VelesDB CLI

Interactive VelesQL REPL and CLI for VelesDB.

Table of Contents

• Installation
• Quick Start
• Getting Started
• REPL Mode
• CLI Subcommands
• VelesQL Examples (Vector, Graph, Hybrid, Sparse, Temporal, Aggregations, Joins, Set Operations)
• Session Settings
• Output Formats
• Shell Completions
• Common Errors
• License

Installation

From crates.io

cargo install velesdb-cli

From source

git clone https://github.com/cyberlife-coder/VelesDB
cd VelesDB
cargo install --path crates/velesdb-cli

Linux .deb package

A cargo-deb configuration is included. Build the package with:

cargo deb -p velesdb-cli
sudo dpkg -i target/debian/velesdb-cli_*.deb

The binary is installed as velesdb in /usr/bin/.

Quick Start

# Start the interactive REPL (default data directory: ./data)
velesdb repl

# Open a specific database
velesdb repl ./my_database

# Execute a one-shot query
velesdb query ./my_database "SELECT * FROM documents LIMIT 10"

# List all collections
velesdb list ./my_database

Getting Started

End-to-end example: create a collection, insert vectors, search, and query the graph.

# 1. Create a vector collection (384-dimensional, cosine similarity)
velesdb create-vector-collection ./data docs --dimension 384 --metric cosine

# 2. Upsert a few points
velesdb upsert ./data docs --id 1 --vector '[0.1, 0.2, ..., 0.384]' \
  --payload '{"title": "Intro to VelesDB", "category": "tech"}'
velesdb upsert ./data docs --id 2 --vector '[0.3, 0.4, ..., 0.384]' \
  --payload '{"title": "Graph Databases 101", "category": "tech"}'

# 3. Search by vector similarity
velesdb query ./data "SELECT * FROM docs WHERE vector NEAR [0.1, 0.2, ..., 0.384] LIMIT 5"

# 4. Create a graph collection and add edges
velesdb create-graph-collection ./data my_graph
velesdb graph add-edge ./data my_graph 1 100 200 "AUTHORED_BY"

# 5. Query the graph
velesdb graph neighbors ./data my_graph 100 --direction both

# 6. Enter the REPL for interactive exploration
velesdb repl ./data

REPL Mode

Start the REPL with velesdb repl [path]. The REPL accepts dot-commands (.help), backslash-commands (\set), and raw VelesQL queries.

The REPL is single-line only -- each command or query must fit on one line. Multi-line input is not supported.

History is persisted across sessions in ~/.local/share/.velesdb_history (Linux) or the equivalent platform data directory.

General Commands

Collection Inspection

Data Operations

Query Analysis

Index Management

Advanced Search

.sparse-search example:

velesdb> .sparse-search my_col sparse_idx [[42,0.8],[137,0.6],[891,0.3]] 10

.hybrid-sparse example:

Dense vector is a JSON array [0.1, 0.2, ...], sparse vector is [[index, weight], ...]:

velesdb> .hybrid-sparse docs [0.1,0.2,0.3,0.4] [[0,1.5],[3,0.8]] 10 --strategy rrf

Valid strategies: rrf, average, max (also accepts maximum).

Graph Commands (REPL)

All graph REPL commands operate on graph collections via .graph <subcommand>:

Naming note: The REPL uses .graph edges while the CLI subcommand uses graph get-edges. Both do the same thing.

Session Commands

Session commands use the backslash prefix (dot prefix also accepted):

Effect of \use <collection>: Sets the collection session setting to the named collection. The REPL verifies that the collection exists (vector, graph, or metadata) and displays its type. Currently, \use records the active collection in the session state (visible via \show), but VelesQL queries still require an explicit FROM <collection> clause. The active collection does not automatically apply to dot-commands or queries.

