeaderValue("Bearer", _token);
        client.DefaultRequestHeaders.Add("anthropic-version", "2023-06-01");
        client.DefaultRequestHeaders.Add("anthropic-beta", "claude-code-20250219,oauth-2025-04-20");
        client.DefaultRequestHeaders.Add("User-Agent", "claude-cli/2.1.2 (external, cli)");
        client.DefaultRequestHeaders.Add("x-app", "cli");
        client.DefaultRequestHeaders.Accept.Add(new MediaTypeWithQualityHeaderValue("application/json"));
        client.DefaultRequestHeaders.Add("anthropic-dangerous-direct-browser-access", "true");
        return client;
    }

    public static async Task WriteSseEvent(HttpResponse response, string eventType, object data, CancellationToken ct = default)
    {
        ct.ThrowIfCancellationRequested();
        var json = JsonSerializer.Serialize(data);
        await response.WriteAsync($"event: {eventType}\ndata: {json}\n\n", ct);
        await response.Body.FlushAsync(ct);
    }

    private static async Task TryWriteSse(HttpResponse response, string eventType, object data, CancellationToken ct)
    {
        try
        {
            if (!ct.IsCancellationRequested)
                await WriteSseEvent(response, eventType, data, ct);
        }
        catch
        {
            // Client bereits disconnected – ignorieren
        }
    }
}

public record ToolResult(string Id, string Result, bool IsError);

public class TokenUsage
{
    public long InputTokens { get; set; }
    public long OutputTokens { get; set; }
    public long CacheCreationTokens { get; set; }
    public long CacheReadTokens { get; set; }
}