Turn a folder of finance files into a simple chat assistant. New, changed, or deleted files are synced to a vector database so your team can ask questions and get clear answers fast. It fits teams that need quick insight from bank statements and other finance documents.

The system watches a local folder for adds, changes, and removals. When a file appears or changes, the flow reads the content, splits it into chunks, and creates embeddings with Mistral AI. Points are stored in Qdrant with helpful metadata like file name and month. If a file is removed, matched points are cleared through the Qdrant API. A chat endpoint then uses Retrieval QA with a Mistral chat model and the Qdrant retriever to answer questions from the latest files.

Setup needs a folder that n8n can access, a reachable Qdrant collection, and a Mistral API key. Expect faster close cycles and fewer manual lookups, with answers in minutes instead of hours. Useful for month end reviews, audits, and vendor checks where people need quick facts from many documents.

• Monitors a local folder for add, change, and delete events with the Local File Trigger
• Reads updated files and prepares clean text for indexing using the file reader and a default data loader
• Splits long documents into chunks with a recursive character text splitter for better recall
• Builds embeddings using Mistral AI and stores them in Qdrant with file based metadata
• Checks for existing points and removes outdated entries through Qdrant API calls
• Runs a chat endpoint that uses a Retrieval QA chain with a Mistral chat model
• Retrieves context from Qdrant via a vector store retriever for relevant answers
• Includes manual test and remap steps to validate paths and variables

• Reduce manual document lookup from hours to minutes
• Keep the knowledge base fresh by syncing every file change
• Improve answer accuracy by storing rich metadata with each point
• Support hundreds of files without slowing down search
• Connect AI and vector search without custom code

1. Import the template into n8n: Create a new workflow in n8n > Click the three dots menu > Select 'Import from File' > Choose the downloaded JSON file.
2. You will need accounts with Mistral AI and Qdrant. See the Tools Required section above for links to create accounts with these services.
3. In the n8n credentials manager, create a Mistral credential. Choose Mistral Cloud, add your API key from the Mistral account dashboard, give it a clear name, and save.
4. In the n8n credentials manager, create a Qdrant credential. Use the API key from your Qdrant service. Name it clearly and save. If you host Qdrant yourself, confirm the host and port match your environment.
5. Open the Set Variables node and confirm the directory path matches a folder that n8n can access. If using Docker, mount the host folder into the n8n container and use that path.
6. Double click the Local File Trigger and set the same folder path. Keep await write finish enabled so large files are not read before they are fully saved.
7. Check the Qdrant collection name in the Set Variables node. Make sure the Qdrant Vector Store nodes point to the same collection.
8. Open the Embeddings and Chat Model nodes and select the Mistral credential you created. Save each node.
9. Open the Qdrant Vector Store nodes and select your Qdrant credential. Confirm host, port, and collection are correct.
10. Click Execute on the Manual Trigger to validate variable mapping. Review the debug panel for any path or auth errors.
11. Add a sample PDF or CSV to the watched folder. Confirm that a point is created or updated in Qdrant by checking the Search For Existing Point node output.
12. Start the workflow, open the Chat Trigger test URL, and ask a question about content in your files. If no results appear, verify the file path metadata, collection name, and that Qdrant is reachable from n8n.