VelesQL Queries

Any input not starting with . or \ is executed as a VelesQL query through the full pipeline (parse → validate → execute). The REPL supports the complete VelesQL surface:

velesdb> SELECT * FROM docs WHERE category = 'tech' LIMIT 10;
velesdb> SELECT * FROM docs WHERE vector NEAR [0.1, 0.2, 0.3] LIMIT 5;
velesdb> SELECT * FROM docs WHERE similarity(vector, [0.1, 0.2]) > 0.8 LIMIT 10;
velesdb> SELECT * FROM docs WHERE content MATCH 'rust programming' LIMIT 10;
velesdb> CREATE COLLECTION test (dimension = 4, metric = 'cosine');
velesdb> DROP COLLECTION test;
velesdb> SELECT EDGES FROM kg WHERE label = 'KNOWS';
velesdb> MATCH (a:Person)-[:KNOWS]->(b) RETURN a, b LIMIT 10;
velesdb> SELECT category, COUNT(*) FROM docs GROUP BY category HAVING COUNT(*) > 5;
velesdb> TRAIN QUANTIZER ON docs WITH (m = 8, k = 256);

Supported VelesQL statements: SELECT, INSERT (with $params), UPSERT, UPDATE, DELETE, CREATE/DROP COLLECTION, CREATE/DROP INDEX, TRUNCATE, FLUSH, ANALYZE, SHOW COLLECTIONS, DESCRIBE COLLECTION, EXPLAIN, TRAIN QUANTIZER, INSERT EDGE, DELETE EDGE, INSERT NODE, SELECT EDGES, MATCH (graph patterns), UNION/INTERSECT/EXCEPT, JOIN.

Supported WHERE clauses: =, !=, >, <, >=, <=, IN, NOT IN, BETWEEN, IS NULL, IS NOT NULL, LIKE, ILIKE, NOT, AND, OR, NEAR (vector), NEAR_FUSED (multi-vector), SPARSE_NEAR, MATCH (BM25), similarity() threshold.

Supported modifiers: LIMIT, OFFSET, ORDER BY, GROUP BY, HAVING, DISTINCT, WITH (mode, ef_search, timeout_ms, rerank, quantization), USING FUSION (rrf, rsf, weighted, maximum).

Limitation: Bind parameters ($v, $query) are not supported in the REPL because there is no mechanism to pass external values. Use literal vectors [0.1, 0.2, ...] for search queries. For INSERT with vectors, use the velesdb upsert CLI command instead. Bind parameters work via the REST API (POST /query with params).

MATCH queries require an active collection set via \use <collection_name>. The REPL tries graph collections first, then vector collections.

CLI Subcommands

All subcommands operate offline against the database directory -- no running server required.

Collections

# Create a vector collection
velesdb create-vector-collection ./data my_vectors \
  --dimension 384 \
  --metric cosine \
  --storage full

# Create a graph collection
velesdb create-graph-collection ./data my_graph --schemaless true

# Create a metadata-only collection (no vectors, no graph -- structured payloads only)
velesdb create-metadata-collection ./data my_metadata

# List all collections
velesdb list ./data
velesdb list ./data --format json

# Show collection details
velesdb show ./data my_vectors
velesdb show ./data my_vectors --samples 5 --format json

# Delete a collection (interactive confirmation)
velesdb delete-collection ./data my_vectors
velesdb delete-collection ./data my_vectors --force

# Database overview
velesdb info ./data

create-vector-collection flags:

create-metadata-collection: Creates a collection that stores only structured JSON payloads -- no vectors, no graph edges. Useful for reference tables, configuration, or any metadata that does not need similarity search. There are no additional flags beyond the database path and collection name.

