openclaw-trapi-config/references/custom-provider-config.md

trapi 模型与配置参考

目录

• Provider 配置
• 预置模型
• 别名
• 图像模型
• 当前 OpenClaw 写入方式
• 故障排查

Provider 配置

trapi 使用混合模型适配器。Provider 级 api 只是默认值；模型级 api 优先。

{
  "baseUrl": "https://lapi.transiglobal.com",
  "apiKey": {
    "source": "env",
    "provider": "default",
    "id": "TRAPI_API_KEY"
  },
  "api": "anthropic-messages"
}

不要在配置示例中放真实 API Key。

预置模型

所有模型暂时保留零费用元数据。不要给任何模型写 reasoning 字段，让 OpenClaw 和上游按模型默认行为处理。

模型 ID	API	输入	contextWindow	maxTokens
GLM-5-Turbo	anthropic-messages	text	200000	64000
GLM-5.2	anthropic-messages	text	200000	64000
GLM-4.5-Air	anthropic-messages	text	200000	64000
MiniMax-M2	anthropic-messages	text	200000	64000
MiniMax-M2.7	anthropic-messages	text	200000	64000
MiniMax-M3	anthropic-messages	text, image	1000000	131072
kimi-for-coding	anthropic-messages	text, image	262144	32768
GLM-5V-Turbo	anthropic-messages	text, image	200000	64000
PaddleOCR-VL-1.5	anthropic-messages	text, image	131072	8192
deepseek-v4-pro	anthropic-messages	text	1000000	384000
deepseek-v4-flash	anthropic-messages	text	1000000	384000
gpt-5.5	openai-responses	text	200000	65536
gpt-5.6-sol	openai-responses	text, image	372000	128000
gpt-5.6-terra	openai-responses	text, image	372000	128000
gpt-5.6-luna	openai-responses	text, image	372000	128000
mimo-v2.5-pro	anthropic-messages	text	200000	64000
mimo-v2.5	anthropic-messages	text, image, video, audio	200000	64000
claude-opus-4.7	anthropic-messages	text	300000	64000
qwen3.7-max	anthropic-messages	text	1000000	100000
qwen3.6-plus	anthropic-messages	text, image	1000000	100000

完整模型数组：

[
  {"id":"GLM-5-Turbo","name":"GLM-5-Turbo (Transiglobal)","api":"anthropic-messages","input":["text"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":200000,"maxTokens":64000},
  {"id":"GLM-5.2","name":"GLM-5.2 (Transiglobal)","api":"anthropic-messages","input":["text"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":200000,"maxTokens":64000},
  {"id":"GLM-4.5-Air","name":"GLM-4.5-Air (Transiglobal)","api":"anthropic-messages","input":["text"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":200000,"maxTokens":64000},
  {"id":"MiniMax-M2","name":"MiniMax M2 (Transiglobal)","api":"anthropic-messages","input":["text"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":200000,"maxTokens":64000},
  {"id":"MiniMax-M2.7","name":"MiniMax-M2.7 (Transiglobal)","api":"anthropic-messages","input":["text"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":200000,"maxTokens":64000},
  {"id":"MiniMax-M3","name":"MiniMax M3 (Transiglobal)","api":"anthropic-messages","input":["text","image"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":1000000,"maxTokens":131072},
  {"id":"kimi-for-coding","name":"kimi-for-coding (Transiglobal)","api":"anthropic-messages","input":["text","image"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":262144,"maxTokens":32768},
  {"id":"GLM-5V-Turbo","name":"GLM-5V-Turbo (Transiglobal)","api":"anthropic-messages","input":["text","image"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":200000,"maxTokens":64000},
  {"id":"PaddleOCR-VL-1.5","name":"PaddleOCR-VL-1.5 (Transiglobal)","api":"anthropic-messages","input":["text","image"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":131072,"maxTokens":8192},
  {"id":"deepseek-v4-pro","name":"DeepSeek V4 Pro (Transiglobal)","api":"anthropic-messages","input":["text"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":1000000,"maxTokens":384000},
  {"id":"deepseek-v4-flash","name":"DeepSeek V4 Flash (Transiglobal)","api":"anthropic-messages","input":["text"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":1000000,"maxTokens":384000},
  {"id":"gpt-5.5","name":"GPT-5.5 (Transiglobal)","api":"openai-responses","input":["text"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":200000,"maxTokens":65536},
  {"id":"gpt-5.6-sol","name":"GPT-5.6 Sol (Transiglobal)","api":"openai-responses","input":["text","image"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":372000,"maxTokens":128000},
  {"id":"gpt-5.6-terra","name":"GPT-5.6 Terra (Transiglobal)","api":"openai-responses","input":["text","image"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":372000,"maxTokens":128000},
  {"id":"gpt-5.6-luna","name":"GPT-5.6 Luna (Transiglobal)","api":"openai-responses","input":["text","image"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":372000,"maxTokens":128000},
  {"id":"mimo-v2.5-pro","name":"MiMo 2.5 Pro (Transiglobal)","api":"anthropic-messages","input":["text"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":200000,"maxTokens":64000},
  {"id":"mimo-v2.5","name":"MiMo 2.5 (Transiglobal)","api":"anthropic-messages","input":["text","image","video","audio"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":200000,"maxTokens":64000},
  {"id":"claude-opus-4.7","name":"Claude Opus 4.7 (Transiglobal)","api":"anthropic-messages","input":["text"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":300000,"maxTokens":64000},
  {"id":"qwen3.7-max","name":"Qwen 3.7 Max (Transiglobal)","api":"anthropic-messages","input":["text"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":1000000,"maxTokens":100000},
  {"id":"qwen3.6-plus","name":"Qwen 3.6 Plus (Transiglobal)","api":"anthropic-messages","input":["text","image"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":1000000,"maxTokens":100000}
]

