Dynamic model fetching, pricing scraping, and context window metadata

.gitignore

@@ -14,3 +14,9 @@
 # OS files
 .DS_Store
 Thumbs.db
+
+# Build artifacts & local-only scripts
+bin/command-code-proxy-updated
+bin/command-code-proxy-new.exe
+run-proxy.bat
+run-proxy.vbs

internal/api/commandcode.go

@@ -33,6 +33,7 @@ type CCChatParams struct {
 	MaxTokens   int         `json:"max_tokens"`
 	Temperature float64     `json:"temperature"`
 	Stream      bool        `json:"stream"`
+	Reasoning   string      `json:"reasoning,omitempty"`
 }
 
 type CCConfig struct {

internal/api/openai.go

@@ -58,8 +58,17 @@ type OpenAIChatRequest struct {
 	PresencePenalty     *float64        `json:"presence_penalty,omitempty"`
 	FrequencyPenalty    *float64        `json:"frequency_penalty,omitempty"`
 	User                string          `json:"user,omitempty"`
+	Reasoning           *ReasoningEffort `json:"reasoning,omitempty"`
 }
 
+type ReasoningEffort string
+
+const (
+	ReasoningEffortLow    ReasoningEffort = "low"
+	ReasoningEffortMedium ReasoningEffort = "medium"
+	ReasoningEffortHigh   ReasoningEffort = "high"
+)
+
 type OpenAIResponsesRequest struct {
 	Model               string   `json:"model"`
 	Input               any      `json:"input"`
@@ -131,10 +140,20 @@ type OpenAIError struct {
 }
 
 type OpenAIModel struct {
-	ID      string `json:"id"`
-	Object  string `json:"object"`
-	Created int64  `json:"created"`
-	OwnedBy string `json:"owned_by"`
+	ID            string `json:"id"`
+	Object        string `json:"object"`
+	Created       int64  `json:"created"`
+	OwnedBy       string `json:"owned_by"`
+	ContextLength int    `json:"context_length,omitempty"`
+	Pricing       *ModelPricing `json:"pricing,omitempty"`
+}
+
+// ModelPricing describes deal/pricing information for a model.
+// Only populated when an active deal exists.
+type ModelPricing struct {
+	Multiplier  string `json:"multiplier,omitempty"`   // e.g. "4x", "2x", "99% off"
+	Description string `json:"description,omitempty"`  // human-readable description
+	Status      string `json:"status,omitempty"`       // "permanent" or expiration date
 }
 
 type OpenAIModelList struct {

internal/proxy/contextmap.go

@@ -0,0 +1,112 @@
+package proxy
+
+// contextLengthMap maps model IDs to their context window size in tokens.
+// Command Code's /v1/models endpoint does not expose context windows, so we
+// maintain a local override. Unknown models fall back to 128K (safe default).
+//
+// Sources:
+//   - Upstream provider documentation (Anthropic, OpenAI, Google, etc.)
+//   - Command Code CLI docs (https://commandcode.ai/docs/reference/cli/models)
+//   - Model release notes
+var contextLengthMap = map[string]int{
+	// MoonshotAI Kimi
+	"moonshotai/Kimi-K2.7-Code":           262144,
+	"moonshotai/Kimi-K2.7-Code-Highspeed": 262144,
+	"moonshotai/Kimi-K2.6":                262144,
+	"moonshotai/Kimi-K2.5":                262144,
+
+	// ZhipuAI GLM
+	"zai-org/GLM-5.2": 1048576, // 1M
+	"zai-org/GLM-5.1": 202752,
+	"zai-org/GLM-5":   202752,
+
+	// MiniMax
+	"MiniMaxAI/MiniMax-M3":        1000000, // 1M
+	"MiniMaxAI/MiniMax-M3-Promo":  1000000, // 1M
+	"MiniMaxAI/MiniMax-M2.7":      204800,
+	"MiniMaxAI/MiniMax-M2.5":      204800,
+
+	// DeepSeek
+	"deepseek/deepseek-v4-pro":   1000000, // 1M
+	"deepseek/deepseek-v4-flash": 1000000, // 1M
+
+	// Qwen
+	"Qwen/Qwen3.6-Max-Preview": 1048576, // 1M
+	"Qwen/Qwen3.6-Plus":        1048576, // 1M
+	"Qwen/Qwen3.7-Max":         1048576, // 1M
+	"Qwen/Qwen3.7-Plus":        1048576, // 1M
+
+	// StepFun
+	"stepfun/Step-3.7-Flash": 262144,
+	"stepfun/Step-3.5-Flash": 262144,
+
+	// Xiaomi MiMo
+	"xiaomi/mimo-v2.5-pro": 1048576, // 1M
+	"xiaomi/mimo-v2.5":     1048576, // 1M
+
+	// NVIDIA Nemotron
+	"nvidia/nemotron-3-ultra-550b-a55b": 131072,
+
+	// Anthropic Claude
+	"claude-sonnet-4-6":              1000000, // 1M
+	"claude-fable-5":                 1000000, // 1M
+	"claude-opus-4-8":                1000000, // 1M
+	"claude-opus-4-7":                1000000, // 1M
+	"claude-opus-4-6":                200000,
+	"claude-haiku-4-5":               200000,
+	"claude-haiku-4-5-20251001":      200000,
+
+	// OpenAI GPT
+	"gpt-5.5":         1050000, // ~1M with reasoning overhead
+	"gpt-5.4":         1050000,
+	"gpt-5.3-codex":   400000,
+	"gpt-5.4-mini":    400000,
+
+	// Google Gemini
+	"google/gemini-3.5-flash":      1048576, // 1M
+	"google/gemini-3.1-flash-lite": 1048576, // 1M
+}
+
+// defaultContextLength is returned for unknown models when we can't determine
+// their context window. Conservative — better to over-estimate and trigger
+// compression early than to under-estimate and blow the window.
+const defaultContextLength = 131072 // 128K
+
+// tasteOneModelID is Command Code's internal model that ships free with all plans.
+// Not exposed via upstream /v1/models API; hardcoded here so it's discoverable
+// through the proxy.
+const tasteOneModelID = "taste-1"
+
+// tasteOneContextLength is the context window for taste-1. Set to a reasonable
+// coding-agent window — tune if Command Code publishes a different number.
+const tasteOneContextLength = 262144 // 256K
+
+// ContextLengthFor returns the context window for a model ID.
+// Falls back to defaultContextLength if the model is not in the map.
+func ContextLengthFor(modelID string) int {
+	if n, ok := contextLengthMap[modelID]; ok {
+		return n
+	}
+	// Try matching by suffix (model names can have provider prefixes we don't track)
+	for knownID, n := range contextLengthMap {
+		if modelMatches(modelID, knownID) {
+			return n
+		}
+	}
+	return defaultContextLength
+}
+
+// modelMatches does fuzzy matching — strips common prefixes and compares.
+// e.g. "minimax/MiniMax-M3" matches "MiniMaxAI/MiniMax-M3"
+func modelMatches(query, known string) bool {
+	// Extract model name after the last "/"
+	getSuffix := func(s string) string {
+		for i := len(s) - 1; i >= 0; i-- {
+			if s[i] == '/' {
+				return s[i+1:]
+			}
+		}
+		return s
+	}
+	return getSuffix(query) == getSuffix(known)
+}

