├── .gitignore ├── tests ├── conftest.py ├── test_query_assistant.py └── cassettes │ └── test_query_assistant │ ├── test_database_assistant_page.yaml │ └── test_table_assistant_page.yaml ├── .github └── workflows │ ├── test.yml │ └── publish.yml ├── pyproject.toml ├── README.md ├── datasette_query_assistant ├── templates │ └── query_assistant.html └── __init__.py └── LICENSE /.gitignore: -------------------------------------------------------------------------------- 1 | .venv 2 | __pycache__/ 3 | *.py[cod] 4 | *$py.class 5 | venv 6 | .eggs 7 | .pytest_cache 8 | *.egg-info 9 | .DS_Store 10 | .vscode 11 | dist 12 | build 13 | -------------------------------------------------------------------------------- /tests/conftest.py: -------------------------------------------------------------------------------- 1 | import os 2 | import pytest 3 | 4 | 5 | @pytest.fixture(autouse=True) 6 | def patch_env(monkeypatch): 7 | if not os.environ.get("OPENAI_API_KEY"): 8 | monkeypatch.setenv("OPENAI_API_KEY", "mock-key") 9 | 10 | 11 | @pytest.fixture(scope="module") 12 | def vcr_config(): 13 | return {"filter_headers": ["authorization"]} 14 | -------------------------------------------------------------------------------- /.github/workflows/test.yml: -------------------------------------------------------------------------------- 1 | name: Test 2 | 3 | on: [push, pull_request] 4 | 5 | permissions: 6 | contents: read 7 | 8 | jobs: 9 | test: 10 | runs-on: ubuntu-latest 11 | strategy: 12 | matrix: 13 | python-version: ["3.10", "3.11", "3.12", "3.13", "3.14"] 14 | steps: 15 | - uses: actions/checkout@v5 16 | - name: Set up Python ${{ matrix.python-version }} 17 | uses: actions/setup-python@v6 18 | with: 19 | python-version: ${{ matrix.python-version }} 20 | cache: pip 21 | cache-dependency-path: pyproject.toml 22 | - name: Install dependencies 23 | run: | 24 | pip install '.[test]' 25 | - name: Run tests 26 | run: | 27 | python -m pytest 28 | 29 | -------------------------------------------------------------------------------- /pyproject.toml: -------------------------------------------------------------------------------- 1 | [project] 2 | name = "datasette-query-assistant" 3 | version = "0.1a5" 4 | description = "Query databases and tables with AI assistance" 5 | readme = "README.md" 6 | authors = [{name = "Simon Willison"}] 7 | license = "Apache-2.0" 8 | classifiers=[ 9 | "Framework :: Datasette" 10 | ] 11 | requires-python = ">=3.10" 12 | dependencies = [ 13 | "datasette>=1.0a21", 14 | "llm>=0.22", 15 | "markdown2", 16 | ] 17 | 18 | [project.urls] 19 | Homepage = "https://github.com/datasette/datasette-query-assistant" 20 | Changelog = "https://github.com/datasette/datasette-query-assistant/releases" 21 | Issues = "https://github.com/datasette/datasette-query-assistant/issues" 22 | CI = "https://github.com/datasette/datasette-query-assistant/actions" 23 | 24 | [project.entry-points.datasette] 25 | query_assistant = "datasette_query_assistant" 26 | 27 | [project.optional-dependencies] 28 | test = ["llm-openai-plugin", "inline-snapshot", "pytest", "pytest-asyncio", "pytest-recording", "sqlite-utils"] 29 | 30 | [tool.pytest.ini_options] 31 | asyncio_mode = "strict" 32 | 33 | [tool.setuptools.package-data] 34 | datasette_query_assistant = ["templates/*"] 35 | -------------------------------------------------------------------------------- /.github/workflows/publish.yml: -------------------------------------------------------------------------------- 1 | name: Publish Python Package 2 | 3 | on: 4 | release: 5 | types: [created] 6 | 7 | permissions: 8 | contents: read 9 | 10 | jobs: 11 | test: 12 | runs-on: ubuntu-latest 13 | strategy: 14 | matrix: 15 | python-version: ["3.10", "3.11", "3.12", "3.13", "3.14"] 16 | steps: 17 | - uses: actions/checkout@v5 18 | - name: Set up Python ${{ matrix.python-version }} 19 | uses: actions/setup-python@v6 20 | with: 21 | python-version: ${{ matrix.python-version }} 22 | cache: pip 23 | cache-dependency-path: pyproject.toml 24 | - name: Install dependencies 25 | run: | 26 | pip install '.[test]' 27 | - name: Run tests 28 | run: | 29 | python -m pytest 30 | deploy: 31 | runs-on: ubuntu-latest 32 | needs: [test] 33 | environment: release 34 | permissions: 35 | id-token: write 36 | steps: 37 | - uses: actions/checkout@v5 38 | - name: Set up Python 39 | uses: actions/setup-python@v6 40 | with: 41 | python-version: "3.12" 42 | cache: pip 43 | cache-dependency-path: pyproject.toml 44 | - name: Install dependencies 45 | run: | 46 | pip install setuptools wheel build 47 | - name: Build 48 | run: | 49 | python -m build 50 | - name: Publish 51 | uses: pypa/gh-action-pypi-publish@release/v1 52 | 53 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # datasette-query-assistant 2 | 3 | [](https://pypi.org/project/datasette-query-assistant/) 4 | [](https://github.com/datasette/datasette-query-assistant/releases) 5 | [](https://github.com/datasette/datasette-query-assistant/actions/workflows/test.yml) 6 | [](https://github.com/datasette/datasette-query-assistant/blob/main/LICENSE) 7 | 8 | Query databases and tables with AI assistance 9 | 10 | **Early alpha**. 11 | 12 | ## Installation 13 | 14 | Install this plugin in the same environment as Datasette. 15 | ```bash 16 | datasette install datasette-query-assistant 17 | ``` 18 | 19 | ## Configuration 20 | 21 | Set the model ID to be used by this plugin by adding this to your Datasette configuration: 22 | 23 | ```yaml 24 | plugins: 25 | datasette-query-assistant: 26 | model: openai/gpt-4.1-mini 27 | key_env_var: OPENAI_API_KEY 28 | ``` 29 | To use the `openai/gpt-4.1-mini` model you'll need to install `llm-openai-plugin` in addition to `datasette-query-assistant`. 30 | 31 | If the model needs an API key, set the `key_env_var` to the name of an environment variable and ensure that environment variable is correctly set. 32 | 33 | You can use any model that is available via an [LLM plugin](https://llm.datasette.io/en/stable/plugins/directory.html) 34 | 35 | ## Usage 36 | 37 | Users with `execute-sql` permission will gain a database action menu item for "Query this database with AI assistance" which will let them ask a question and be redirected to a commented SQL query that will hopefully answer it. 38 | 39 | ## Development 40 | 41 | To set up this plugin locally, first checkout the code. Then create a new virtual environment: 42 | ```bash 43 | cd datasette-query-assistant 44 | python3 -m venv venv 45 | source venv/bin/activate 46 | ``` 47 | Now install the dependencies and test dependencies: 48 | ```bash 49 | pip install -e '.