73 lines
3.9 KiB
Markdown
73 lines
3.9 KiB
Markdown
---
|
|
description: Learn about the Different Models Supported by Dify.
|
|
---
|
|
|
|
# Model Configuration
|
|
|
|
Dify supports major model providers like OpenAI's GPT series and Anthropic's Claude series. Each model's capabilities and parameters differ, so select a model provider that suits your application's needs. **Obtain the API key from the model provider's official website before using it in Dify.**
|
|
|
|
## Model Types in Dify
|
|
|
|
Dify classifies models into 4 types, each for different uses:
|
|
|
|
1. **System Inference Models:** Used in applications for tasks like chat, name generation, and suggesting follow-up questions.
|
|
|
|
> Providers include [OpenAI](https://platform.openai.com/account/api-keys)、[Azure OpenAI Service](https://azure.microsoft.com/en-us/products/ai-services/openai-service/)、[Anthropic](https://console.anthropic.com/account/keys)、Hugging Face Hub、Replicate、Xinference、OpenLLM、[iFLYTEK SPARK](https://www.xfyun.cn/solutions/xinghuoAPI)、[WENXINYIYAN](https://console.bce.baidu.com/qianfan/ais/console/applicationConsole/application)、[TONGYI](https://dashscope.console.aliyun.com/api-key\_management?spm=a2c4g.11186623.0.0.3bbc424dxZms9k)、[Minimax](https://api.minimax.chat/user-center/basic-information/interface-key)、ZHIPU(ChatGLM) [Ollama](https://docs.dify.ai/tutorials/model-configuration/ollama)、[LocalAI](https://github.com/mudler/LocalAI)、.
|
|
2. **Embedding Models:** Employed for embedding segmented documents in knowledge and processing user queries in applications.
|
|
|
|
> Providers include OpenAI, ZHIPU (ChatGLM), Jina AI([Jina Embeddings 2](https://jina.ai/embeddings/)).
|
|
3. [**Rerank Models**](https://docs.dify.ai/advanced/retrieval-augment/rerank)**:** Enhance search capabilities in LLMs.
|
|
|
|
> Provider: Cohere.
|
|
4. **Speech-to-Text Models:** Convert spoken words to text in conversational applications.
|
|
|
|
> Provider: OpenAI.
|
|
|
|
Dify plans to add more LLM providers as technology and user needs evolve.
|
|
|
|
## Hosted Model Trial Service
|
|
|
|
Dify offers trial quotas for cloud service users to experiment with different models. Set up your model provider before the trial ends to ensure uninterrupted application use.
|
|
|
|
* OpenAI Hosted Model Trial: Includes 200 invocations for models like GPT3.5-turbo, GPT3.5-turbo-16k, text-davinci-003 models.
|
|
|
|
## Setting the Default Model
|
|
|
|
Dify automatically selects the default model based on usage. Configure this in `Settings > Model Provider`.
|
|
|
|
<figure><img src="../../.gitbook/assets/image-default-models.png" alt=""><figcaption></figcaption></figure>
|
|
|
|
## Model Integration Settings
|
|
|
|
Choose your model in Dify's `Settings > Model Provider`.
|
|
|
|
<figure><img src="../../.gitbook/assets/image-20231210143654461.png" alt=""><figcaption></figcaption></figure>
|
|
|
|
Model providers fall into two categories:
|
|
|
|
1. Proprietary Models: Developed by providers such as OpenAI and Anthropic.
|
|
2. Hosted Models: Offer third-party models, like Hugging Face and Replicate.
|
|
|
|
Integration methods differ between these categories.
|
|
|
|
**Proprietary Model Providers:** Dify connects to all models from an integrated provider. Set the provider's API key in Dify to integrate.
|
|
|
|
{% hint style="info" %}
|
|
Dify uses [PKCS1\_OAEP](https://pycryptodome.readthedocs.io/en/latest/src/cipher/oaep.html) encryption to protect your API keys. Each user (tenant) has a unique key pair for encryption, ensuring your API keys remain confidential.
|
|
{% endhint %}
|
|
|
|
**Hosted Model Providers:** Integrate third-party models individually.
|
|
|
|
Specific integration methods are not detailed here.
|
|
|
|
* [Hugging Face](https://docs.dify.ai/advanced/model-configuration/hugging-face)
|
|
* [Replicate](https://docs.dify.ai/advanced/model-configuration/replicate)
|
|
* [Xinference](https://docs.dify.ai/advanced/model-configuration/xinference)
|
|
* [OpenLLM](https://docs.dify.ai/advanced/model-configuration/openllm)
|
|
|
|
## Using Models
|
|
|
|
Once configured, these models are ready for application use.
|
|
|
|
<figure><img src="../../.gitbook/assets/choice-model-in-app.png" alt=""><figcaption></figcaption></figure>
|