internal/proxy/model.go

@@ -13,6 +13,12 @@ func MapModel(name string) string {
 		return "MiniMaxAI/MiniMax-M2.7"
 	case "minimax-m2.5", "minimax2.5", "minimax":
 		return "MiniMaxAI/MiniMax-M2.5"
+	case "minimax-m3", "minimax3":
+		return "MiniMaxAI/MiniMax-M3"
+	case "minimax-m3-promo", "minimax3-promo":
+		return "MiniMaxAI/MiniMax-M3-Promo"
+	case "glm-5.2", "glm-52":
+		return "zai-org/GLM-5.2"
 	case "glm-5.1":
 		return "zai-org/GLM-5.1"
 	case "glm-5":
@@ -21,14 +27,52 @@ func MapModel(name string) string {
 		return "moonshotai/Kimi-K2.6"
 	case "kimi-k2.5", "kimi2.5":
 		return "moonshotai/Kimi-K2.5"
+	case "kimi-k2.7-code", "kimi2.7-code":
+		return "moonshotai/Kimi-K2.7-Code"
+	case "kimi-k2.7-code-highspeed", "kimi2.7-code-highspeed":
+		return "moonshotai/Kimi-K2.7-Code-Highspeed"
 	case "qwen-3.6-max-preview", "qwen3.6-max":
 		return "Qwen/Qwen3.6-Max-Preview"
 	case "qwen-3.6-plus", "qwen3.6-plus", "qwen3.6":
 		return "Qwen/Qwen3.6-Plus"
+	case "qwen-3.7-max", "qwen3.7-max":
+		return "Qwen/Qwen3.7-Max"
+	case "qwen-3.7-plus", "qwen3.7-plus", "qwen3.7":
+		return "Qwen/Qwen3.7-Plus"
 	case "step-3.5-flash", "step3.5":
 		return "stepfun/Step-3.5-Flash"
+	case "step-3.7-flash", "step3.7":
+		return "stepfun/Step-3.7-Flash"
+	case "mimo-v2.5-pro", "mimo2.5-pro":
+		return "xiaomi/mimo-v2.5-pro"
+	case "mimo-v2.5", "mimo2.5", "mimo":
+		return "xiaomi/mimo-v2.5"
+	case "nemotron-3-ultra", "nemotron":
+		return "nvidia/nemotron-3-ultra-550b-a55b"
+	case "claude-sonnet-4-6", "sonnet-4-6", "sonnet":
+		return "claude-sonnet-4-6"
+	case "claude-fable-5", "fable-5", "fable":
+		return "claude-fable-5"
+	case "claude-opus-4-8", "opus-4-8", "opus":
+		return "claude-opus-4-8"
+	case "claude-opus-4-7", "opus-4-7":
+		return "claude-opus-4-7"
+	case "claude-opus-4-6", "opus-4-6":
+		return "claude-opus-4-6"
+	case "claude-haiku-4-5", "haiku-4-5", "haiku":
+		return "claude-haiku-4-5"
+	case "gpt-5.5":
+		return "gpt-5.5"
+	case "gpt-5.4":
+		return "gpt-5.4"
+	case "gpt-5.3-codex", "codex":
+		return "gpt-5.3-codex"
+	case "gpt-5.4-mini", "gpt-mini":
+		return "gpt-5.4-mini"
 	case "gemini-3.1-flash-lite", "gemini-flash-lite":
 		return "google/gemini-3.1-flash-lite"
+	case "gemini-3.5-flash", "gemini-flash":
+		return "google/gemini-3.5-flash"
 	default:
 		return name // pass through as-is
 	}