[test]' 50 | ``` 51 | To run the tests: 52 | ```bash 53 | pytest 54 | ``` 55 | To re-generate the tests with refreshed examples from the API: 56 | ```bash 57 | pytest -x --record-mode=rewrite --inline-snapshot=fix 58 | ``` 59 | -------------------------------------------------------------------------------- /datasette_query_assistant/templates/query_assistant.html: -------------------------------------------------------------------------------- 1 | {% extends "base.html" %} 2 | 3 | {% block title %}Query assistant for {% if table %}{{ table }}{% else %}{{ database }}{% endif %}{% endblock %} 4 | 5 | {% block crumbs %} 6 | {{ crumbs.nav(request=request, database=database, table=table) }} 7 | {% endblock %} 8 | 9 | {% block content %} 10 | 11 |
{{ schema }}
31 | CREATE TABLE foo (id integer primary key, name text)" 49 | in response.text 50 | ) 51 | # Submit the form 52 | csrftoken = response.cookies["ds_csrftoken"] 53 | post_response = await datasette.client.post( 54 | "/test/-/assistant", 55 | cookies={ 56 | "ds_csrftoken": csrftoken, 57 | }, 58 | data={ 59 | "question": "Show me all the data in the foo table", 60 | "csrftoken": csrftoken, 61 | }, 62 | ) 63 | assert post_response.status_code == 302 64 | qs = dict(urllib.parse.parse_qsl(post_response.headers["location"].split("?")[1])) 65 | assert qs["sql"] == snapshot( 66 | """\ 67 | -- SQL query to select all columns and all rows from the table 'foo' 68 | SELECT * FROM foo;\ 69 | """ 70 | ) 71 | 72 | 73 | @pytest.mark.asyncio 74 | async def test_table_assistant_page(datasette): 75 | response = await datasette.client.get("/test/-/assistant?table=foo") 76 | assert response.status_code == 200 77 | assert "Query assistant for foo" in response.text 78 | assert ( 79 | "
CREATE TABLE foo (id integer primary key, name text)" 80 | in response.text 81 | ) 82 | # Submit the form 83 | csrftoken = response.cookies["ds_csrftoken"] 84 | post_response = await datasette.client.post( 85 | "/test/-/assistant", 86 | cookies={ 87 | "ds_csrftoken": csrftoken, 88 | }, 89 | data={ 90 | "question": "Count of rows in foo", 91 | "csrftoken": csrftoken, 92 | }, 93 | ) 94 | assert post_response.status_code == 302 95 | qs = dict(urllib.parse.parse_qsl(post_response.headers["location"].split("?")[1])) 96 | assert qs["sql"] == snapshot( 97 | """\ 98 | -- Count the total number of rows in the table 'foo' 99 | SELECT COUNT(*) 100 | FROM foo;\ 101 | """ 102 | ) 103 | -------------------------------------------------------------------------------- /tests/cassettes/test_query_assistant/test_database_assistant_page.yaml: -------------------------------------------------------------------------------- 1 | interactions: 2 | - request: 3 | body: '{"input": [{"role": "system", "content": "You answer questions by generating 4 | SQL queries using SQLite schema syntax.\nAlways start with -- SQL comments explaining 5 | what you are about to do.\nNo yapping. Output SQL with extensive SQL comments 6 | in a sql tagged\nfenced markdown code block.\n\nReturn only one SQL SELECT query. 7 | Follow the query with an explanation\nof what the query does and how it works, 8 | which should include bold for\nemphasis where appropriate.\n\nExample question:\n\nHow 9 | many rows in the sqlite_master table?\n\nExample output (shown between ----):\n----\n```sql\nselect 10 | count(*) from sqlite_master\n```\nCount the **number of rows** in the `sqlite_master` 11 | table.