This is a plugin for Koishi that stroages SNS message records and create vectorized embeddings.
⚠️ WARNING! This plugin will automatically embed SNS messages to vector database! Please make sure you have enough money in your account!
- [✅] Dump all SNS messages to database. SQLite as default, MySQL supported.
- [❌] More platform adapter support. Only Onebot is tested now.
- [✅] Store more Metadata for SNS messages so that we can use them in complicated RAG callback rules.
- [✅] Store MultiModal messages (Image, Video, Audio, etc.) to koishi assets. Local storage as default, S3 compatible or others not tested yet.
- [❌] More OSS provider support. Alibaba Cloud, Tencent Cloud, Cloudflare R2, etc.
- [❌] Use ASR model, VLM model, optical flow, etc. for MultiModal semantic understanding. Create vectorized embeddings for MultiModal messages. This is for further RAG callback.
- [✅] Create vectorized embeddings with OpenAI compatible API for SNS messages. Only Zhipu API is tested now.
- [✅] Store vectorized embeddings to vector database. Only Milvus is supported now.
- [❌] Create more RAG callback rules and NLP models workflow for message reply. Prompt engineering nessitates in this step.
- [❌] Cluster deployment parameters for Database, Vector Database, Object Storage, etc.
| Category | Field | Description |
|---|---|---|
| Database | table_name |
Table name for SQL Database. |
| Milvus | db_name |
DB name for Milvus. Only set as "default" can create collection automatically. Otherwise, you need to create collection manually. |
| Milvus | collection_name |
Collection name for Milvus. |
| Milvus | consistency_level |
Consistency level for Milvus. Select one of "Strong", "Bounded Staleness", "Session", "Eventually". This parameter is only effective in cluster deployment. |
| Milvus/Embedding | embeddingDimension |
Embedding dimension for vectorized embeddings and vector database. |
| Embedding | model |
Embedding model name. |
| Embedding | apiUrl |
API URL for embedding model. Including endpoint. |
| Embedding | apiKey |
API Key for embedding model. |
| Field | Parameters |
|---|---|
table_name |
sns_record |
db_name |
default |
collection_name |
sns_record |
consistency_level |
Eventually |
embeddingDimension |
1536 |
model |
embedding-3 |
apiUrl |
https://open.bigmodel.cn/api/paas/v4/embeddings |
apiKey |
your_api_key_here |