# Build a tax code assistant with Qdrant and Mistral: n8n workflow

> Started by someone opening a chat with it, and works across Qdrant. It puts OpenAI behind a retrieval step, so answers are drawn from documents you supply...

Source: https://millionrounds.com/workflow/build-a-tax-code-assistant-with-qdrant-and-mistral (Millionrounds; every figure on the page links to its filing or article)

[Automation library](https://millionrounds.com/workflows) · [Money admin](https://millionrounds.com/workflows?category=Money+admin#library)

# Build a tax code assistant with Qdrant and Mistral

Started by someone opening a chat with it, and works across Qdrant. It puts OpenAI behind a retrieval step, so answers are drawn from documents you supply rather than from the model alone. 31 steps, 1 branch, 2 loops or waits: large, and worth reading before you trust it with anything live.

Large Calls a model: OpenAI Retrieval

[Download free](https://millionrounds.com/login?next=%2Fapp%2Fworkflow%2Fbuild-a-tax-code-assistant-with-qdrant-and-mistral)

**31 steps** Large: read it before trusting it with anything live

**1** Branch: places the flow splits

**2** Loops or waits

**1** Service it connects to

## The flow

Starts it 3 Model 10 Logic 12 Service 6

The steps in the order they run, from what starts it.

- 1 **When clicking ‘Test workflow’** Manual trigger *Starts it*
- 2 **Embeddings Mistral Cloud** Embeddings Mistral Cloud *Model*
- 3 **Recursive Character Text Splitter** Text Splitter Recursive Character Text Splitter *Model*
- 4 **Execute Workflow Trigger** Execute Workflow trigger *Starts it*
- 5 **Window Buffer Memory** Memory Buffer Window *Model*
- 6 **Window Buffer Memory1** Memory Buffer Window *Model*
- 7 **OpenAI Chat Model** Lm Chat Open Ai *Model*
- 8 **Ask Tool** Tool Workflow *Model*
- 9 **Search Tool** Tool Workflow *Model*
- 10 **Get Tax Code Zip File** Http Request *Service*
- 11 **Qdrant Vector Store** Vector Store Qdrant *Model*
- 12 **Default Data Loader** Document Default Data Loader *Model*
- 13 **Switch** Switch *Logic*
- 14 **AI Agent** Agent *Model*
- 15 **When chat message received** Chat trigger *Starts it*
- 16 **Extract Zip Files** Compression *Service*
- 17 **1sec** Wait *Logic*
- 18 **Get Mistral Embeddings** Http Request *Service*
- 19 **Use Qdrant Scroll API** Http Request *Service*
- 20 **Files as Items** Split Out *Logic*
- 21 **For Each Section...** Split In Batches *Logic*
- 22 **Use Qdrant Search API1** Http Request *Service*
- 23 **Get Search Response** Set *Logic*
- 24 **Extract PDF Contents** Extract From File *Service*
- 25 **Content Chunking @ 50k Chars** Set *Logic*
- 26 **Get Ask Response** Set *Logic*
- 27 **Extract From Chapter** Set *Logic*
- 28 **Split Out Chunks** Split Out *Logic*
- 29 **Map To Sections** Set *Logic*
- 30 **Sections To List** Split Out *Logic*
- 31 **Only Valid Sections** Filter *Logic*

## How it's built

- **Starts on:** Someone opening a chat with it
- **Steps:** 31, with 1 branch, 2 loops or waits
- **Size:** Large: read it before trusting it with anything live
- **Connects to:** Qdrant
- **Model:** OpenAI
- **Retrieval:** Answers are drawn from documents you supply, not from the model alone

## Questions

What does it need to run?

An n8n account, plus your own Qdrant account, and a key for OpenAI. You connect your own accounts after importing it; none are included.

What starts it?

It starts on someone opening a chat with it.

Does it use AI?

Yes. It calls OpenAI. Its answers draw on documents you give it, not on the model alone. Model calls are billed by the provider, not by n8n.

How long does it take to set up?

Take your time. At 31 steps it is large: read it before trusting it with anything live.

How do I import it?

Copy the workflow, open a new workflow in n8n and paste. The steps appear on the canvas, ready for your own accounts.

**Before you run it**

Nothing here has been tested by us. It is somebody's working automation, read and described from its own steps. Treat the connections it wants as a list of accounts it will touch, and read the 31 steps before you point it at anything that matters.

## Run it in n8n

- **Create your n8n account** Free to start. n8n is where the workflow runs. [Create an n8n account ↗](https://millionrounds.com/n8n)
- **Download the workflow** Free with a Millionrounds account. [Download free](https://millionrounds.com/login?next=%2Fapp%2Fworkflow%2Fbuild-a-tax-code-assistant-with-qdrant-and-mistral)
- **Paste it into n8n** Open a new workflow in n8n and paste. Then connect your own accounts.

Stored with the author's traces removed: canvas notes, instance identifiers, saved credential names and any live document they had left behind.

## Connects to

Each opens the other workflows that use it.

[Qdrant](https://millionrounds.com/workflows?service=Qdrant#library)

## Workflows like it

Also money admin, with the same services where possible, closest in size first.

- [**PDF proposal knowledge base with OpenAI and Qdrant RAG** 14 steps · calls a model · Qdrant, S3](https://millionrounds.com/workflow/pdf-proposal-knowledge-base-with-openai-and-qdrant-rag)
- [**Generate AI-curated sales quotes with OpenAI and Qdrant** 12 steps · calls a model · Email, Qdrant](https://millionrounds.com/workflow/generate-ai-curated-sales-quotes-with-openai-and-qdrant)