\n----\nThe table schema is:\nCREATE TABLE foo (id integer primary key, 12 | name text)"}, {"role": "user", "content": "Show me all the data in the foo table"}], 13 | "model": "gpt-4.1-mini", "stream": false}' 14 | headers: 15 | accept: 16 | - application/json 17 | accept-encoding: 18 | - gzip, deflate 19 | connection: 20 | - keep-alive 21 | content-length: 22 | - '879' 23 | content-type: 24 | - application/json 25 | host: 26 | - api.openai.com 27 | user-agent: 28 | - AsyncOpenAI/Python 1.76.2 29 | x-stainless-arch: 30 | - arm64 31 | x-stainless-async: 32 | - async:asyncio 33 | x-stainless-lang: 34 | - python 35 | x-stainless-os: 36 | - MacOS 37 | x-stainless-package-version: 38 | - 1.76.2 39 | x-stainless-read-timeout: 40 | - '600' 41 | x-stainless-retry-count: 42 | - '0' 43 | x-stainless-runtime: 44 | - CPython 45 | x-stainless-runtime-version: 46 | - 3.12.0 47 | method: POST 48 | uri: https://api.openai.com/v1/responses 49 | response: 50 | body: 51 | string: !!binary | 52 | H4sIAAAAAAAAA3RUTW/bMAy951cIunQLliJ27XztOHSnDsPW3prBlmU60SaLrkSlDYr898Gy4yRF 53 | ezEsPvKJfCT1OmKMq5KvGLfgmmy2iG5kms6jRTpdRMu4iIWsimU5lUV5E5dyKorlNJkv42ohkqS6 54 | 4V9aAiz+gqQjCRoHnV1aEARlJlosmiezKJ5Fs3nAHAnyro2RWDcaCMouCKxFy1fMeK2DQZmjR1YC 55 | CaXdJerIekkKzYW9Fi8Zemo8ZYT/4A2IJej27k1Dk+Q6mtTKqEk8jdPJNJlESV9WCOcr9jhijLHX 56 | 8B30qt1mkEsmxaKVa5nMFiKN5+ksSpfLJH5XrsBB+wYCCzgnNnACPtIlgBINgTmldJ7WBe2xcnih 57 | ITo4CGOQxFGtxz8XYHBfMZ7nuXvSazOZsPtfd+zJg90zQuZAgyQmtGYSta+NY8KU4Wzx2bHKYs1o 58 | C4xEoYFdVYhXa3N/e3f77YGN2fffP3+wCvHr2uR5vjYPW+V6cgtkFezAsfG4pWtJSkFiPGaO0ELJ 59 | lAnGvELMe/5i32ekzIbBruXp0mKfvGtt+Tj/zCq0PWjx+UgTCK75UPyh/xv04BZ1UFI4pxwJQ51z 60 | 6xiceCOs0Bp0Rog6k0KHuSTroYMt7BR6lx1XIgtjM4ygBeHQKLPhq76FHKoKLZ05tePg61rYfW8c 61 | MXbotgfsTknISEG7K7yESnjd9ZoHxc5zIagbsIJ8MEfX094aut1fXqGtxel8NkvBbyi+u7+reYtK 62 | diJ5Qj4Ap8nihE3WnN9pvZFh/ELWyrV96Bffh0UYElLmYnf7d+MNcPYiDHlLIbdQniKnXe599Ns3 63 | IV28B7zHO3TsI2pCEvoExnE66OUdXLxpNZBo57vlP4wO/wEAAP//AwDxf5SbhwUAAA== 64 | headers: 65 | CF-RAY: 66 | - 939189000ab27d23-LAX 67 | Connection: 68 | - keep-alive 69 | Content-Encoding: 70 | - gzip 71 | Content-Type: 72 | - application/json 73 | Date: 74 | - Thu, 01 May 2025 19:02:48 GMT 75 | Server: 76 | - cloudflare 77 | Set-Cookie: 78 | - __cf_bm=A36mulovkt818TSk94rwODLlLq0hFHW19akmZLwqUc0-1746126168-1.0.1.1-7rsETv8YOrhJVMZHG7IHyaC8jaEbY.6KFR8bhgkTDPOcZ7NqTkBBFZeinIWrN85B3kTI9HU6S4FdFOVay0_sKjXGP907SgZACZUxRlILDAo; 79 | path=/; expires=Thu, 01-May-25 19:32:48 GMT; domain=.api.openai.com; HttpOnly; 80 | Secure; SameSite=None 81 | - _cfuvid=sZmpoXnD5vU7mZaYpIcqJwzoWujUPogspVn8ZctaUVY-1746126168965-0.0.1.1-604800000; 82 | path=/; domain=.api.openai.com; HttpOnly; Secure; SameSite=None 83 | Transfer-Encoding: 84 | - chunked 85 | X-Content-Type-Options: 86 | - nosniff 87 | alt-svc: 88 | - h3=":443"; ma=86400 89 | cf-cache-status: 90 | - DYNAMIC 91 | openai-organization: 92 | - user-r3e61fpak04cbaokp5buoae4 93 | openai-processing-ms: 94 | - '1874' 95 | openai-version: 96 | - '2020-10-01' 97 | strict-transport-security: 98 | - max-age=31536000; includeSubDomains; preload 99 | x-request-id: 100 | - req_59dde03c1f28080593ba14b98013ccdd 101 | status: 102 | code: 200 103 | message: OK 104 | version: 1 105 | -------------------------------------------------------------------------------- /tests/cassettes/test_query_assistant/test_table_assistant_page.yaml: -------------------------------------------------------------------------------- 1 | interactions: 2 | - request: 3 | body: '{"input": [{"role": "system", "content": "You answer questions by generating 4 | SQL queries using SQLite schema syntax.