import {McpServer} from '@modelcontextprotocol/sdk/server/mcp.js';
import {Pinecone} from '@pinecone-database/pinecone';
import {z} from 'zod';
const INSTRUCTIONS = 'Create a Pinecone index with integrated inference';
const SCHEMA = {
name: z.string().describe('A unique name to identify the new index.'),
embed: z
.object({
model: z
.enum(['multilingual-e5-large', 'llama-text-embed-v2', 'pinecone-sparse-english-v0'])
.describe(
`Choose an embedding model:
- "multilingual-e5-large" is an efficient dense embedding model
trained on a mixture of multilingual datasets. It works well on messy
data and short queries expected to return medium-length passages of
text (1-2 paragraphs).
- "llama-text-embed-v2" is a high-performance dense embedding model
optimized for text retrieval and ranking tasks. It is trained on a
diverse range of text corpora and provides strong performance on
longer passages and structured documents.
- "pinecone-sparse-english-v0" is a sparse embedding model for
converting text to sparse vectors for keyword or hybrid search. The
model directly estimates the lexical importance of tokens by
leveraging their context.`,
),
fieldMap: z
.object({
text: z.string().describe(
`The name of the field in the data records that contains the text
content to embed. Records in the index must contain this field.`,
),
})
.describe('Identify which field from your data records will be embedded.'),
})
.describe('Configure an embedding model that converts text into a vector.'),
};
export function addCreateIndexForModelTool(server: McpServer, pc: Pinecone) {
server.tool('create-index-for-model', INSTRUCTIONS, SCHEMA, async ({name, embed}) => {
// Check if the index already exists
const existingIndexes = await pc.listIndexes();
const existingIndex = existingIndexes.indexes?.find((index) => index.name === name);
if (existingIndex) {
return {
content: [
{
type: 'text',
text: `Index not created. An index named "${name}" already exists:
${JSON.stringify(existingIndex, null, 2)}`,
},
],
};
}
// Create the new index
const indexInfo = await pc.createIndexForModel({
name,
cloud: 'aws',
region: 'us-east-1',
embed,
tags: {
source: 'mcp',
embedding_model: embed.model,
},
waitUntilReady: true,
});
return {
content: [
{
type: 'text',
text: JSON.stringify(indexInfo, null, 2),
},
],
};
});
}
