import os
from mcp.server.fastmcp import FastMCP
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_community.vectorstores import SKLearnVectorStore
# Define common path to the repo locally
PATH = os.path.dirname(os.path.abspath(__file__))
# Create an MCP server
mcp = FastMCP("Gemini-CLI-Docs-MCP-Server")
# Add a tool to query the Gemini CLI documentation
@mcp.tool()
def gemini_cli_query_tool(query: str):
"""
Query the Gemini CLI documentation using a retriever.
Args:
query (str): The query to search the documentation with
Returns:
str: A str of the retrieved documents
"""
retriever = SKLearnVectorStore(
embedding=HuggingFaceEmbeddings(model_name="BAAI/bge-large-en-v1.5"),
persist_path=os.path.join(PATH, "gemini_cli_vectorstore.parquet"),
serializer="parquet").as_retriever(search_kwargs={"k": 3}
)
relevant_docs = retriever.invoke(query)
print(f"Retrieved {len(relevant_docs)} relevant documents")
formatted_context = "\n\n".join([f"==DOCUMENT {i+1}==\n{doc.page_content}" for i, doc in enumerate(relevant_docs)])
return formatted_context
# The @mcp.resource() decorator is meant to map a URI pattern to a function that provides the resource content
@mcp.resource("docs://gemini-cli/full")
def get_all_gemini_cli_docs() -> str:
"""
Get all the Gemini CLI documentation. Returns the contents of the file gemini_cli_docs.txt,
which contains a curated set of Gemini CLI documentation. This is useful
for a comprehensive response to questions about Gemini CLI.
Args: None
Returns:
str: The contents of the Gemini CLI documentation
"""
# Local path to the Gemini CLI documentation
doc_path = os.path.join(PATH, "gemini_cli_docs.txt")
try:
with open(doc_path, 'r') as file:
return file.read()
except Exception as e:
return f"Error reading log file: {str(e)}"
if __name__ == "__main__":
# Initialize and run the server
mcp.run(transport='stdio')
