provider.go

v0.7.0

package provider

// StaticModel represents a model available from a provider (no API fetch needed).
type StaticModel struct {
	ID     string
	Name   string
	Family string
}

// Plan describes one billing/access plan for a provider with multiple base URLs.
type Plan struct {
	Key         string // e.g. "api", "coding"
	Name        string // display name, e.g. "API (pay per token)"
	Description string
	BaseURL     string
	Compat      string // optional override; when set, switches the provider's Compat mode
	AuthEnvVar  string // optional override; when set, switches the auth env var
}

var zaiModels = []StaticModel{
	{ID: "glm-5", Name: "GLM-5", Family: "GLM-5"},
	{ID: "glm-5-turbo", Name: "GLM-5 Turbo", Family: "GLM-5"},
	{ID: "glm-4.7", Name: "GLM-4.7", Family: "GLM-4.7"},
	{ID: "glm-4.7-flash", Name: "GLM-4.7 Flash", Family: "GLM-4.7"},
	{ID: "glm-4.7-flashx", Name: "GLM-4.7 FlashX", Family: "GLM-4.7"},
	{ID: "glm-4.6", Name: "GLM-4.6", Family: "GLM-4.6"},
	{ID: "glm-4.5", Name: "GLM-4.5", Family: "GLM-4.5"},
	{ID: "glm-4.5-air", Name: "GLM-4.5 Air", Family: "GLM-4.5"},
	{ID: "glm-4.5-x", Name: "GLM-4.5 X", Family: "GLM-4.5"},
	{ID: "glm-4.5-airx", Name: "GLM-4.5 AirX", Family: "GLM-4.5"},
	{ID: "glm-4.5-flash", Name: "GLM-4.5 Flash", Family: "GLM-4.5"},
	{ID: "glm-4-32b-0414-128k", Name: "GLM-4 32B", Family: "GLM-4"},
}

var zaiCodingModels = []StaticModel{
	{ID: "glm-5.1", Name: "GLM-5.1", Family: "GLM-5"},
	{ID: "glm-5", Name: "GLM-5", Family: "GLM-5"},
	{ID: "glm-5-turbo", Name: "GLM-5 Turbo", Family: "GLM-5"},
	{ID: "glm-4.7", Name: "GLM-4.7", Family: "GLM-4.7"},
	{ID: "glm-4.7-flash", Name: "GLM-4.7 Flash", Family: "GLM-4.7"},
	{ID: "glm-4.7-flashx", Name: "GLM-4.7 FlashX", Family: "GLM-4.7"},
	{ID: "glm-4.5-air", Name: "GLM-4.5 Air", Family: "GLM-4.5"},
}

// Kimi Code exposes an Anthropic-compatible gateway at /coding/ and does not
// serve a listable models endpoint there, so we ship a static catalog.
var kimiCodeModels = []StaticModel{
	{ID: "kimi-k2.6", Name: "Kimi K2.6", Family: "Kimi K2"},
	{ID: "kimi-k2.5", Name: "Kimi K2.5", Family: "Kimi K2"},
	{ID: "kimi-k2-turbo-preview", Name: "Kimi K2 Turbo (preview)", Family: "Kimi K2"},
}

// StaticModels maps provider keys to their available models.
var StaticModels = map[string][]StaticModel{
	"zai":        zaiModels,
	"zai-coding": zaiCodingModels,
	"kimi":       kimiCodeModels,
}

// Provider defines an alternative Claude Code API provider.
type Provider struct {
	Name        string
	Description string
	BaseURL     string
	Model       string
	SmallModel  string
	SonnetModel string
	OpusModel   string
	HaikuModel  string
	TimeoutMS   string
	// AuthEnvVar is the env var name used for the API key (some use AUTH_TOKEN, some API_KEY).
	AuthEnvVar string
	// Compat is "anthropic" (default) or "openai". When "openai", a local
	// translation proxy is started to convert Anthropic ↔ OpenAI API format.
	Compat string
	// ContextWindow is the model's context window size in tokens.
	// Claude Code assumes 200k by default; when this is smaller,
	// CLAUDE_AUTOCOMPACT_PCT_OVERRIDE is set so compaction triggers before
	// the real limit is hit.
	ContextWindow int
	// NoAuth indicates the provider does not require an API key (e.g. local Ollama).
	NoAuth bool
	// Plans lists alternative billing/access plans with different base URLs.
	// When non-empty, the TUI prompts the user to choose a plan.
	// BaseURL holds the default plan's URL.
	Plans []Plan
}