Points

# Upsert a point with vector and payload
velesdb upsert ./data my_vectors \
  --id 1 \
  --vector '[0.1, 0.2, 0.3]' \
  --payload '{"title": "Hello World"}'

# Upsert with payload only (no --vector flag)
# The vector defaults to an empty array. This will fail if the collection
# expects a specific dimension -- the collection validates vector dimensions.
velesdb upsert ./data my_vectors \
  --id 2 \
  --payload '{"title": "No vector"}'

# Get a point by ID (default output format: json)
velesdb get ./data my_vectors 42
velesdb get ./data my_vectors 42 --format table

# Delete points
velesdb delete-points ./data my_vectors 1 2 3

The upsert subcommand operates on vector collections only. The --vector flag is optional (an empty vector is used when omitted), but the collection will reject vectors whose dimension does not match the configured dimension -- this produces a dimension mismatch error. The --id flag is required.

Import / Export

# Import from JSONL
velesdb import data.jsonl \
  --database ./data \
  --collection documents \
  --dimension 768 \
  --metric cosine \
  --batch-size 1000

# Import from CSV (custom column names)
velesdb import embeddings.csv \
  --database ./data \
  --collection docs \
  --id-column doc_id \
  --vector-column embedding

# Export to JSON (vector collections only)
velesdb export ./data documents --output documents.json

# Export metadata only (no vectors)
velesdb export ./data documents --output meta.json --no-include-vectors

Note on --include-vectors: This is a Clap boolean flag that defaults to true. To disable it, use --no-include-vectors (Clap's automatic negation syntax). Writing --include-vectors false does not work as expected with Clap.

JSONL format for import:

Each line must be a valid JSON object with the following fields:

{"id": 1, "vector": [0.1, 0.2, 0.3, 0.4], "payload": {"title": "Doc A", "category": "tech"}}
{"id": 2, "vector": [0.5, 0.6, 0.7, 0.8], "payload": {"title": "Doc B"}}
{"id": 3, "vector": [0.9, 0.0, 0.1, 0.2]}

Lines with mismatched vector dimensions are counted as errors and skipped.

import flags:

export flags:

Export operates on vector collections only. Attempting to export a graph or metadata collection produces a "not found" error.

Search

# Execute a single VelesQL query
velesdb query ./data "SELECT * FROM documents LIMIT 10"
velesdb query ./data "SELECT * FROM docs WHERE category = 'tech' LIMIT 5" --format json

# Multi-query fusion search
velesdb multi-search ./data my_vectors \
  '[[0.1, 0.2, 0.3], [0.4, 0.5, 0.6]]' \
  -k 10 \
  --strategy rrf \
  --rrf-k 60

# Explain a query plan (default format: tree)
velesdb explain ./data "SELECT * FROM docs WHERE vector NEAR [0.1, 0.2] LIMIT 5"
velesdb explain ./data "SELECT * FROM docs LIMIT 10" --format json

multi-search flags:

multi-search strategies: average (also avg), maximum (also max), rrf (default), weighted

The weighted strategy uses fixed weights (avg_weight=0.5, max_weight=0.3, hit_weight=0.2) that are not configurable via the CLI.

explain formats:

The default format is tree, which renders a human-readable execution plan tree. Use --format json for machine-readable output.

Example tree output:

velesdb> .explain SELECT * FROM docs WHERE category = 'tech' LIMIT 10
Plan:
  Scan: docs
    Filter: category = 'tech'
    Limit: 10
  Estimated cost: 0.150 ms

Analyze

# Collection statistics (point count, deletion ratio, index stats, column stats)
velesdb analyze ./data my_vectors
velesdb analyze ./data my_vectors --format json

Graph

All graph CLI subcommands require a database path and graph collection name:

# Add an edge
velesdb graph add-edge ./data my_graph 1 100 200 "AUTHORED_BY"

# List edges
velesdb graph get-edges ./data my_graph
velesdb graph get-edges ./data my_graph --label "AUTHORED_BY" --format json

# Node degree
velesdb graph degree ./data my_graph 100

# Traverse (BFS/DFS)
velesdb graph traverse ./data my_graph 100 \
  --algorithm bfs \
  --max-depth 3 \
  --limit 50 \
  --rel-types "AUTHORED_BY,CITES"

