ai-servers/new-api/CHANNELS.md

new-api Channel Configuration

After first start, access the new-api web UI at http://<server>:4000 to configure channels.

Default admin credentials: root / 123456 — change immediately.

API Token for Open WebUI

Create an API token in new-api's token management. Use this token as OPENWEBUI_API_KEY in .env.

Channels to Create

Configure each channel via Channels > Add Channel in the web UI.

1. DeepInfra (Priority 1)

Field	Value
Name	DeepInfra
Type	OpenAI
Base URL	https://api.deepinfra.com/v1/openai
Key	$DEEPINFRA_API_KEY
Priority	1
Models	See model mapping below

2. SiliconFlow (Priority 2)

Field	Value
Name	SiliconFlow
Type	OpenAI
Base URL	https://api.siliconflow.com/v1
Key	$SILICONFLOW_API_KEY
Priority	2
Models	See model mapping below

3. OpenRouter (Priority 3)

Field	Value
Name	OpenRouter
Type	OpenAI
Base URL	https://openrouter.ai/api/v1
Key	$OPENROUTER_API_KEY
Priority	3
Models	See model mapping below

4. Groq (Priority 1)

Field	Value
Name	Groq
Type	OpenAI
Base URL	https://api.groq.com/openai/v1
Key	$GROQ_API_KEY
Priority	1
Models	llama-3.3-70b

5. Cerebras (Priority 1)

Field	Value
Name	Cerebras
Type	OpenAI
Base URL	https://api.cerebras.ai/v1
Key	$CEREBRAS_API_KEY
Priority	1
Models	llama-3.3-70b-cerebras

Model Mapping per Channel

new-api uses model aliasing: the "model name" is what clients see, the "actual model" is what's sent to the provider.

DeepInfra Models

Client Model Name	Actual Provider Model
deepseek-v3.2	deepseek-ai/DeepSeek-V3.2
deepseek-r1	deepseek-ai/DeepSeek-R1
gpt-oss	openai/gpt-oss-120b
gpt-oss-20b	openai/gpt-oss-20b
nemotron-super	nvidia/Llama-3.3-Nemotron-Super-49B-v1.5
nemotron-nano	nvidia/NVIDIA-Nemotron-Nano-9B-v2
devstral	mistralai/Devstral-Small-2505
glm-4.6	zai-org/GLM-4.6
glm-4.7	zai-org/GLM-4.7
glm-5	zai-org/GLM-5
kimi-k2	moonshotai/Kimi-K2-Instruct-0905
kimi-k2.5	moonshotai/Kimi-K2.5
deepseek-v3-free	deepseek-ai/DeepSeek-V3

SiliconFlow Models

Client Model Name	Actual Provider Model
deepseek-v3.2	deepseek-ai/DeepSeek-V3.2
glm-4.7	THUDM/GLM-4-32B-0414
kimi-k2	moonshotai/Kimi-K2-Instruct-0905
qwen3-coder	Qwen/Qwen3-Coder-480B-A35B-Instruct
qwen3-coder-30b	Qwen/Qwen3-Coder-30B-A3B-Instruct

OpenRouter Models

Client Model Name	Actual Provider Model
deepseek-v3.2	deepseek/deepseek-chat-v3-0324
deepseek-v3-free	deepseek/deepseek-chat-v3-0324:free
kimi-k2.5	moonshotai/kimi-k2.5
minimax-m2.5	minimax/minimax-m2.5
gpt-4.1-mini	openai/gpt-4.1-mini
gpt-4.1	openai/gpt-4.1
gemini-3-flash-preview	google/gemini-3-flash-preview
gemini-2.5-pro	google/gemini-2.5-pro-preview
claude-sonnet	anthropic/claude-sonnet-4
trinity-large-preview	arcee-ai/trinity-large-preview

Groq Models

Client Model Name	Actual Provider Model
llama-3.3-70b	llama-3.3-70b-versatile

Cerebras Models

Client Model Name	Actual Provider Model
llama-3.3-70b-cerebras	llama-3.3-70b

Fallback Behavior

new-api handles fallbacks via priority levels:

• When a model exists on multiple channels, the highest priority (lowest number) channel is tried first
• If it fails, it automatically falls back to the next priority level

For example, deepseek-v3.2 exists on:

1. DeepInfra (priority 1) — tried first
2. SiliconFlow (priority 2) — fallback
3. OpenRouter (priority 3) — last resort

Grafana Setup

After first start, access Grafana at http://<server>:3001:

1. Login with admin / $GRAFANA_ADMIN_PASSWORD
2. Add data source: Prometheus with URL http://victoriametrics:8428
3. Import dashboards:
   • Node Exporter Full: dashboard ID 1860
   • Redis: dashboard ID 763