var Registry = map[string]Provider{
	"deepseek": {
		Name:          "DeepSeek",
		Description:   "DeepSeek Reasoner / V3 via Anthropic-compatible API",
		BaseURL:       "https://api.deepseek.com/anthropic",
		Model:         "deepseek-reasoner",
		SmallModel:    "deepseek-chat",
		SonnetModel:   "deepseek-reasoner",
		OpusModel:     "deepseek-reasoner",
		HaikuModel:    "deepseek-chat",
		TimeoutMS:     "600000",
		AuthEnvVar:    "ANTHROPIC_AUTH_TOKEN",
		ContextWindow: 128_000,
	},
	"zai": {
		Name:          "z.AI API",
		Description:   "GLM-5 / GLM-4.7 via z.AI (pay-per-token)",
		BaseURL:       "https://api.z.ai/api/anthropic",
		Model:         "glm-5",
		SmallModel:    "glm-4.5-air",
		SonnetModel:   "glm-5",
		OpusModel:     "glm-5",
		HaikuModel:    "glm-4.5-air",
		TimeoutMS:     "3000000",
		AuthEnvVar:    "ANTHROPIC_AUTH_TOKEN",
		ContextWindow: 205_000,
	},
	"zai-coding": {
		Name:          "z.AI Coding",
		Description:   "GLM-5.1 / GLM-5 via z.AI (subscription)",
		BaseURL:       "https://api.z.ai/api/coding/anthropic",
		Model:         "glm-5.1",
		SmallModel:    "glm-4.5-air",
		SonnetModel:   "glm-5.1",
		OpusModel:     "glm-5.1",
		HaikuModel:    "glm-4.5-air",
		TimeoutMS:     "3000000",
		AuthEnvVar:    "ANTHROPIC_AUTH_TOKEN",
		ContextWindow: 205_000,
	},
	"minimax": {
		Name:          "MiniMax",
		Description:   "MiniMax-M2.5 via MiniMax Anthropic gateway",
		BaseURL:       "https://api.minimax.io/anthropic",
		Model:         "MiniMax-M2.5",
		SmallModel:    "MiniMax-M2.5",
		SonnetModel:   "MiniMax-M2.5",
		OpusModel:     "MiniMax-M2.5",
		HaikuModel:    "MiniMax-M2.5",
		TimeoutMS:     "3000000",
		AuthEnvVar:    "ANTHROPIC_AUTH_TOKEN",
		ContextWindow: 205_000,
	},
	"kimi": {
		Name:          "Kimi Code",
		Description:   "Kimi K2.5 / K2 via Kimi Code (International)",
		BaseURL:       "https://api.kimi.com/coding/",
		Model:         "kimi-k2.5",
		SmallModel:    "kimi-k2-turbo-preview",
		SonnetModel:   "kimi-k2.5",
		OpusModel:     "kimi-k2.5",
		HaikuModel:    "kimi-k2-turbo-preview",
		TimeoutMS:     "600000",
		AuthEnvVar:    "ANTHROPIC_AUTH_TOKEN",
		ContextWindow: 256_000,
	},
	"kimi-api-cn": {
		Name:          "Kimi API (China)",
		Description:   "Kimi via Moonshot Open Platform (api.moonshot.cn)",
		BaseURL:       "https://api.moonshot.cn/v1/",
		Model:         "kimi-k2.5",
		SmallModel:    "kimi-k2-turbo-preview",
		SonnetModel:   "kimi-k2.5",
		OpusModel:     "kimi-k2.5",
		HaikuModel:    "kimi-k2-turbo-preview",
		TimeoutMS:     "600000",
		AuthEnvVar:    "ANTHROPIC_AUTH_TOKEN",
		ContextWindow: 256_000,
		Compat:        "openai",
	},
	"kimi-api-intl": {
		Name:          "Kimi API (Intl)",
		Description:   "Kimi via Moonshot Open Platform (api.moonshot.ai)",
		BaseURL:       "https://api.moonshot.ai/v1/",
		Model:         "kimi-k2.5",
		SmallModel:    "kimi-k2-turbo-preview",
		SonnetModel:   "kimi-k2.5",
		OpusModel:     "kimi-k2.5",
		HaikuModel:    "kimi-k2-turbo-preview",
		TimeoutMS:     "600000",
		AuthEnvVar:    "ANTHROPIC_AUTH_TOKEN",
		ContextWindow: 256_000,
		Compat:        "openai",
	},
	"mistral": {
		Name:          "Mistral",
		Description:   "Devstral / Mistral Large via OpenAI-compatible API",
		BaseURL:       "https://api.mistral.ai/v1",
		Model:         "devstral-latest",
		SmallModel:    "devstral-small-latest",
		SonnetModel:   "devstral-latest",
		OpusModel:     "devstral-latest",
		HaikuModel:    "devstral-small-latest",
		TimeoutMS:     "600000",
		AuthEnvVar:    "ANTHROPIC_AUTH_TOKEN",
		Compat:        "openai",
		ContextWindow: 256_000,
	},
	"copilot": {
		Name:          "GitHub Copilot",
		Description:   "Claude via GitHub Copilot",
