98 lines
3.1 KiB
C#
98 lines
3.1 KiB
C#
namespace AnchorCli.OpenRouter;
|
|
|
|
/// <summary>
|
|
/// Tracks token usage and calculates costs for the session.
|
|
/// </summary>
|
|
internal sealed class TokenTracker
|
|
{
|
|
private readonly ChatSession _session;
|
|
|
|
public TokenTracker(ChatSession session)
|
|
{
|
|
_session = session;
|
|
}
|
|
|
|
public string Provider { get; set; } = "Unknown";
|
|
public long SessionInputTokens => _session.SessionInputTokens;
|
|
public long SessionOutputTokens => _session.SessionOutputTokens;
|
|
public int RequestCount => _session.RequestCount;
|
|
|
|
/// <summary>Maximum context window for the model (tokens). 0 = unknown.</summary>
|
|
public int ContextLength { get; set; }
|
|
|
|
/// <summary>Input tokens from the most recent API response — approximates current context size.</summary>
|
|
public int LastInputTokens { get; private set; }
|
|
|
|
/// <summary>USD per input token.</summary>
|
|
public decimal InputPrice { get; set; }
|
|
|
|
/// <summary>USD per output token.</summary>
|
|
public decimal OutputPrice { get; set; }
|
|
|
|
/// <summary>Fixed USD per API request.</summary>
|
|
public decimal RequestPrice { get; set; }
|
|
/// <summary>
|
|
/// Record usage from one response (may span multiple LLM rounds).
|
|
/// </summary>
|
|
public void AddUsage(int inputTokens, int outputTokens)
|
|
{
|
|
_session.SessionInputTokens += inputTokens;
|
|
_session.SessionOutputTokens += outputTokens;
|
|
LastInputTokens = inputTokens;
|
|
_session.RequestCount++;
|
|
}
|
|
public void Reset()
|
|
{
|
|
_session.SessionInputTokens = 0;
|
|
_session.SessionOutputTokens = 0;
|
|
_session.RequestCount = 0;
|
|
LastInputTokens = 0;
|
|
}
|
|
|
|
|
|
private const int MaxContextReserve = 150_000;
|
|
|
|
/// <summary>
|
|
/// Returns true if the context is getting too large and should be compacted.
|
|
/// Triggers at min(75% of model context, 150K tokens).
|
|
/// </summary>
|
|
public bool ShouldCompact()
|
|
{
|
|
if (LastInputTokens <= 0) return false;
|
|
|
|
int threshold = ContextLength > 0
|
|
? Math.Min((int)(ContextLength * 0.75), MaxContextReserve)
|
|
: MaxContextReserve;
|
|
|
|
return LastInputTokens >= threshold;
|
|
}
|
|
|
|
/// <summary>Context usage as a percentage (0-100). Returns -1 if context length is unknown.</summary>
|
|
public double ContextUsagePercent =>
|
|
ContextLength > 0 && LastInputTokens > 0
|
|
? (double)LastInputTokens / ContextLength * 100.0
|
|
: -1;
|
|
|
|
/// <summary>
|
|
/// Calculate cost for a single response.
|
|
/// </summary>
|
|
public decimal CalculateCost(int inputTokens, int outputTokens) =>
|
|
inputTokens * InputPrice +
|
|
outputTokens * OutputPrice +
|
|
RequestPrice;
|
|
|
|
/// <summary>
|
|
/// Total session cost.
|
|
/// </summary>
|
|
public decimal SessionCost =>
|
|
SessionInputTokens * InputPrice +
|
|
SessionOutputTokens * OutputPrice +
|
|
RequestCount * RequestPrice;
|
|
|
|
public static string FormatTokens(long count) =>
|
|
count >= 1_000 ? $"{count / 1_000.0:F1}k" : count.ToString("N0");
|
|
|
|
public static string FormatCost(decimal cost) =>
|
|
cost < 0.01m ? $"${cost:F4}" : $"${cost:F2}";
|
|
}
|