# Get neighbors
velesdb graph neighbors ./data my_graph 100 --direction both --format json

# Store payload on a graph node
velesdb graph store-payload ./data my_graph 100 '{"name": "Alice", "role": "author"}'

# Retrieve node payload
velesdb graph get-payload ./data my_graph 100

# Remove an edge by ID
velesdb graph remove-edge ./data my_graph 1

# Count edges and nodes
velesdb graph count ./data my_graph
velesdb graph count ./data my_graph --format json

# Search graph nodes by embedding similarity (requires graph with embeddings)
velesdb graph search ./data my_graph '[0.1, 0.2, 0.3]' -k 10
velesdb graph search ./data my_graph '[0.1, 0.2, 0.3]' --format json

# List all nodes (paginated, 20 per page)
velesdb graph nodes ./data my_graph
velesdb graph nodes ./data my_graph --page 2
velesdb graph nodes ./data my_graph --format json

traverse flags:

Index Management

# Create a secondary index
velesdb index create ./data my_vectors category

# Create a property index
velesdb index create ./data my_vectors name --index-type property --label Person

# Create a range index
velesdb index create ./data my_vectors price --index-type range --label Product

# List indexes
velesdb index list ./data my_vectors
velesdb index list ./data my_vectors --format json

# Drop an index
velesdb index drop ./data my_vectors Person name

SIMD Diagnostics

# Show SIMD dispatch configuration
velesdb simd info

# Run SIMD benchmarks (redirects to cargo bench)
velesdb simd benchmark

License Management

# Show current license status
velesdb license show

# Activate a license
velesdb license activate <license_key>

# Verify a license (without activating)
velesdb license verify <license_key> --public-key <base64_public_key>

Both license show and license activate read the public key from the VELESDB_LICENSE_PUBLIC_KEY environment variable. If the variable is not set, a development fallback key is used with a warning. Set it with:

export VELESDB_LICENSE_PUBLIC_KEY=<base64_encoded_public_key>

Shell Completions

# Generate completions for your shell
velesdb completions bash       > /etc/bash_completion.d/velesdb
velesdb completions zsh        > ~/.zfunc/_velesdb
velesdb completions fish       > ~/.config/fish/completions/velesdb.fish
velesdb completions powershell > velesdb.ps1
velesdb completions elvish     > velesdb.elv

VelesQL Examples

VelesQL is a SQL-like query language with vector and graph extensions. These examples work directly in the REPL.

Vector Search

-- Basic vector similarity search
SELECT * FROM documents
WHERE vector NEAR [0.15, 0.25, 0.35, 0.45]
LIMIT 10;

-- Vector search with metadata filter
SELECT * FROM documents
WHERE vector NEAR [0.1, 0.2, 0.3, 0.4]
AND category = 'tech'
AND views > 100
LIMIT 5;

The distance metric is defined at collection creation time and applies to all searches on that collection. All five metrics are supported: Cosine, Euclidean, DotProduct, Hamming, Jaccard.

Multi-Vector Fusion (NEAR_FUSED)

-- RRF fusion with multiple query vectors
SELECT * FROM documents
WHERE vector NEAR_FUSED [$v1, $v2, $v3] USING FUSION 'rrf' (k = 60)
LIMIT 10;

-- Weighted fusion
SELECT * FROM documents
WHERE vector NEAR_FUSED [$query1, $query2] USING FUSION 'weighted'
LIMIT 10;

-- Maximum fusion
SELECT * FROM documents
WHERE vector NEAR_FUSED [$v1, $v2] USING FUSION 'maximum'
LIMIT 10;

-- Default (RRF) -- USING FUSION clause is optional
SELECT * FROM documents
WHERE vector NEAR_FUSED [$v1, $v2]
LIMIT 10;

Hybrid Search (USING FUSION)

The USING FUSION clause at query level combines results from multiple search strategies (vector + text, dense + sparse):

