Semantic Search
🔎 Semantic search cli.
Setup
Installation
You can compile the program from source by executing:
The compiled binary will be located at ./target/release/sense. You can move it to a directory in your PATH to use it globally.
Configuration
Grab an API key from SiliconCloud (free) and set it in the configuration file. The configuration file is located at .sense/config.toml and should look like this:
[]
= 8080 # Default port for the server (Optional)
[]
= "sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx" # API key for SiliconCloud (Required)
= "BAAI/bge-large-zh-v1.5" # Model to use for embedding (Optional)
# Available models: BAAI/bge-large-zh-v1.5, BAAI/bge-large-en-v1.5, netease-youdao/bce-embedding-base_v1, BAAI/bge-m3, Pro/BAAI/bge-m3
[] # Only required for `sense bot`
= "1234567890:xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx" # Telegram bot token (Required)
= 1234567890 # Telegram user ID of the bot owner (Required)
= [] # Whitelisted user IDs - Only these users can use the bot (Optional, all users can use the bot if not set or set to an empty array)
= "meme" # Sticker set id prefix for the bot (Optional, an additional `_by_<bot_username>` will be appended to form the full sticker set id)
= 8 # Number of results to return (Optional)
Indexing
(Re-)Index the files you want to search for by executing the following command:
This will generate or update index of the files, their hashes, labels and embeddings in .sense/index.db3. Note that each time you add or remove files, you need to re-run this process.
If files are created or changed, running this command will prompt you to label them (again). You can use any tool of your choice to label them automatically. See DEV.md for more information on the database schema.
Usage
Command Line Interface
To search for files based on labels, execute the following command:
You can specify how many results you want to display using the --num-results flag, or -n for short:
(TBD) Also, you can specify the regular expression for the path of the files using the --path flag, or -p for short:
Telegram Bot
You can start a Telegram bot to search for files using a chat interface:
See TELEGRAM.md for detailed instructions on deploying the bot.
Server
You can start a server to search for files using a REST API:
Which will start a server on port 8080. You can then search for files using the following endpoint:
}
☑️ TODOs
- Implement aforementioned features
- Indexing
- Labeling
- Embedding
- Searching
- Telegram bot
- Server
- Incremental update
- Better error handling
- Reduce using of
unwrap(), so as to make the program more robust to network failures - Error logging
- Reduce using of
- Reducing clone
- Enhance Telegram bot
- Graceful shutdown
- Parallel processing
- Using chunks to upload stickers
- Auto release binary