LitSynth

A Model Context Protocol (MCP) server for intelligent academic paper discovery and semantic search using ArXiv and Datasets from Hugging Face..

python

0

GitHub Stars

python

Language

7 months ago

First Indexed

3 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "rayanechch-dev-litsynth-mcp-server": {
      "command": "fastmcp",
      "args": [
        "run",
        "my_server.py:mcp",
        "--transport",
        "http",
        "--port",
        "<your_port>"
      ]
    }
  }
}

You can run the LitSynth MCP Server to search ArXiv papers and perform semantic discovery using state-of-the-art embeddings. This MCP server exposes practical tools for querying papers, evaluating semantic relevance, and exploring related Hugging Face datasets, all accessible through a local, controllable MCP interface.

How to use

To use the LitSynth MCP Server, connect with an MCP client and invoke the available tools to search, rank, and explore academic content. You can perform basic ArXiv searches, run semantic similarity comparisons to surface the most relevant papers, and explore related datasets. Start by ensuring your server is running, then issue requests to search ArXiv, followed by optional semantic ranking on a set of papers. You can also look up Hugging Face datasets and obtain detailed dataset information or previews.

How to install

Prerequisites: Python 3.8 or newer and pip.

Step 1: Clone the LitSynth MCP Server repository.

git clone https://github.com/RayaneChCh-dev/LitSynth-MCP-Server.git

Step 2: Install dependencies from the requirements file.

pip install -r requirements.txt

Step 3: Run the MCP server using the provided start command.

fastmcp run my_server.py:mcp --transport http --port <your_port>

Notes and usage details

The server provides an ArXiv search tool with automatic URL encoding for complex terms, a semantic search tool to rank papers by relevance, and utilities to query Hugging Face datasets.
Error handling is designed to gracefully handle network issues, malformed data, and empty results, giving informative messages to guide next steps.
The system uses a semantic model to compare text and rank results, with a focus on academic content quality and relevance.

Troubleshooting

If you encounter URL encoding issues or slow semantic searches, ensure you are running the latest dependencies and that the transformer model loads correctly on first use. Initialize the semantic model once to speed up subsequent queries.

If no papers are returned, verify your query terms, try broader topics, and confirm the MCP server is reachable on the specified port.

For memory issues, the semantic model may require additional RAM. Consider adjusting batch sizes or running on a machine with more memory.

Examples

Basic ArXiv search for transformers: search_query_arxiv("attention mechanisms transformers", 5).

Semantic discovery: get papers on a broad topic, then rank them for a specific query.

Handle complex queries like: search_query_arxiv("machine learning & deep learning: survey") or search_query_arxiv("reinforcement learning (RL) applications") without manual encoding.

Available tools

greet

Simple greeting function for testing server connectivity.

search_query_arxiv

Search ArXiv database for papers matching your query; results include title, authors, abstract, arXiv link, and status.

search_semantic_arxiv

Perform semantic search on a list of papers to find the most relevant ones, returning ranked results with similarity scores.

search_hf_datasets

Search Hugging Face datasets by topic and return a ranked list of matching datasets.

get_dataset_details

Retrieve detailed information about a specific Hugging Face dataset, including metadata and usage examples.

explore_dataset_files

Explore the structure of a dataset without fully downloading it, showing available files and a sample preview.

explore_dataset_structure

Explore dataset structure with detailed split information for a given dataset.