feat(server): API relay gateway backend M0-M4

Gin + GORM + pure-Go SQLite. Users/auth (JWT), API key management with
quotas, proxy gateway with weighted channel failover and health checks,
usage/billing ledger, cross-protocol conversion (Anthropic Messages /
OpenAI Chat Completions / OpenAI Responses), and channel/model admin API.
Channels declare native API formats and auto-convert the rest.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

server/internal/proxy/usage.go

@@ -0,0 +1,48 @@
+package proxy
+
+import (
+	"encoding/json"
+
+	"openteam/server/internal/proxy/claude"
+	"openteam/server/internal/proxy/openai"
+	"openteam/server/internal/proxy/responses"
+)
+
+// parseUsageForProtocol extracts token usage from a non-stream response body.
+func parseUsageForProtocol(proto Protocol, body []byte) *tokenUsage {
+	switch proto {
+	case ProtocolAnthropic:
+		var mr claude.MessageResponse
+		if err := json.Unmarshal(body, &mr); err != nil || mr.Usage == nil {
+			return nil
+		}
+		return &tokenUsage{
+			input:         mr.Usage.InputTokens,
+			output:        mr.Usage.OutputTokens,
+			cacheRead:     mr.Usage.CacheReadInputTokens,
+			cacheCreation: mr.Usage.CacheCreationInputTokens,
+		}
+	case ProtocolOpenAIResponses:
+		var r responses.Response
+		if err := json.Unmarshal(body, &r); err != nil || r.Usage == nil {
+			return nil
+		}
+		return &tokenUsage{
+			input:     r.Usage.InputTokens,
+			output:    r.Usage.OutputTokens,
+			cacheRead: r.Usage.InputTokensDetails.CachedTokens,
+		}
+	default:
+		var cc openai.ChatCompletion
+		if err := json.Unmarshal(body, &cc); err != nil || cc.Usage == nil {
+			return nil
+		}
+		return &tokenUsage{input: cc.Usage.PromptTokens, output: cc.Usage.CompletionTokens}
+	}
+}
+
+// estimateTokensFromText is a coarse fallback used when the upstream omits usage.
+func estimateTokensFromText(s string) int64 {
+	// ~4 chars per token, per OpenAI's common heuristic.
+	return int64(len(s)/4 + 1)
+}