		BaseURL:       "https://api.githubcopilot.com",
		Model:         "claude-sonnet-4.6",
		SmallModel:    "claude-haiku-4.5",
		SonnetModel:   "claude-sonnet-4.6",
		OpusModel:     "claude-opus-4.6",
		HaikuModel:    "claude-haiku-4.5",
		TimeoutMS:     "600000",
		AuthEnvVar:    "ANTHROPIC_AUTH_TOKEN",
		Compat:        "openai",
		ContextWindow: 200_000,
	},
	"nvidia": {
		Name:          "NVIDIA NIM",
		Description:   "Access NVIDIA NIM models via OpenAI-compatible API",
		BaseURL:       "https://integrate.api.nvidia.com/v1",
		Model:         "nvidia/llama-3.1-nemotron-ultra-253b-v1",
		SmallModel:    "nvidia/llama-3.3-nemotron-super-49b-v1.5",
		SonnetModel:   "nvidia/llama-3.1-nemotron-ultra-253b-v1",
		OpusModel:     "nvidia/llama-3.1-nemotron-ultra-253b-v1",
		HaikuModel:    "nvidia/llama-3.3-nemotron-super-49b-v1.5",
		TimeoutMS:     "600000",
		AuthEnvVar:    "ANTHROPIC_AUTH_TOKEN",
		Compat:        "openai",
		ContextWindow: 128_000,
	},
	"openrouter": {
		Name:          "OpenRouter",
		Description:   "Access multiple models via OpenRouter",
		BaseURL:       "https://openrouter.ai/api/v1",
		Model:         "anthropic/claude-sonnet-4.6",
		SmallModel:    "anthropic/claude-haiku-4.5",
		SonnetModel:   "anthropic/claude-sonnet-4.6",
		OpusModel:     "anthropic/claude-opus-4.6",
		HaikuModel:    "anthropic/claude-haiku-4.5",
		TimeoutMS:     "600000",
		AuthEnvVar:    "ANTHROPIC_AUTH_TOKEN",
		Compat:        "openai",
		ContextWindow: 200_000,
	},
	"vkproxy": {
		Name:          "VK LLM Proxy",
		Description:   "Import models from claude-code-config.zip",
		BaseURL:       "https://llm-proxy.vkteam.ru",
		Model:         "deepseek-reasoner",
		SmallModel:    "deepseek-reasoner",
		SonnetModel:   "deepseek-reasoner",
		OpusModel:     "deepseek-reasoner",
		HaikuModel:    "deepseek-reasoner",
		TimeoutMS:     "600000",
		AuthEnvVar:    "ANTHROPIC_AUTH_TOKEN",
		ContextWindow: 128_000,
	},
	"ollama": {
		Name:          "Ollama (local)",
		Description:   "Local Ollama with native Anthropic-compatible API",
		BaseURL:       "http://localhost:11434",
		Model:         "qwen3-coder",
		SmallModel:    "qwen3-coder",
		SonnetModel:   "qwen3-coder",
		OpusModel:     "qwen3-coder",
		HaikuModel:    "qwen3-coder",
		TimeoutMS:     "600000",
		AuthEnvVar:    "ANTHROPIC_AUTH_TOKEN",
		ContextWindow: 32_000,
		NoAuth:        true,
	},
	"ollama-cloud": {
		Name:          "Ollama Cloud",
		Description:   "Cloud models via ollama.com API",
		BaseURL:       "https://ollama.com",
		Model:         "qwen3.5:397b",
		SmallModel:    "devstral-small-2:24b",
		SonnetModel:   "devstral-2:123b",
		OpusModel:     "qwen3.5:397b",
		HaikuModel:    "devstral-small-2:24b",
		TimeoutMS:     "600000",
		AuthEnvVar:    "ANTHROPIC_AUTH_TOKEN",
		ContextWindow: 32_000,
	},
	"lmstudio": {
		Name:          "LM Studio",
		Description:   "Local LM Studio server with OpenAI-compatible API",
		BaseURL:       "http://localhost:1234/v1",
		Model:         "local-model",
		SmallModel:    "local-model",
		SonnetModel:   "local-model",
		OpusModel:     "local-model",
		HaikuModel:    "local-model",
		TimeoutMS:     "600000",
		AuthEnvVar:    "ANTHROPIC_AUTH_TOKEN",
		ContextWindow: 128_000,
		Compat:        "openai",
		NoAuth:        true,
	},
	"claude": {
		Name:        "Claude (Anthropic)",
		Description: "Native Claude API via Anthropic (passthrough proxy for accounting)",
		BaseURL:     "https://api.anthropic.com",
		Model:       "claude-sonnet-4-6-20250514",
		SmallModel:  "claude-haiku-4-5-20251001",
		SonnetModel: "claude-sonnet-4-6-20250514",
		OpusModel:   "claude-opus-4-6-20250610",
		HaikuModel:  "claude-haiku-4-5-20251001",
		TimeoutMS:   "600000",
		NoAuth:      true,
	},
}

