QQL — Qdrant Query Language

SQL-like query language and CLI for Qdrant vector database.

Write INSERT, SELECT, SEARCH, SCROLL, RECOMMEND, UPDATE, DELETE, and CREATE COLLECTION statements instead of Python SDK calls. Supports hybrid dense+sparse vector search, grouped search (GROUP BY), cross-encoder reranking, quantization (scalar, turbo, binary, product), SQL-style WHERE filters, script execution, and collection dump/restore.

qql> INSERT INTO COLLECTION notes VALUES {'text': 'Qdrant is a vector database', 'author': 'alice', 'year': 2024}
✓ Inserted 1 point [3f2e1a4b-8c91-4d0e-b123-abc123def456]

qql> SEARCH notes SIMILAR TO 'vector storage engines' LIMIT 3 WHERE year >= 2023
✓ Found 1 result(s)
 Score  │ ID                                   │ Payload
────────┼──────────────────────────────────────┼──────────────────────────────────────
 0.8931 │ 3f2e1a4b-8c91-4d0e-b123-abc123def456 │ {'text': 'Qdrant is a ...', 'author': 'alice', 'year': 2024}

qql> SEARCH notes SIMILAR TO 'vector databases' LIMIT 5 USING HYBRID RERANK
✓ Found 1 result(s) (hybrid, reranked)
 Score  │ ID                                   │ Payload
────────┼──────────────────────────────────────┼──────────────────────────────────────
 5.3754 │ 3f2e1a4b-8c91-4d0e-b123-abc123def456 │ {'text': 'Qdrant is a ...', 'author': 'alice', 'year': 2024}

How It Works

QQL is a thin translation layer between a SQL-like query language and the Qdrant Python client. Every statement you type goes through three stages:

Your query string
      │
      ▼
  [ Lexer ]      — tokenizes the input into keywords, identifiers, literals
      │
      ▼
  [ Parser ]     — builds a typed AST node (e.g. InsertStmt, SearchStmt)
      │
      ▼
  [ Executor ]   — maps the AST node to a Qdrant client call
      │
      ▼
  Qdrant instance

When you run INSERT, the text field is automatically converted into a dense vector using Fastembed. In hybrid mode (USING HYBRID), a sparse BM25 vector is also generated alongside the dense vector, and searches use Qdrant's Reciprocal Rank Fusion (RRF) by default to merge the results of both retrieval methods. You can switch hybrid search to DBSF with FUSION 'dbsf'.

QQL also exposes a programmatic API for use inside Python applications — no CLI required:

from qql import Connection

with Connection("http://localhost:6333") as conn:
    conn.run_query("INSERT INTO COLLECTION notes VALUES {'text': 'Qdrant is fast'}")
    result = conn.run_query("SEARCH notes SIMILAR TO 'vector database' LIMIT 5")
    for hit in result.data:
        print(hit["score"], hit["payload"])

Installation

Requirements: Python 3.12+, a running Qdrant instance.

pip install qql-cli

Connect to a Qdrant instance:

# Local
qql connect --url http://localhost:6333

# Qdrant Cloud
qql connect --url https://<your-cluster>.qdrant.io --secret <your-api-key>

Then type qql to open the interactive shell.

Documentation

Full documentation lives in the docs/ folder and at pavanjava.github.io/qql:

Quick Syntax Reference

-- Insert
INSERT INTO COLLECTION articles VALUES {'text': '...', 'year': 2024}
INSERT BULK INTO COLLECTION articles VALUES [{'text': '...'}, {'text': '...'}]

-- Search
SEARCH articles SIMILAR TO 'query' LIMIT 10
SEARCH articles SIMILAR TO 'query' LIMIT 10 WHERE year >= 2020
SEARCH articles SIMILAR TO 'query' LIMIT 10 WHERE active = true
SEARCH articles SIMILAR TO 'query' LIMIT 10 WITH { mmr_diversity: 0.5, mmr_candidates: 50 }
SEARCH articles SIMILAR TO 'query' LIMIT 10 USING HYBRID
SEARCH articles SIMILAR TO 'query' LIMIT 10 USING HYBRID FUSION 'dbsf'
SEARCH articles SIMILAR TO 'query' LIMIT 10 WITH { indexed_only: true }
SEARCH articles SIMILAR TO 'query' LIMIT 10 WITH { quantization: { ignore: true, oversampling: 2 } }
SEARCH articles SIMILAR TO 'query' LIMIT 10 USING HYBRID RERANK

-- Scroll
SCROLL FROM articles LIMIT 50
SCROLL FROM articles WHERE year >= 2024 LIMIT 50
SCROLL FROM articles AFTER 'cursor-id' LIMIT 50

-- Recommend
RECOMMEND FROM articles POSITIVE IDS (1001, 1002) LIMIT 5

-- Select (retrieve a point by ID)
SELECT * FROM articles WHERE id = '3f2e1a4b-...'

-- Collections
CREATE COLLECTION articles
CREATE COLLECTION articles HYBRID
CREATE COLLECTION articles WITH HNSW { payload_m: 16 }
CREATE COLLECTION articles WITH VECTORS { on_disk: true } WITH HNSW { full_scan_threshold: 10000 }
ALTER COLLECTION articles WITH OPTIMIZERS { indexing_threshold: 10000 }
ALTER COLLECTION articles WITH PARAMS { read_fan_out_factor: 4, on_disk_payload: false }
ALTER COLLECTION articles QUANTIZE DISABLED
CREATE COLLECTION articles QUANTIZE SCALAR
CREATE COLLECTION articles QUANTIZE TURBO
CREATE COLLECTION articles QUANTIZE TURBO BITS 2
CREATE COLLECTION articles QUANTIZE TURBO BITS 1.5 ALWAYS RAM
CREATE INDEX ON COLLECTION articles FOR year TYPE integer
CREATE INDEX ON COLLECTION articles FOR tenant_id TYPE keyword WITH { is_tenant: true, on_disk: true }
CREATE INDEX ON COLLECTION articles FOR doc_id TYPE uuid
CREATE INDEX ON COLLECTION articles FOR title TYPE text WITH { tokenizer: 'word', min_token_len: 2, lowercase: true }
SHOW COLLECTIONS
SHOW COLLECTION articles
DROP COLLECTION articles

-- Search with grouping
SEARCH articles SIMILAR TO 'query' LIMIT 5 GROUP BY category
SEARCH articles SIMILAR TO 'query' LIMIT 5 GROUP BY category GROUP_SIZE 3
SEARCH articles SIMILAR TO 'query' LIMIT 5 WHERE year >= 2020 GROUP BY category GROUP_SIZE 2
SEARCH articles SIMILAR TO 'query' LIMIT 5 USING HYBRID GROUP BY category

-- Update
UPDATE articles SET VECTOR WHERE id = '3f2e1a4b-...' [0.1, 0.2, 0.3, 0.4]
UPDATE articles SET PAYLOAD WHERE id = '3f2e1a4b-...' {'year': 2025, 'status': 'active'}
UPDATE articles SET PAYLOAD WHERE category = 'draft' {'status': 'published'}

-- Delete
DELETE FROM articles WHERE id = '3f2e1a4b-...'
DELETE FROM articles WHERE year < 2020

-- Scripts
EXECUTE /path/to/script.qql
DUMP articles /path/to/backup.qql

Running Tests

Tests do not require a running Qdrant instance — the Qdrant client is mocked.

pytest tests/ -v

Expected: 549 tests passing.

License

MIT © Kameshwara Pavan Kumar Mantha