feat(relay): hosted-tool conversion fidelity, reasoning normalization, and billing usage integrity (#7137)

* feat(relaykit): preserve hosted tools across conversions

- add protocol-neutral hosted-tool DTOs, conversion metadata, and loss policies
- bridge citations, grounding metadata, and hosted-tool stream lifecycles
- document the public conversion behavior and channel policy controls

* refactor(relaykit): normalize reasoning and thinking intent

- centralize provider-neutral reasoning intent, effort, and budget mappings
- parse model suffixes at the host entry boundary while preserving provider-owned tails
- keep adaptive Claude thinking and explicit zero-token compatibility consistent

* fix(billing): preserve authoritative usage across relay hops

- carry native BillingUsage sidecars through direct and streamed protocol bridges
- merge partial and terminal usage monotonically with safe fallback settlement
- retain cache metadata, penultimate usage, and per-call Gemini tool surcharges

* feat(relay): bridge Responses with Claude and Gemini protocols

- add direct request, response, and stream converters across supported relay formats
- expose Claude count_tokens and Chat-to-Responses compatibility endpoints
- carry conversion diagnostics through the host while retaining the curated public goldens

* fix(relay): wire relaykit conversions into host channels

- connect handlers, adaptors, and channel settings to the standalone conversion layer
- keep model mapping, pricing identity, retries, and provider-specific suffix behavior aligned
- ignore local audit artifacts and retain focused public regression coverage

relay/common/conversion_diagnostics.go

@@ -0,0 +1,71 @@
+package common
+
+import (
+	"context"
+	"fmt"
+
+	"github.com/QuantumNous/new-api/logger"
+	"github.com/QuantumNous/new-api/relaykit/types"
+	"github.com/gin-gonic/gin"
+)
+
+const maxConversionDiagnostics = 32
+
+type conversionDiagnosticKey struct {
+	code     string
+	path     string
+	severity types.ConversionDiagnosticSeverity
+	from     types.RelayFormat
+	to       types.RelayFormat
+}
+
+// RecordConversionDiagnostics retains conversion losses for the consume log
+// and emits one request-correlated warning per distinct diagnostic. The cap
+// prevents malformed streams from growing request state without bound.
+func (info *RelayInfo) RecordConversionDiagnostics(ctx context.Context, diagnostics []types.ConversionDiagnostic) {
+	if info == nil || len(diagnostics) == 0 {
+		return
+	}
+	if ginCtx, ok := ctx.(*gin.Context); ok && ginCtx == nil {
+		ctx = nil
+	}
+	if info.conversionDiagnosticKeys == nil {
+		info.conversionDiagnosticKeys = make(map[conversionDiagnosticKey]struct{})
+	}
+	for _, diagnostic := range diagnostics {
+		key := conversionDiagnosticKey{
+			code:     diagnostic.Code,
+			path:     diagnostic.Path,
+			severity: diagnostic.Severity,
+			from:     diagnostic.From,
+			to:       diagnostic.To,
+		}
+		if _, exists := info.conversionDiagnosticKeys[key]; exists {
+			continue
+		}
+		if len(info.conversionDiagnostics) >= maxConversionDiagnostics {
+			if !info.conversionDiagnosticsTruncated {
+				info.conversionDiagnosticsTruncated = true
+				logger.LogWarn(ctx, fmt.Sprintf("conversion diagnostics truncated after %d distinct entries", maxConversionDiagnostics))
+			}
+			continue
+		}
+		info.conversionDiagnosticKeys[key] = struct{}{}
+		info.conversionDiagnostics = append(info.conversionDiagnostics, diagnostic)
+		logger.LogWarn(ctx, fmt.Sprintf(
+			"conversion diagnostic: code=%q severity=%q from=%q to=%q path=%q message=%q",
+			diagnostic.Code, diagnostic.Severity, diagnostic.From, diagnostic.To, diagnostic.Path, diagnostic.Message,
+		))
+	}
+}
+
+func (info *RelayInfo) ConversionDiagnostics() []types.ConversionDiagnostic {
+	if info == nil || len(info.conversionDiagnostics) == 0 {
+		return nil
+	}
+	return append([]types.ConversionDiagnostic(nil), info.conversionDiagnostics...)
+}
+
+func (info *RelayInfo) ConversionDiagnosticsTruncated() bool {
+	return info != nil && info.conversionDiagnosticsTruncated
+}