// Order defines the display order of providers.
var Order = []string{"claude", "deepseek", "mistral", "zai", "zai-coding", "minimax", "kimi", "kimi-api-cn", "kimi-api-intl", "copilot", "nvidia", "openrouter", "ollama", "ollama-cloud", "lmstudio", "vkproxy"}


Source Files

provider.go

1	package provider
2
3	// StaticModel represents a model available from a provider (no API fetch needed).
4	type StaticModel struct {
5	ID string
6	Name string
7	Family string
8	}
9
10	// Plan describes one billing/access plan for a provider with multiple base URLs.
11	type Plan struct {
12	Key string // e.g. "api", "coding"
13	Name string // display name, e.g. "API (pay per token)"
14	Description string
15	BaseURL string
16	Compat string // optional override; when set, switches the provider's Compat mode
17	AuthEnvVar string // optional override; when set, switches the auth env var
18	}
19
20	var zaiModels = []StaticModel{
21	{ID: "glm-5", Name: "GLM-5", Family: "GLM-5"},
22	{ID: "glm-5-turbo", Name: "GLM-5 Turbo", Family: "GLM-5"},
23	{ID: "glm-4.7", Name: "GLM-4.7", Family: "GLM-4.7"},
24	{ID: "glm-4.7-flash", Name: "GLM-4.7 Flash", Family: "GLM-4.7"},
25	{ID: "glm-4.7-flashx", Name: "GLM-4.7 FlashX", Family: "GLM-4.7"},
26	{ID: "glm-4.6", Name: "GLM-4.6", Family: "GLM-4.6"},
27	{ID: "glm-4.5", Name: "GLM-4.5", Family: "GLM-4.5"},
28	{ID: "glm-4.5-air", Name: "GLM-4.5 Air", Family: "GLM-4.5"},
29	{ID: "glm-4.5-x", Name: "GLM-4.5 X", Family: "GLM-4.5"},
30	{ID: "glm-4.5-airx", Name: "GLM-4.5 AirX", Family: "GLM-4.5"},
31	{ID: "glm-4.5-flash", Name: "GLM-4.5 Flash", Family: "GLM-4.5"},
32	{ID: "glm-4-32b-0414-128k", Name: "GLM-4 32B", Family: "GLM-4"},
33	}
34
35	var zaiCodingModels = []StaticModel{
36	{ID: "glm-5.1", Name: "GLM-5.1", Family: "GLM-5"},
37	{ID: "glm-5", Name: "GLM-5", Family: "GLM-5"},
38	{ID: "glm-5-turbo", Name: "GLM-5 Turbo", Family: "GLM-5"},
39	{ID: "glm-4.7", Name: "GLM-4.7", Family: "GLM-4.7"},
40	{ID: "glm-4.7-flash", Name: "GLM-4.7 Flash", Family: "GLM-4.7"},
41	{ID: "glm-4.7-flashx", Name: "GLM-4.7 FlashX", Family: "GLM-4.7"},
42	{ID: "glm-4.5-air", Name: "GLM-4.5 Air", Family: "GLM-4.5"},
43	}
44
45	// Kimi Code exposes an Anthropic-compatible gateway at /coding/ and does not
46	// serve a listable models endpoint there, so we ship a static catalog.
47	var kimiCodeModels = []StaticModel{
48	{ID: "kimi-k2.6", Name: "Kimi K2.6", Family: "Kimi K2"},
49	{ID: "kimi-k2.5", Name: "Kimi K2.5", Family: "Kimi K2"},
50	{ID: "kimi-k2-turbo-preview", Name: "Kimi K2 Turbo (preview)", Family: "Kimi K2"},
51	}
52
53	// StaticModels maps provider keys to their available models.
54	var StaticModels = map[string][]StaticModel{
55	"zai": zaiModels,
56	"zai-coding": zaiCodingModels,
57	"kimi": kimiCodeModels,
58	}
59
60	// Provider defines an alternative Claude Code API provider.
