n8n Workflow: Index Candidate Records to Qdrant with Ollama

Index your candidate database into Qdrant using Ollama embeddings (n8n workflow)

This n8n workflow automatically reads a local JSON file of candidate records, generates semantic embeddings with Ollama (using nomic-embed-text), and stores everything in a Qdrant collection for fast, metadata-aware search.

What this workflow does

• Manually starts an indexing run via a Manual Trigger.
• Reads and parses a local JSON candidate database from disk (default path: /data/candidates.json).
• Builds embed-ready documents by converting each candidate record into a text block (the workflow creates a pageContent representation for embedding).
• Extracts and attaches metadata per candidate, including: candidate_id, title, employer, region, seniority, last_activity_date, and consent_withdrawn.
• Generates embeddings using the Ollama embedding model nomic-embed-text.
• Upserts into Qdrant by inserting documents, embeddings, and metadata into the candidate_database collection.
• Outputs a data-quality summary, specifically listing candidates missing last_activity_date so you can correct source data before downstream search excludes them.

Use cases

• Semantic search over a local ATS/candidate export using Qdrant vectors plus candidate metadata.
• Keeping an embeddings index in sync whenever your candidates.json file updates.
• QAing candidate data completeness (e.g., identifying missing last_activity_date) to prevent inaccurate downstream filtering.

Technical details

• n8n nodes used include: Code, Sticky Note, Manual Trigger, Read/Write File, and Extract from File.
• Embeddings: n8n-nodes-langchain Ollama embeddings (model: nomic-embed-text).
• Vector store: Qdrant (collection name: candidate_database).

Set your JSON path, ensure Ollama has nomic-embed-text pulled, configure Qdrant credentials, and run—this workflow will index your candidate records end-to-end.