\nAlways start with -- SQL comments explaining 5 | what you are about to do.\nNo yapping. Output SQL with extensive SQL comments 6 | in a sql tagged\nfenced markdown code block.\n\nReturn only one SQL SELECT query. 7 | Follow the query with an explanation\nof what the query does and how it works, 8 | which should include bold for\nemphasis where appropriate.\n\nExample question:\n\nHow 9 | many rows in the sqlite_master table?\n\nExample output (shown between ----):\n----\n```sql\nselect 10 | count(*) from sqlite_master\n```\nCount the **number of rows** in the `sqlite_master` 11 | table.\n----\nThe table schema is:\nCREATE TABLE foo (id integer primary key, 12 | name text)"}, {"role": "user", "content": "Count of rows in foo"}], "model": 13 | "gpt-4.1-mini", "stream": false}' 14 | headers: 15 | accept: 16 | - application/json 17 | accept-encoding: 18 | - gzip, deflate 19 | connection: 20 | - keep-alive 21 | content-length: 22 | - '862' 23 | content-type: 24 | - application/json 25 | host: 26 | - api.openai.com 27 | user-agent: 28 | - AsyncOpenAI/Python 1.76.2 29 | x-stainless-arch: 30 | - arm64 31 | x-stainless-async: 32 | - async:asyncio 33 | x-stainless-lang: 34 | - python 35 | x-stainless-os: 36 | - MacOS 37 | x-stainless-package-version: 38 | - 1.76.2 39 | x-stainless-read-timeout: 40 | - '600' 41 | x-stainless-retry-count: 42 | - '0' 43 | x-stainless-runtime: 44 | - CPython 45 | x-stainless-runtime-version: 46 | - 3.12.0 47 | method: POST 48 | uri: https://api.openai.com/v1/responses 49 | response: 50 | body: 51 | string: !!binary | 52 | H4sIAAAAAAAAA3RUTW/bMAy951cQurQNmiJ2Pp0di+60rcDWnZrBVmTa0SZLrkS1K4r898Gy4yRd 53 | egkcPvKRfHrS2wCAyZytgFl0dTpfRhMxmyXTZSKWURJv5rN4kWwmSSHGcVIU4zhaJMs5LnJRRFE8 54 | ZtcNgdn8RkF7EqMdtnFhkRPmKW+waDGdR/E8micBc8TJu6ZGmKpWSJi3RWitsWwF2isVAlLvM9Ic 55 | iUvlTlFH1guSRp/EK/43NZ5qTymZP/gONDmqpndZ02h6E40qqeUoHsez0Xg6iqbdWqGcreBxAADw 56 | Fn57vSpX9nLxGQ9y8TyZTvJ8VuQJ5oLHZ+UKHPRaY2BB53iJB+AjXQIojCbUh5GOxzqh3W+Of6mv 57 | Dglca0N8r9bjrxMwpK+AZVnmntRaj0Zwa7wmoC0CGeIKtK82aMEUYM2LA6lbjG8UwkVhzMVa/7j7 58 | cnf7ALf3P789XA6vYK0/f7//CoUxn9Y6y7K1fthKB08e7SuIht8FkuHwXIvhcN8kK4zJulabV/BO 59 | 6rIF9q0y4GVpseSEUHgdTAEvWym2YJG81W2j8ztYdF4ROCS4DCHpQHCHwJVqEwtrqnaKqxvW67br 60 | vnopmTUqHAJ3TjrimtrkJjEksZpbrhSqlIxRqeAqWJqsxxa2+CyNd+n+NqXBcb17LXJntNQlW3Wn 61 | z7AojKWjpMZJvqq4fe2CA4Bde/HQPkuBKUlsrhnLseBetTZhjozF41kIqxotJx/C0c24iwajdM0L 62 | Yyt++H9kw5DXL9/2b3feGilakTwZ1gMHUzIydVof97Rei+DcMLV0jRG6N8OHO9QPJPXJtY/mk+v/ 63 | gaPHpJ9bcLHF/FA5bmfvqt8/J4voHHCOtz+xj6iD8w9gPJn2enmHJ89hhcRzTrzh3w12/wAAAP// 64 | AwDTjNIhwgUAAA== 65 | headers: 66 | CF-RAY: 67 | - 9391890dab042aa1-LAX 68 | Connection: 69 | - keep-alive 70 | Content-Encoding: 71 | - gzip 72 | Content-Type: 73 | - application/json 74 | Date: 75 | - Thu, 01 May 2025 19:02:50 GMT 76 | Server: 77 | - cloudflare 78 | Set-Cookie: 79 | - __cf_bm=KI41m5goq0MdStN7jaOvz0xBcEmATwG5A6PKwoMaD9A-1746126170-1.0.1.1-UhUtcNacquruYpq_QWWNF2qCcFuLWFWZXkoqtOsVx1wJ5zwiMjZVJc9H7t4NM1vIOwfZ8hTOqDd8E9qxuoGBx1uFrHo8paFEMH9NHMHbWuU; 