Archiver is a command-line application designed to interact with large language models (LLMs) for testing purposes. It leverages Ollama for model integration and SurrealDB as the vector store. The application uses Langchain-rust as the framework for interacting with the Ollama and SurrealDB.
Prerequisites
Before setting up Archiver, ensure you have the following installed on your system:
- Rust (latest stable version)
just
(command runner)- Pandoc (load text when adding to vector db)
- Ollama
- SurrealDB
Quick Start
Ensure docker is running and make sure docker compose is setup
Install Just to run the setup and Pandocs for file indexing to surrealdb
Once Pandocs and Just are installed run
Add file to SurrealDB
Or
Add directory of Markdown
Finally Run Archiver and use it
Change Log
v0.3.2
- Quick Start with Docker: Update to Justfile and Readme to walk through setting up Archiver with docker
v0.3.1
- Check for Pandoc: Ensure pandoc exists before running
v0.3.0
- Index Directory: Add Directory Indexing for bulk Markdown file importing
v0.2.0
- Prompt: Add prompt identifier and colored response from LLM
- Error Handling: Better error handling with SurrealDB and Ollama
- Chuncking: Added document chunking when adding files to SurrealDB
- Mode Switch: Switched LLM chain to a Conversational Retriever from Conversational, making the LLM more efficient with responses from the VectorDB
v0.1.0
- Conversational Memory: Maintain context across interactions with the LLM.
- Vector Stores: Efficient storage and retrieval of high-dimensional vectors through SurrealDB.
- Ollama Integration: Easy setup and use of Ollama for model management.
- Indexing Files: Add Markdown files to SurrealDB.
Install Archiver
-
Install using Cargo:
Make sure you have SurrealDB setup and Ollama before running Archiver
-
Download and Install SurrealDB:
|
-
Run SurrealDB:
Setting up SurrealDB
Once Surreal is up and running you can use https://surrealist.app/ to access the instance of Surreal. But before doing that a Namespace and Table need to be setup. Fastest way to get this done is to run Archiver once.
-
Set Namespace and Table:
-
Add Markdown File to the SurrealDB:
Archiver
- Setup from Source:
License
This project is licensed under the Apache Livense (Version 2.0). See the LICENSE file for details.