-- Dense vector + BM25 full-text combined with RRF
SELECT * FROM documents
WHERE vector NEAR $v AND content MATCH 'rust programming'
USING FUSION(strategy = 'rrf', k = 60)
LIMIT 10;

-- Dense + sparse vector fusion with RSF
SELECT * FROM documents
WHERE vector NEAR $dense AND vector SPARSE_NEAR $sparse
USING FUSION(strategy = 'rsf', dense_w = 0.7, sparse_w = 0.3)
LIMIT 10;

-- Weighted fusion
SELECT * FROM docs
WHERE vector NEAR $v
USING FUSION(strategy = 'weighted', vector_weight = 0.7, graph_weight = 0.3)
LIMIT 10;

-- Maximum fusion (take best score)
SELECT * FROM docs
WHERE vector NEAR $v
USING FUSION(strategy = 'maximum')
LIMIT 10;

-- Default USING FUSION (defaults to RRF)
SELECT * FROM docs
WHERE vector NEAR $v
USING FUSION
LIMIT 10;

Graph MATCH Queries

-- Find authors of documents similar to a query
MATCH (doc:Document)-[:AUTHORED_BY]->(author:Person)
WHERE similarity(doc.embedding, $question) > 0.8
RETURN author.name, doc.title
ORDER BY similarity() DESC
LIMIT 5;

-- Multi-hop traversal with depth range
MATCH (user:User)-[:FOLLOWS*1..3]->(target:User)
WHERE user.name = 'Alice'
RETURN target.name, target.bio
LIMIT 20;

-- Undirected relationship (both directions)
MATCH (a:Person)-[:KNOWS]-(b:Person)
WHERE a.city = 'Paris'
RETURN a.name, b.name
LIMIT 10;

-- Incoming relationships
MATCH (doc:Document)<-[:AUTHORED_BY]-(author:Person)
RETURN doc.title, author.name
LIMIT 10;

-- Node property filtering in pattern
MATCH (doc:Document {status: 'published'})-[:HAS_TAG]->(tag:Tag)
RETURN doc.title, tag.name
LIMIT 20;

-- Combined: graph + vector + metadata in WHERE
SELECT * FROM articles
WHERE category = 'tech' AND MATCH (d:Doc)-[:HAS_TAG]->(tag)
LIMIT 10;

Cross-collection MATCH: Use \use <collection> to set the primary collection (the one with graph edges) before running a MATCH query. Nodes annotated with @collection will have their payloads enriched from the named collection after traversal. Example:

velesdb> \use catalog_graph
velesdb> MATCH (p:Product)-[:STORED_IN]->(inv:Inventory@inventory) RETURN p.name, inv.price LIMIT 20;

Metadata-Only Collections

Metadata collections store structured data without vectors. They support full VelesQL SELECT queries:

-- Browse all items in a metadata collection
SELECT * FROM my_metadata LIMIT 10;

-- Filter by field
SELECT * FROM my_metadata WHERE status = 'active' LIMIT 20;

-- Count items
SELECT * FROM my_metadata WHERE price > 100 LIMIT 100;

Tip: Use .sample my_metadata, .browse my_metadata, and .export my_metadata in the REPL — all three commands work for Metadata collections (no "vector" column is shown).

Metadata Filters

-- Equality, comparison, pattern matching
SELECT * FROM docs WHERE category = 'tech' LIMIT 10;
SELECT * FROM docs WHERE price >= 50 AND price <= 200 LIMIT 10;
SELECT * FROM docs WHERE title LIKE '%rust%' LIMIT 10;

-- Case-insensitive pattern matching
SELECT * FROM docs WHERE title ILIKE '%Rust%' LIMIT 10;

-- IN, BETWEEN, NULL checks
SELECT * FROM docs WHERE status IN ('published', 'featured') LIMIT 10;
SELECT * FROM docs WHERE score BETWEEN 0.5 AND 1.0 LIMIT 10;
SELECT * FROM docs WHERE author IS NOT NULL LIMIT 10;