61	type Provider struct {
62	Name string
63	Description string
64	BaseURL string
65	Model string
66	SmallModel string
67	SonnetModel string
68	OpusModel string
69	HaikuModel string
70	TimeoutMS string
71	// AuthEnvVar is the env var name used for the API key (some use AUTH_TOKEN, some API_KEY).
72	AuthEnvVar string
73	// Compat is "anthropic" (default) or "openai". When "openai", a local
74	// translation proxy is started to convert Anthropic ↔ OpenAI API format.
75	Compat string
76	// ContextWindow is the model's context window size in tokens.
77	// Claude Code assumes 200k by default; when this is smaller,
78	// CLAUDE_AUTOCOMPACT_PCT_OVERRIDE is set so compaction triggers before
79	// the real limit is hit.
80	ContextWindow int
81	// NoAuth indicates the provider does not require an API key (e.g. local Ollama).
82	NoAuth bool
83	// Plans lists alternative billing/access plans with different base URLs.
84	// When non-empty, the TUI prompts the user to choose a plan.
85	// BaseURL holds the default plan's URL.
86	Plans []Plan
87	}
88
89	var Registry = map[string]Provider{
90	"deepseek": {
91	Name: "DeepSeek",
92	Description: "DeepSeek Reasoner / V3 via Anthropic-compatible API",
93	BaseURL: "https://api.deepseek.com/anthropic",
94	Model: "deepseek-reasoner",
95	SmallModel: "deepseek-chat",
96	SonnetModel: "deepseek-reasoner",
97	OpusModel: "deepseek-reasoner",
98	HaikuModel: "deepseek-chat",
99	TimeoutMS: "600000",
100	AuthEnvVar: "ANTHROPIC_AUTH_TOKEN",
101	ContextWindow: 128_000,
102	},
103	"zai": {
104	Name: "z.AI API",
105	Description: "GLM-5 / GLM-4.7 via z.AI (pay-per-token)",
106	BaseURL: "https://api.z.ai/api/anthropic",
107	Model: "glm-5",
108	SmallModel: "glm-4.5-air",
109	SonnetModel: "glm-5",
110	OpusModel: "glm-5",
111	HaikuModel: "glm-4.5-air",
112	TimeoutMS: "3000000",
113	AuthEnvVar: "ANTHROPIC_AUTH_TOKEN",
114	ContextWindow: 205_000,
115	},
116	"zai-coding": {
117	Name: "z.AI Coding",
118	Description: "GLM-5.1 / GLM-5 via z.AI (subscription)",
119	BaseURL: "https://api.z.ai/api/coding/anthropic",
120	Model: "glm-5.1",
121	SmallModel: "glm-4.5-air",
122	SonnetModel: "glm-5.1",
123	OpusModel: "glm-5.1",
124	HaikuModel: "glm-4.5-air",
125	TimeoutMS: "3000000",
126	AuthEnvVar: "ANTHROPIC_AUTH_TOKEN",
127	ContextWindow: 205_000,
128	},
129	"minimax": {
130	Name: "MiniMax",
131	Description: "MiniMax-M2.5 via MiniMax Anthropic gateway",
132	BaseURL: "https://api.minimax.io/anthropic",
133	Model: "MiniMax-M2.5",
134	SmallModel: "MiniMax-M2.5",
135	SonnetModel: "MiniMax-M2.5",
136	OpusModel: "MiniMax-M2.5",
137	HaikuModel: "MiniMax-M2.5",
138	TimeoutMS: "3000000",
139	AuthEnvVar: "ANTHROPIC_AUTH_TOKEN",
140	ContextWindow: 205_000,
141	},
142	"kimi": {
143	Name: "Kimi Code",
144	Description: "Kimi K2.5 / K2 via Kimi Code (International)",
