Flowise is an open-source, low-code builder for LLM apps and agents. Its OpenAI Custom Model chat model node is built for OpenAI-compatible APIs: you type the model name and set a base path. BroadNet Inference serves chat completions at https://inference.broadnet.ai/v1, so that node is the one to use. BroadRouter is BroadNet’s own site; the steps follow the node’s source in Flowise 3.1.4, because Flowise’s docs have no page for this node.
Setup
-
Create a credential. In Flowise, open Credentials, add an OpenAI API credential, name it “BroadNet” and paste your BroadNet key into OpenAI Api Key.
-
Add the node. In a chatflow, add Chat Models → OpenAI Custom Model.
-
Fill in the node:
Field Value Connect Credential the BroadNet credential Model Name broadnet/glm-5.3Temperature 0.9 by default; lower it for extraction Streaming (Additional Parameters) on (the default) Max Tokens (Additional Parameters) 4000 Base Path (Additional Parameters) https://inference.broadnet.ai/v1 -
Connect it to a chain or agent node, save, and send a test message in the chat panel. A reply means the key, URL and model name are right.
What the fields do
- Model Name is sent as the model ID. Type BroadNet’s ID with the
broadnet/prefix. The node doesn’t fetch a model list, so BroadNet’s/v1/models, which listed only the transcription models when we checked on 16 November 2026, doesn’t get in the way. - Base Path becomes the client’s base URL. Flowise appends
/chat/completions. - Max Tokens is sent as
max_tokens. DeepSeek V4.1 Flash, GLM 5.3 and Qwen3.8 27B reason before they answer, and the reasoning counts toward it, so don’t set it low. - Base Options takes JSON headers to add to every request. BroadNet doesn’t need any.
Why not the standard ChatOpenAI node?
The ChatOpenAI node also has a Base Path field, but in Flowise 3.1.4 its Model Name is a drop-down filled by a listModels method, with gpt-4o-mini as the default. The OpenAI Custom Model node has a free-text Model Name instead, and its source describes it as being for an “OpenAI Chat compatible API”, which is what BroadNet provides.
Which model to pick
| Flow | Model |
|---|---|
| Tool agents, structured output parsers | broadnet/glm-5.3 |
| Agents that read images or Arabic text | broadnet/qwen3.8-27b |
| Conversational and retrieval chains, long documents | broadnet/deepseek-v4.1-flash |
| Trying Flowise on the starting credit | broadnet/gemma-4-26b-a4b (32,768-token context) |
BroadNet’s docs list tools and structured output for GLM 5.3 and Qwen3.8 27B only. Prices for every model.
How we tested this
We did not run a Flowise server. On 2026-11-15 we read the OpenAI Custom Model node’s source at the flowise@3.1.4 tag, then built the same ChatOpenAI object it builds, using @langchain/openai 1.2.5, the version Flowise 3.1.4 pins. We sent a streamed message and a tool call to a local mock server that speaks BroadNet’s chat-completions shape. Both requests went to /v1/chat/completions with the model ID broadnet/glm-5.3 unchanged, a Bearer key, max_tokens: 4000, temperature: 0.9 and stream: true, and the tool call came back parsed. We had no BroadNet key, so we did not test real model output. The field labels come from the source; check them against your Flowise version.
Troubleshooting
- 401. The credential holds the wrong key, or the key is expired or out of allowance.
- 404. Base Path must end in
/v1. Model Name must match BroadNet’s ID exactly. Paid models need a top-up; the $5 credit covers onlybroadnet/gemma-4-26b-a4band the two Transcribe models. - Empty replies. Raise Max Tokens. Why answers come back empty.
- The agent answers without using tools. Switch to
broadnet/glm-5.3orbroadnet/qwen3.8-27b. - 429 under load. A pay-as-you-go key allows 30 requests a minute and 4 in flight. See errors and retries.