-- NOT and OR operators
SELECT * FROM docs WHERE NOT category = 'draft' LIMIT 10;
SELECT * FROM docs WHERE category = 'tech' OR category = 'science' LIMIT 10;

-- Nested field access (dot notation)
SELECT metadata.source, profile.type FROM docs WHERE metadata.lang = 'en' LIMIT 10;

-- Full-text search (BM25)
SELECT * FROM docs WHERE content MATCH 'rust programming' LIMIT 10;

Similarity Threshold

-- Return all documents above a similarity threshold (not just top-K)
SELECT * FROM docs
WHERE similarity(vector, $query) > 0.8
LIMIT 20;

-- Combine similarity threshold with metadata filters
SELECT * FROM docs
WHERE similarity(embedding, $ref) >= 0.9 AND category = 'tech'
LIMIT 10;

Sparse Vector Search

-- Sparse vector similarity search
SELECT * FROM docs
WHERE vector SPARSE_NEAR $sparse_vector
LIMIT 10;

-- Sparse search with inline literal
SELECT * FROM docs
WHERE vector SPARSE_NEAR {12: 0.8, 45: 0.3, 891: 0.1}
LIMIT 10;

-- Sparse search on a named index
SELECT * FROM docs
WHERE vector SPARSE_NEAR $sv USING 'my_sparse_index'
LIMIT 10;

Temporal Queries

-- Filter by current time
SELECT * FROM events WHERE timestamp > NOW() LIMIT 10;

-- Last 7 days
SELECT * FROM logs WHERE created_at > NOW() - INTERVAL '7 days' LIMIT 50;

-- Last hour
SELECT * FROM alerts WHERE fired_at > NOW() - INTERVAL '1 hour' LIMIT 20;

-- Next week (scheduling)
SELECT * FROM tasks WHERE due_date < NOW() + INTERVAL '7 days' LIMIT 20;

-- Shorthand units: s, m, h, d, w, month
SELECT * FROM metrics WHERE ts > NOW() - INTERVAL '30 min' LIMIT 100;

DISTINCT

-- Deduplicate results
SELECT DISTINCT category FROM documents LIMIT 50;

SELECT DISTINCT status, priority FROM tasks LIMIT 20;

Aggregations

SELECT category, COUNT(*) as cnt
FROM documents
GROUP BY category
HAVING cnt > 5
ORDER BY cnt DESC
LIMIT 10;

-- Multiple aggregates
SELECT category, COUNT(*) as cnt, AVG(price) as avg_price, MIN(price), MAX(price)
FROM products
GROUP BY category
LIMIT 20;

-- SUM aggregate
SELECT region, SUM(quantity) as total
FROM orders
GROUP BY region
ORDER BY total DESC
LIMIT 10;

-- GROUP BY nested fields
SELECT metadata.source, COUNT(*) as cnt
FROM documents
GROUP BY metadata.source
LIMIT 20;

ORDER BY

-- Multiple sort keys
SELECT * FROM docs ORDER BY category ASC, created_at DESC LIMIT 20;

-- Order by similarity score
SELECT * FROM docs
WHERE vector NEAR $v
ORDER BY similarity(vector, $v) DESC
LIMIT 10;

OFFSET (Pagination)

-- Skip first 20 results, return next 10
SELECT * FROM docs LIMIT 10 OFFSET 20;

Set Operations

-- UNION: combine results from two queries
SELECT id, title FROM news WHERE category = 'tech'
UNION
SELECT id, title FROM blog WHERE category = 'tech';

-- UNION ALL: include duplicates
SELECT id FROM collection_a
UNION ALL
SELECT id FROM collection_b;

-- INTERSECT: rows in both queries
SELECT id FROM favorites INTERSECT SELECT id FROM published;