145	BaseURL: "https://api.kimi.com/coding/",
146	Model: "kimi-k2.5",
147	SmallModel: "kimi-k2-turbo-preview",
148	SonnetModel: "kimi-k2.5",
149	OpusModel: "kimi-k2.5",
150	HaikuModel: "kimi-k2-turbo-preview",
151	TimeoutMS: "600000",
152	AuthEnvVar: "ANTHROPIC_AUTH_TOKEN",
153	ContextWindow: 256_000,
154	},
155	"kimi-api-cn": {
156	Name: "Kimi API (China)",
157	Description: "Kimi via Moonshot Open Platform (api.moonshot.cn)",
158	BaseURL: "https://api.moonshot.cn/v1/",
159	Model: "kimi-k2.5",
160	SmallModel: "kimi-k2-turbo-preview",
161	SonnetModel: "kimi-k2.5",
162	OpusModel: "kimi-k2.5",
163	HaikuModel: "kimi-k2-turbo-preview",
164	TimeoutMS: "600000",
165	AuthEnvVar: "ANTHROPIC_AUTH_TOKEN",
166	ContextWindow: 256_000,
167	Compat: "openai",
168	},
169	"kimi-api-intl": {
170	Name: "Kimi API (Intl)",
171	Description: "Kimi via Moonshot Open Platform (api.moonshot.ai)",
172	BaseURL: "https://api.moonshot.ai/v1/",
173	Model: "kimi-k2.5",
174	SmallModel: "kimi-k2-turbo-preview",
175	SonnetModel: "kimi-k2.5",
176	OpusModel: "kimi-k2.5",
177	HaikuModel: "kimi-k2-turbo-preview",
178	TimeoutMS: "600000",
179	AuthEnvVar: "ANTHROPIC_AUTH_TOKEN",
180	ContextWindow: 256_000,
181	Compat: "openai",
182	},
183	"mistral": {
184	Name: "Mistral",
185	Description: "Devstral / Mistral Large via OpenAI-compatible API",
186	BaseURL: "https://api.mistral.ai/v1",
187	Model: "devstral-latest",
188	SmallModel: "devstral-small-latest",
189	SonnetModel: "devstral-latest",
190	OpusModel: "devstral-latest",
191	HaikuModel: "devstral-small-latest",
192	TimeoutMS: "600000",
193	AuthEnvVar: "ANTHROPIC_AUTH_TOKEN",
194	Compat: "openai",
195	ContextWindow: 256_000,
196	},
197	"copilot": {
198	Name: "GitHub Copilot",
199	Description: "Claude via GitHub Copilot",
200	BaseURL: "https://api.githubcopilot.com",
201	Model: "claude-sonnet-4.6",
202	SmallModel: "claude-haiku-4.5",
203	SonnetModel: "claude-sonnet-4.6",
204	OpusModel: "claude-opus-4.6",
205	HaikuModel: "claude-haiku-4.5",
206	TimeoutMS: "600000",
207	AuthEnvVar: "ANTHROPIC_AUTH_TOKEN",
208	Compat: "openai",
209	ContextWindow: 200_000,
210	},
211	"nvidia": {
212	Name: "NVIDIA NIM",
213	Description: "Access NVIDIA NIM models via OpenAI-compatible API",
214	BaseURL: "https://integrate.api.nvidia.com/v1",
215	Model: "nvidia/llama-3.1-nemotron-ultra-253b-v1",
216	SmallModel: "nvidia/llama-3.3-nemotron-super-49b-v1.5",
217	SonnetModel: "nvidia/llama-3.1-nemotron-ultra-253b-v1",
218	OpusModel: "nvidia/llama-3.1-nemotron-ultra-253b-v1",
219	HaikuModel: "nvidia/llama-3.3-nemotron-super-49b-v1.5",
220	TimeoutMS: "600000",
221	AuthEnvVar: "ANTHROPIC_AUTH_TOKEN",
222	Compat: "openai",
223	ContextWindow: 128_000,
224	},
225	"openrouter": {
226	Name: "OpenRouter",
227	Description: "Access multiple models via OpenRouter",
