IVFFlat Indexing in pgvector

Vector databases and AI-powered applications continue to grow rapidly, and PostgreSQL has joined the movement with pgvector, a powerful extension that adds vector similarity search directly to Postgres. With the release of pgvector 0.5+, one of the most widely used indexing strategies is IVFFlat, an approximate nearest neighbor (ANN) index that dramatically speeds up similarity queries on large vector datasets.

What is IVFFlat in pgvector?

IVFFlat (Inverted File with Flat Vectors) is an Approximate Nearest Neighbor (ANN) index. Unlike a brute-force scan, which compares a query vector against every vector in the table, IVFFlat partitions vectors into multiple “lists” (or clusters). During a query, only the most relevant lists are searched.

Key Benefits

• Faster similarity search on large datasets
• Approximate, but accuracy is tunable
• Great fit for high-dimensional embeddings

How IVFFlat Works?

IVFFlat uses a centroid-based clustering approach:

1. Training Step:

• Vectors are clustered into lists using k-means.
• Each list represents a centroid.

2. Index Structure

• Each vector is assigned to the closest centroid/list.
• The index stores lists of vectors (inverted lists).

3. Query Execution

• Query vector is compared to all centroids.
• The probes most similar lists are selected.
• Only vectors within those lists are compared.

What do we control?

• lists - number of clusters
• probes - number of clusters searched during query Increasing the number of probes increases accuracy but reduces speed.

Implementing IVFFlat Indexing in pgvector

1. Install pgvector (if not already installed)

CREATE EXTENSION IF NOT EXISTS vector;

2. Create a table with vector embeddings

CREATE TABLE documents (
    id bigserial PRIMARY KEY,
    embedding vector(768)
);

3. Create IVFFlat Index

CREATE INDEX vector_ivfflat_idx
ON documents
USING ivfflat (embedding vector_cosine_ops)
WITH (lists = 1000);

Note:

• The index must be created after inserting enough rows (for better k-means training).
• Try to have at least 1,000 rows per list.

4. Querying with IVFFlat
Example cosine similarity search:

SET ivfflat.probes = 20;

SELECT id
FROM documents
ORDER BY embedding <-> '[0.5, 0.3, …]'
LIMIT 10;

# Set globally:
ALTER SYSTEM SET ivfflat.probes = 20;

Tuning probes in IVFFlat

• Probes control the number of IVF lists scanned during a query.
• Lower probes - faster but less accurate, because fewer clusters are searched.
• Higher probes - better accuracy, but more clusters scanned means slower performance.
• Choosing the optimal value depends on how much you prioritize speed vs recall.

Recommended Ranges

Low probes (1–10)

• ✔ Fastest search performance
• ✔ Best for real-time or high-throughput workloads
• ✘ Lower accuracy and recall
• ✘ Might miss similar vectors if clusters are coarse

Medium probes (~10% of total lists)

• ✔ Balanced between speed and accuracy
• ✔ Suitable for most production workloads
• ✔ Good recall without major performance sacrifice
• ✘ Slightly slower than low probe settings

High probes (50–100% of lists)

• ✔ Near-exact search results (high recall)
• ✔ Good for quality-sensitive workloads (e.g., search relevance)
• ✘ Much slower due to scanning many lists
• ✘ Reduces the performance benefit of ANN indexing

Maintenance Tasks: REINDEX, ANALYZE, VACUUM

IVFFlat indexes must be maintained correctly to keep search performance stable.

1. ANALYZE: Improve Query Planning
PostgreSQL needs fresh statistics to choose the best plan. Run ANALYZE after large batches of insertions or schedule autovacuum/analyze.

ANALYZE documents;

# Check the last ANALYZE time for your table
SELECT relname, last_analyze, last_autoanalyze
FROM pg_stat_all_tables
WHERE relname = 'documents';

2. REINDEX: Required After Massive Data Changes
If many vectors are inserted or deleted, list centroids can drift and degrade performance.

REINDEX INDEX vector_ivfflat_idx;

# If you want to keep the table live during rebuilds
REINDEX INDEX CONCURRENTLY vector_ivfflat_idx;

When to REINDEX:

• After inserting millions of new rows
• After deleting a large portion of data
• If recall noticeably decreases

3. VACUUM: Keep Storage Clean
Vector columns don’t produce unusual bloat, but regular VACUUM helps maintain table and index health.

VACUUM (VERBOSE, ANALYZE) documents;

Enable autovacuum for continuous maintenance.

IVFFlat is a powerful ANN indexing method available in pgvector, offering a balance of performance, memory efficiency, and simplicity.
With proper configuration and maintenance, IVFFlat can deliver high-performance vector search right inside PostgreSQL, no external database required.