Forge (Voxell) for Dify

Author: voxell
Version: 0.0.1
Type: model provider (text embedding)
Source: https://github.com/VoxellInc/dify-plugin-forge

Adds Forge, Voxell's OpenAI-compatible embedding API, as a text-embedding
model provider in Dify. Use it for knowledge base indexing and retrieval.

Models

Model	Dimensions	Max input
	1024	8,192 tokens
	2560	8,192 tokens
	4096	8,192 tokens

is free and rate-limited; and are billed per token. Current prices:
https://voxell.ai/pricing

Query and document embeddings

Forge embeds retrieval queries and indexed documents differently (). This plugin passes
Dify's input type through: knowledge base chunks are embedded as , and the question at
retrieval time is embedded as . The generic OpenAI-API-compatible provider sends neither, so
queries are embedded as documents.

Setup

Create an account at https://dash.voxell.ai (no card needed for ) and create an API key.
In Dify, open Settings > Model Provider, find Forge (Voxell), and click Setup.
Paste the API key.
Select a Forge model as the embedding model of a knowledge base, or as the system default
embedding model.

Requirements and connections

A Forge API key.
Outbound HTTPS from the Dify plugin runtime to .
The plugin calls only .

Privacy

See PRIVACY.md.

License

MIT