Add live steering injection into running chains

Let the user inject a message into a running chain instead of only being
able to abort it. While a chain streams, typing + Enter sends the text to
a new /api/chat/steer endpoint; it is drained at the next iteration
boundary, wrapped with a hint so the model actively classifies it
("complements the task -> keep working" vs. "changes course -> adjust"),
and stored as a Type="steer" user message so it survives reload.

Shared infrastructure (provider-agnostic):
- SteeringQueue: out-of-band queue keyed by sessionId, mirrored from
  StreamingTracker (Enqueue/Drain/Clear).
- Program.cs registers it as a singleton and passes it to AnthropicService.
- ChatEndpoint: POST /api/chat/steer enqueues only while IsStreaming,
  Clear() in the finally.
- ChatService.BuildHistoryFromDb groups Type="steer" so no two
  consecutive user turns are produced on re-chat.
- Frontend api.js/chat.js/bubbles.js/history.js + chat.css render the
  optimistic steer bubble and the steer_injected SSE event; Enter during
  a stream sends a steer instead of being blocked.

Anthropic API path:
- Drain point B (tool boundary, merged into the tool_result user turn)
  and C (turn end -> inject + continue instead of finalize).

Claude CLI path (kill-and-restart):
- The agentic loop runs inside the claude child, so true mid-turn inject
  is impossible. On a pending steer the child is killed at the next tool
  boundary, progress is rebuilt from the DB as a <work_in_progress>
  transcript, and the CLI is restarted with the steer appended.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
MelekiandClaude Opus 4.8 committed 2026-06-02 06:49:58 +02:00
1 parent a58f279eaa
commit f2fd493e5f
12 files changed
+555 -132

No files matched your search

+54
View File
@@ -0,0 +1,54 @@
using System.Collections.Concurrent;
using System.Collections.Generic;
namespace Gott.Services;
/// <summary>
/// Out-of-band-Kanal für "Steering"-Einwürfe: Nachrichten, die der User schickt
/// WÄHREND eine Chain läuft (vgl. /btw bzw. Queued-Steering in Claude Code).
/// 1:1 gespiegelt vom <see cref="StreamingTracker"/> – keyed by sessionId.
///
/// Der Streaming-Loop (AnthropicService.StreamChat) draint die Queue an den
/// Iterations-Grenzen (nach einer Tool-Runde bzw. bevor die Chain finalisiert)
/// und injiziert die Texte als User-Message in die laufende Konversation.
/// Ein laufender messages.create-Request ist unveränderlich – Injektion geht
/// nur ZWISCHEN den API-Calls, nicht mitten im Stream.
///
/// Im CLI-Pfad (AnthropicService.StreamChatViaCli) läuft der Agentic-Loop im
/// claude-Child; dort draint ein Kill-and-Restart am nächsten Tool-Boundary
/// (Child killen, Verlauf + Einwurf an den Prompt hängen, CLI neu starten).
/// </summary>
public class SteeringQueue
{
private readonly ConcurrentDictionary<string, ConcurrentQueue<string>> _queues = new ConcurrentDictionary<string, ConcurrentQueue<string>>();
/// <summary>Reiht einen Einwurf für eine Session ein (FIFO).</summary>
public void Enqueue(string sessionId, string text)
{
if (string.IsNullOrEmpty(sessionId) || string.IsNullOrWhiteSpace(text)) return;
var q = _queues.GetOrAdd(sessionId, _ => new ConcurrentQueue<string>());
q.Enqueue(text);
}
/// <summary>
/// Entnimmt ALLE pending Einwürfe einer Session (FIFO) und gibt sie zurück.
/// Leere Liste wenn nichts ansteht.
/// </summary>
public List<string> Drain(string sessionId)
{
var result = new List<string>();
if (_queues.TryGetValue(sessionId, out var q))
{
while (q.TryDequeue(out var text))
result.Add(text);
}
return result;
}
/// <summary>Prüft ob für eine Session Einwürfe anstehen.</summary>
public bool HasPending(string sessionId)
=> _queues.TryGetValue(sessionId, out var q) && !q.IsEmpty;
/// <summary>Entfernt die Queue einer Session (beim Chain-Ende aufräumen).</summary>
public void Clear(string sessionId) => _queues.TryRemove(sessionId, out _);
}