别名

模型 ID	别名
GLM-5-Turbo	glm5t
GLM-5.2	glm52
GLM-4.5-Air	glm45a
MiniMax-M2	mxm2
MiniMax-M2.7	mxm27
MiniMax-M3	mxm3
kimi-for-coding	kimi
GLM-5V-Turbo	glm5v
PaddleOCR-VL-1.5	pocr
deepseek-v4-pro	dsv4p
deepseek-v4-flash	dsv4f
gpt-5.5	gpt55
gpt-5.6-sol	g56s
gpt-5.6-terra	g56t
gpt-5.6-luna	g56l
mimo-v2.5-pro	mimo25p
mimo-v2.5	mimo25
claude-opus-4.7	opus47
qwen3.7-max	qwn37
qwen3.6-plus	qwn36

别名没有强制五字符限制。优先简短、清晰、稳定，不自动改动用户已有别名。

对话中使用：

/model glm5t
/model mxm3
/model g56s

图像模型

支持图片输入：

• MiniMax-M3
• kimi-for-coding
• GLM-5V-Turbo
• PaddleOCR-VL-1.5
• gpt-5.6-sol
• gpt-5.6-terra
• gpt-5.6-luna
• mimo-v2.5
• qwen3.6-plus

imageModel 仅用于图像理解，不等同于图像生成模型。

当前 OpenClaw 写入方式

优先使用：

openclaw config set models.providers.trapi '<provider-json>' --strict-json --merge
openclaw config set models.providers.trapi.models '<model-array-json>' --strict-json --merge
openclaw models aliases add <alias> trapi/<model-id>
openclaw models set-image trapi/<model-id>
openclaw models image-fallbacks add trapi/<model-id>

不要把模型或别名更新写成 Agent 侧 gateway config.patch。当前 OpenClaw 会保护全局模型配置路径，且 Patch 中数组不是按模型 ID 自动追加。

故障排查

• 401/403：检查 Gateway 服务是否能解析 TRAPI_API_KEY SecretRef。
• Model not found：核对模型 ID 大小写及 trapi 实际开放目录。
• 协议错误：核对模型级 api，GPT 模型使用 openai-responses，其余目录条目使用各自声明值。
• 别名冲突：执行 openclaw models aliases list --json，由用户决定保留或重定向。
• 配置写入被拒绝：重新读取目标路径并使用 --merge；不要直接改用 --replace。
• 图像未传入模型：确认模型 input 包含 image，并检查 imageModel 配置。
• available: true 但调用失败：该状态不证明上游推理可用，需要经用户确认后做最小实时请求。