80 | path=/; expires=Thu, 01-May-25 19:32:50 GMT; domain=.api.openai.com; HttpOnly; 81 | Secure; SameSite=None 82 | - _cfuvid=5.jV5jhOJnJ12IaU9aYvciYDPL_v0ZzAMlUR1kAaRNw-1746126170883-0.0.1.1-604800000; 83 | path=/; domain=.api.openai.com; HttpOnly; Secure; SameSite=None 84 | Transfer-Encoding: 85 | - chunked 86 | X-Content-Type-Options: 87 | - nosniff 88 | alt-svc: 89 | - h3=":443"; ma=86400 90 | cf-cache-status: 91 | - DYNAMIC 92 | openai-organization: 93 | - user-r3e61fpak04cbaokp5buoae4 94 | openai-processing-ms: 95 | - '1595' 96 | openai-version: 97 | - '2020-10-01' 98 | strict-transport-security: 99 | - max-age=31536000; includeSubDomains; preload 100 | x-request-id: 101 | - req_303e97d0c7b10f4659ad92ddb0579f82 102 | status: 103 | code: 200 104 | message: OK 105 | version: 1 106 | -------------------------------------------------------------------------------- /datasette_query_assistant/__init__.py: -------------------------------------------------------------------------------- 1 | from datasette import hookimpl, Response, Forbidden 2 | from datasette.resources import DatabaseResource 3 | import dataclasses 4 | from llm import get_async_model 5 | import re 6 | import urllib 7 | from markupsafe import escape 8 | import markdown2 9 | from datasette.utils import sqlite3 10 | import itsdangerous 11 | from typing import Tuple, Optional, Set 12 | 13 | SYSTEM_PROMPT = """ 14 | You answer questions by generating SQL queries using SQLite schema syntax. 15 | Always start with -- SQL comments explaining what you are about to do. 16 | No yapping. Output SQL with extensive SQL comments in a sql tagged 17 | fenced markdown code block. 18 | 19 | Return only one SQL SELECT query. Follow the query with an explanation 20 | of what the query does and how it works, which should include bold for 21 | emphasis where appropriate. 22 | 23 | Example question: 24 | 25 | How many rows in the sqlite_master table? 26 | 27 | Example output (shown between ----): 28 | ---- 29 | ```sql 30 | select count(*) from sqlite_master 31 | ``` 32 | Count the **number of rows** in the `sqlite_master` table. 33 | ---- 34 | The table schema is: 35 | """.lstrip() 36 | 37 | SCHEMA_SQL = """ 38 | select group_concat(sql, '; 39 | ') from sqlite_master where type != 'trigger' 40 | """ 41 | SCHEMA_SQL_SPECIFIC = """ 42 | select group_concat(sql, '; 43 | ') from sqlite_master where tbl_name in (PARAMS) and type != 'trigger' 44 | """ 45 | 46 | 47 | @dataclasses.dataclass 48 | class Config: 49 | model_id: str 50 | 51 | 52 | def config(datasette): 53 | return 54 | 55 | 56 | async def get_schema(db, table=None): 57 | if table: 58 | 59 | def _related(conn): 60 | return get_related_tables(conn, table) 61 | 62 | tables = await db.execute_fn(_related) 63 | tables.add(table) 64 | sql = SCHEMA_SQL_SPECIFIC.replace("PARAMS", ",".join("?" for _ in tables)) 65 | return (await db.execute(sql, tuple(tables))).first()[0] 66 | else: 67 | return (await db.execute(SCHEMA_SQL)).first()[0] 68 | 69 | 70 | async def has_permission(datasette, actor, database): 71 | return await datasette.allowed( 72 | action="execute-sql", 73 | resource=DatabaseResource(database), 74 | actor=actor, 75 | ) 76 | 77 | 78 | _sql_re = re.compile(r"```sql\n(?P