Introduction to SiliconCloud
SiliconCloud Provide cost-effective GenAI services based on excellent open source base models.
Unlike most big model cloud service platforms that only provide their own big model APIs.SiliconCloudA variety of open-source large language models and image generation models, including Qwen, DeepSeek, GLM, Yi, Mistral, LLaMA 3, SDXL, InstantID, have been uploaded to the shelf, and users are free to switch between models suitable for different application scenarios.
More importantly, SiliconCloud offersout-of-the-boxs Large Model Inference Acceleration Service to bring a more efficient user experience to your GenAI applications.
For developers, SiliconCloud provides one-click access to top open source models. This allows developers to have better application development speed and experience while significantly reducing the cost of trial and error in application development.
Official website address:/zh-cn/siliconcloud
present (sb for a job etc)
Microsoft previously open-sourced the graphrag code, project address:/microsoft/graphrag。
It is a dotnet version implemented by a big guy in China with reference to GraphRag using Semantic Kernel.
Project Address:/AIDotNet/
Don't forget to give the big guy a Star!
For those who want to learn GraphRag, you can combine these two programs together.
A quick taste of GraphRag
git clone locally and configure it according to the guidelines in the README:
Here I am using the API provided by Silicon Flow:
Silicon-based mobility provides not only dialog models but also embedded models, and is also compatible with the OpenAI format, so it is very easy to access. Here the dialog model used is Qwen/Qwen2-72B-Instruct, and the embedding model used is BAAI/bge-large-zh-v1.5.
Start the project and enter http://localhost:5000/Go to page:
Prepared a txt text about the story of Chang'e's journey to the moon, view console:
It's in the Generate Knowledge Graph now, waiting for the generation to complete.
After generating the knowledge graph, view the knowledge graph:
See what is stored in the database.
The vector database used is SQLite, which holds some vector data:
View the node table in the business database:
View the boundary table:
Probably because there is not a lot of text and there is no information in the other tables.
Then it's time to Rag based on the knowledge graph.
Ask a question like, "What gift did Chang'e give to the emperor?" :
Answer correctly.
If after the experience, you are really interested in GraphRag, you can delve deeper into the principles of GraphRag and take a look at the two programs together.