Heidelberg AICurriculum

I Built an AI App to Analyze My Own Business Data (No Code)

Tech With Tim ·2026-07-06 ·2 min read

Summary written by us from the video's transcript. The video, and everything in it, is Tech With Tim's work.

Learn how to build a no‑code Retrieval Augmented Generation (RAG) app in KNIME that vectorizes data from MongoDB, stores it in a vector database, and queries it with OpenAI embeddings.

Takeaways

  • No‑code RAG pipelines can be built in under 30 minutes with KNIME’s node editor.
  • Vectorizing text lets you perform semantic search instead of brittle keyword queries.
  • KNIME Secrets store API keys safely, enabling reusable authentication across workflows.
  • MongoDB data is ingested via connector → reader → JSON‑to‑Table nodes before vectorization.

Why use a RAG app and the no‑code approach

The presenter needed fast insights from 200+ student messages—sentiment, topics, interview discussions—without writing thousands of lines of code. A RAG pipeline lets you turn any text data (messages, PDFs, logs) into searchable semantic vectors, then query them with an LLM for natural‑language answers. Using KNIME’s drag‑and‑drop nodes avoids manual coding and speeds development to under 30 minutes.

Setting up the KNIME environment

Download KNIME from knime.com, create a free account, and install the application. After launching, open the Home tab, create a new workflow (named “vector store creator”), and install all extensions via *Install Extensions* to ensure required nodes are available.

Connecting to your data source

Add a MongoDB Connector node followed by a MongoDB Reader node. Configure host, username, password, and authentication database in the connector, then specify the database (e.g., “messages”) and collection in the reader. Execute to pull JSON records. Convert the JSON output to table format with a *JSON to Table* node, then filter out rows where the "clean content" field is empty using a Row Filter node.

Creating the vector store

Add an OpenAI Authenticator node and retrieve your API key securely via KNIME Secrets: create a secret named “OpenAI API key” in the profile’s *Secrets* tab, then use a Secrets Retriever node to pull it into a flow variable. Connect this variable to the authenticator’s input. Next, add an OpenAI AI Embedding Model Selector node, choose a lightweight model (e.g., "small"), and link it to the authenticator. These nodes will generate numeric embeddings for each message text.

Storing and querying vectors

(The transcript stops before completing this step, but the intended flow is to feed the cleaned table rows into an embedding node, write the resulting vectors to a vector database node, and later add a query node that sends a user prompt to OpenAI along with relevant retrieved vectors for answer generation.)

Running and iterating

Execute each node sequentially—connector, reader, conversion, filter, authenticator, model selector—to ensure data flows correctly. Use KNIME’s visual workflow to adjust filters or switch embedding models without rewriting code. The final app can answer questions like “What do students say about interviews?” by retrieving semantically similar messages and summarizing them via the LLM.