# Typesense Vector Search AI Agent Connect > Typesense Vector Search lets you connect your vector database to your AI agent. You can create collections, index JSON data, and run semantic queries with text filters without writing extra code. It handles the heavy lifting of vector similarity so you can focus on your RAG pipeline. ## Overview - **Category:** loved-by-devs - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_bmlNohSGzzJmxN6goNUndx2uYrEvS88xXUYpe92K/ai-agent-connect - **Tags:** vector-search, semantic-search, rag, embedding-management, document-indexing ## Description Managing vector embeddings usually means jumping between your database console and your code editor to make sure your data is actually getting indexed correctly. This Connector changes that by giving your AI agent direct access to your vector database environment. Instead of writing custom scripts or manually formatting CURL commands to update your collections, you can just tell your agent to do it. You can have it set up new schemas for specific embedding dimensions, push new JSON payloads into your index, or run complex queries that combine text filters with vector similarity. It's a huge time-saver for anyone building RAG systems who needs to move fast. By using this via the Vinkius catalog, you get a clean way to manage your data lifecycle through simple conversation. You can check your collection geometries to make sure your fields are mapped right or wipe out old documents instantly. It turns your database management into a natural conversation with your agent, letting you focus on the actual logic of your application rather than the plumbing of your vector store. You don't have to worry about the specific syntax for multi-dimensional arrays or the nuances of similarity scores anymore. Your agent handles the heavy lifting of data entry and retrieval, so you can spend your time refining your prompts and improving your application's accuracy. ## Tools ### search_vectors Run a similarity search using a vector query and optional text filters. This helps you find relevant data quickly. ### create_collection Build a new search collection with a custom JSON schema. Use this to set up new datasets for your AI agent. ### delete_document Permanently remove a document from a collection using its unique ID. This is useful for cleaning up old data. ### list_vector_collections See a list of all collections currently in your Typesense instance. This gives you a quick overview of your setup. ### get_collection_details View the schema and metadata for a specific collection. Use this to verify your data structure is correct. ### index_document Add or update a JSON document within your search collection. This lets you push new data without writing code. ## Prompt Examples **Prompt:** ``` Show me all the collections I have in my Typesense instance right now. ``` **Response:** ``` I've listed the following active collections in your instance: | Collection Name | Status | | :--- | :--- | | `customer_kb_index` | Active | | `products_inventory` | Active | Would you like to see the schema details for either of these? ``` **Prompt:** ``` I need to add a new product to the 'inventory' collection. Here is the data: {"id": "p-99", "name": "Wireless Mouse", "popularity": 45, "embeddings": [0.12, 0.88, ...]} ``` **Response:** ``` I've successfully pushed that new product into the `products_inventory` collection. The document is now live and ready for semantic search. ``` **Prompt:** ``` What is the schema for my 'user_kb' collection? ``` **Response:** ``` The `user_kb` collection is configured with the following schema: - `user_id` (string) - `query_text` (string) - `embeddings` (float[]) It's currently set up to handle 768-dimensional vectors. ``` ## Capabilities ### Provision vector collections Create new semantic schema datasets holding complex vector embedding structures. ### Index JSON payloads Insert or update documents into your database without writing manual REST integrations. ### Run hybrid semantic searches Execute combined text-filtering and vector similarity queries natively through chat. ### Inspect schema geometries Retrieve absolute schema mappings to ensure fields are correctly mapped. ### Delete specific documents Permanently remove documents from a collection by their unique ID. ### List all active collections View all collections across your entire vector instance at once. ## Use Cases ### Fixing a broken RAG index A developer asks the agent to delete a faulty document and re-index a corrected JSON payload using `delete_document` and `index_document`. ### Rapid prototyping of new categories An app builder creates three new collections for different product categories in minutes using `create_collection`. ### Data auditing and verification A data engineer asks the agent to list all collections and check the schema of the main knowledge base using `list_vector_collections` and `get_collection_details`. ### Testing hybrid search accuracy A developer runs several `search_vectors` queries to see how specific text filters impact the final similarity scores. ## Benefits - Skip the manual coding for indexing by using `index_document` to push JSON data directly into your vector store. - Quickly set up new datasets with `create_collection` to match your specific embedding dimensions in seconds. - Run complex hybrid queries with `search_vectors` to combine text filters and vector similarity in one go. - Audit your database structure instantly with `get_collection_details` to ensure your fields are mapped correctly. - Manage your data lifecycle easily by using `delete_document` to remove outdated or incorrect records. - Get a birds-eye view of your entire setup with `list_vector_collections` to track all active indexes. ## How It Works The bottom line is you get a direct line from your chat interface to your vector database. 1. Connect your Typesense Host URL and Admin API Key to the Connector. 2. Describe the collection, schema, or data updates you want to perform to your agent. 3. Receive immediate confirmation of the index update or the results of your semantic search. ## Frequently Asked Questions **Can the Typesense Vector Search MCP create new collections?** Yes, it can provision new collections with specific schemas. You just describe the fields and embedding dimensions to your agent, and it handles the creation for you. **How do I update a document in my vector index?** You can just tell your agent to update a document using its ID and provide the new JSON data. The Connector will handle the update automatically. **Can I run hybrid searches with this?** Yes, it supports combined text-filtering and vector similarity queries. This allows you to narrow down results by category or name while still using semantic search. **Is this for managing my embeddings?** This Connector manages the indexing, storage, and retrieval of your embeddings. It connects your AI agent directly to your existing Typesense vector database. **Can I delete specific records using the Typesense Vector Search MCP?** Yes, you can ask your agent to permanently remove a specific document from any collection by providing its unique ID. **Does this work for RAG systems?** It's a perfect fit for RAG. It allows you to manage your knowledge base documents and perform the actual semantic searches that power RAG applications. **Can the agent perform vector plus text-filtering search combined natively?** Yes. Provide the agent with the collection name alongside the text payload and tell it the exact vector structure. It leverages internal filters querying natively and returns the nearest neighbors with exact accuracy scores. **How do I make the AI create a semantic collection ready for embeddings (OpenAI 1536 dims)?** Ask the agent to use 'create_collection'. Provide standard JSON declaring the name, the field structure, and explicitly define the `float[]` field tracking the 1536 dims length. The cluster will spin the framework up instantly. **Can it delete problematic vectors holding bad geometry data manually?** Absolutely. Supplying the explicit collection target and the item 'id' to the delete_document prompt securely wipes out all traces from the dataset. Use this sparingly as it can't be undone easily.