-- EXCEPT: rows in first but not second
SELECT id FROM all_items EXCEPT SELECT id FROM archived;

JOIN

-- INNER JOIN (default)
SELECT o.id, c.name
FROM orders AS o
JOIN customers AS c ON o.customer_id = c.id
LIMIT 20;

-- LEFT JOIN
SELECT d.title, a.name
FROM documents AS d
LEFT JOIN authors AS a ON d.author_id = a.id
LIMIT 20;

-- JOIN with vector search
SELECT o.id, c.name
FROM orders AS o
JOIN customers AS c ON o.customer_id = c.id
WHERE similarity(o.embedding, $q) > 0.7
LIMIT 20;

Note: LEFT JOIN and RIGHT JOIN are parsed but raise runtime errors. INNER JOIN is fully supported.

Subqueries (parsed, not yet executable)

Note: Subqueries are recognized by the VelesQL parser but raise runtime errors during execution. This syntax is reserved for future support.

-- IN subquery (parsed, execution not yet supported)
SELECT * FROM docs WHERE id IN (SELECT doc_id FROM comments) LIMIT 10;

-- Scalar subquery comparison (parsed, execution not yet supported)
SELECT * FROM products WHERE price > (SELECT AVG(price) FROM products) LIMIT 20;

EXPLAIN

-- Show the query execution plan
EXPLAIN SELECT * FROM documents
WHERE vector NEAR [0.1, 0.2, 0.3, 0.4]
AND category = 'tech'
LIMIT 10;

TRAIN QUANTIZER

-- Train a product quantizer on a collection
TRAIN QUANTIZER ON documents WITH (m = 8, k = 256);

-- With oversampling and force retrain
TRAIN QUANTIZER ON large_docs WITH (m = 16, k = 256, oversampling = 4, force = true);

WITH Clause (Per-Query Options)

SELECT * FROM docs WHERE vector NEAR $v LIMIT 10
WITH (mode = 'accurate');

SELECT * FROM docs WHERE vector NEAR $v LIMIT 10
WITH (ef_search = 512, timeout_ms = 5000, rerank = true);

-- Quantization hints for dual-precision search
SELECT * FROM docs WHERE vector NEAR $v LIMIT 10
WITH (quantization = 'dual', oversampling = 4);

Escaped Identifiers

Use backticks or double-quotes to use reserved words as column names:

-- Backtick style
SELECT `select`, `from`, `order` FROM docs LIMIT 10;

-- Double-quote style (SQL standard)
SELECT "select", "from", "order" FROM docs LIMIT 10;

Session Settings

Session settings control REPL search behavior. Set with \set, view with \show, reset with \reset.

Examples:

velesdb> \set mode accurate
velesdb> \set ef_search 512
velesdb> \set timeout 5000
velesdb> \set rerank no
velesdb> \show
velesdb> \show mode
velesdb> \reset ef_search
velesdb> \reset
velesdb> \use documents

Output Formats

Two output formats are available, controlled by .format in the REPL or --format on CLI subcommands:

• table (default) -- UTF-8 formatted table with colored headers. The id column is always first.
• json -- Pretty-printed JSON array.

velesdb> .format json
velesdb> SELECT * FROM documents LIMIT 3;

velesdb query ./data "SELECT * FROM docs LIMIT 3" --format json

Shell Completions

Generate tab-completion scripts for your shell:

# Bash
velesdb completions bash > /etc/bash_completion.d/velesdb

# Zsh (add ~/.zfunc to fpath in .zshrc)
velesdb completions zsh > ~/.zfunc/_velesdb

# Fish
velesdb completions fish > ~/.config/fish/completions/velesdb.fish

# PowerShell
velesdb completions powershell | Out-String | Invoke-Expression

# Elvish
velesdb completions elvish > ~/.config/elvish/lib/velesdb.elv

Common Errors

License

MIT License

See LICENSE for details.