228	BaseURL: "https://openrouter.ai/api/v1",
229	Model: "anthropic/claude-sonnet-4.6",
230	SmallModel: "anthropic/claude-haiku-4.5",
231	SonnetModel: "anthropic/claude-sonnet-4.6",
232	OpusModel: "anthropic/claude-opus-4.6",
233	HaikuModel: "anthropic/claude-haiku-4.5",
234	TimeoutMS: "600000",
235	AuthEnvVar: "ANTHROPIC_AUTH_TOKEN",
236	Compat: "openai",
237	ContextWindow: 200_000,
238	},
239	"vkproxy": {
240	Name: "VK LLM Proxy",
241	Description: "Import models from claude-code-config.zip",
242	BaseURL: "https://llm-proxy.vkteam.ru",
243	Model: "deepseek-reasoner",
244	SmallModel: "deepseek-reasoner",
245	SonnetModel: "deepseek-reasoner",
246	OpusModel: "deepseek-reasoner",
247	HaikuModel: "deepseek-reasoner",
248	TimeoutMS: "600000",
249	AuthEnvVar: "ANTHROPIC_AUTH_TOKEN",
250	ContextWindow: 128_000,
251	},
252	"ollama": {
253	Name: "Ollama (local)",
254	Description: "Local Ollama with native Anthropic-compatible API",
255	BaseURL: "http://localhost:11434",
256	Model: "qwen3-coder",
257	SmallModel: "qwen3-coder",
258	SonnetModel: "qwen3-coder",
259	OpusModel: "qwen3-coder",
260	HaikuModel: "qwen3-coder",
261	TimeoutMS: "600000",
262	AuthEnvVar: "ANTHROPIC_AUTH_TOKEN",
263	ContextWindow: 32_000,
264	NoAuth: true,
265	},
266	"ollama-cloud": {
267	Name: "Ollama Cloud",
268	Description: "Cloud models via ollama.com API",
269	BaseURL: "https://ollama.com",
270	Model: "qwen3.5:397b",
271	SmallModel: "devstral-small-2:24b",
272	SonnetModel: "devstral-2:123b",
273	OpusModel: "qwen3.5:397b",
274	HaikuModel: "devstral-small-2:24b",
275	TimeoutMS: "600000",
276	AuthEnvVar: "ANTHROPIC_AUTH_TOKEN",
277	ContextWindow: 32_000,
278	},
279	"lmstudio": {
280	Name: "LM Studio",
281	Description: "Local LM Studio server with OpenAI-compatible API",
282	BaseURL: "http://localhost:1234/v1",
283	Model: "local-model",
284	SmallModel: "local-model",
285	SonnetModel: "local-model",
286	OpusModel: "local-model",
287	HaikuModel: "local-model",
288	TimeoutMS: "600000",
289	AuthEnvVar: "ANTHROPIC_AUTH_TOKEN",
290	ContextWindow: 128_000,
291	Compat: "openai",
292	NoAuth: true,
293	},
294	"claude": {
295	Name: "Claude (Anthropic)",
296	Description: "Native Claude API via Anthropic (passthrough proxy for accounting)",
297	BaseURL: "https://api.anthropic.com",
298	Model: "claude-sonnet-4-6-20250514",
299	SmallModel: "claude-haiku-4-5-20251001",
300	SonnetModel: "claude-sonnet-4-6-20250514",
301	OpusModel: "claude-opus-4-6-20250610",
302	HaikuModel: "claude-haiku-4-5-20251001",
303	TimeoutMS: "600000",
304	NoAuth: true,
305	},
306	}
307
308	// Order defines the display order of providers.
309	var Order = []string{"claude", "deepseek", "mistral", "zai", "zai-coding", "minimax", "kimi", "kimi-api-cn", "kimi-api-intl", "copilot", "nvidia", "openrouter", "ollama", "ollama-cloud", "lmstudio", "vkproxy"}
310