Pixeltable Cloud is in Limited Beta. Email contact@pixeltable.com if you are interested.

Quickstart | Documentation | CLI | Cloud | Discord

Install into your agent: Agent Skill · MCP server · Cursor plugin · ChatGPT plugin

Reading this as an agent? Start at get-started.md, then skill.md and llms-full.txt.

Pixeltable is the database, orchestration, and serving layers. Images, video, audio, and documents live in tables. A transform is a computed column. An index is a declaration, and so is an HTTP route. Insert a row and everything below it runs. Object storage, a vector database, an orchestrator, and the endpoint code that copies between them collapse into one application file. That file is what a coding agent can hold: it writes the whole application at once, and you review the change in one diff instead of across four systems.

pip install 'pixeltable[serve]'
pxt init
pxt service example --out app.py
pxt schema update app.py my_app
pxt service update app.py my_app

pxt schema update creates the catalog my_app and its tables; it does not start HTTP. pxt service update starts HTTP; it does not create tables.

pxt service example writes this application file.

import pixeltable as pxt
import pixeltable.functions as pxtf
from pixeltable.serving import FastAPIRouter

TableModel = pxt.model_base()


@pxt.udf                                        # a Python function the columns below can call
def excerpt(text: str, n: int = 12) -> str:
    return text if len(text) <= n else f'{text[:n]}...'


class Docs(TableModel, name='docs'):
    id = pxt.Column(value=pxtf.uuid.uuid7(), primary_key=True)  # a generated key: provided automatically on insert
    title: pxt.String
    body: pxt.String | None
    title_upper = pxtf.string.upper(title)      # an assignment: computed on insert and on update
    summary = excerpt(title)                    # a computed column over a udf this file defines


ingest = FastAPIRouter(name='ingest')
ingest.add_insert_route(                        # POST /docs inserts and returns the computed columns
    Docs, path='/docs', inputs=[Docs.title, Docs.body], outputs=[Docs.id, Docs.title_upper, Docs.summary]
)
ingest.add_update_route(                        # POST /docs/update takes id plus the new values
    Docs, path='/docs/update', inputs=[Docs.title], outputs=[Docs.id, Docs.title_upper]
)
ingest.add_compute_route(Docs, path='/titles', inputs=[Docs.title], outputs=[Docs.title_upper])

The same file holds pxt.Image, pxt.Video, pxt.Audio, or pxt.Document columns, and a computed column over one of them is another assignment: media pipelines, RAG. The port is assigned, so read it back rather than hardcoding it:

URL=$(pxt service list --json | jq -r '.[0].endpoint')
curl -X POST "$URL/docs" \
  -H 'Content-Type: application/json' \
  -d '{"title": "Hello", "body": "world"}'
# {"id":"...","title_upper":"HELLO","summary":"Hello"}

The same file runs on Pixeltable Cloud. Create an API key in the Cloud dashboard, set PIXELTABLE_API_KEY, name the database in pixeltable.toml, then target it by URI. pxt db update creates or updates the hosted database; it does not insert rows. pxt service run is local only and cannot target Cloud.

pxt db update pxt://org:mydb
pxt schema update app.py pxt://org:mydb
pxt service update app.py pxt://org:mydb

A @pxt.udf in that same app.py is in the image pxt db update builds.

uvx pixeltable-new copies one app from the starter kit. The default copy is a chat app; pass agent as the last argument to pxt schema update. --video copies video search; pass videointel.

uvx pixeltable-new myapp
cd myapp
uv sync
pxt schema update app.py agent
pxt service update app.py agent

Inserting into the knowledge table needs no API key. The /ask route needs ANTHROPIC_API_KEY.

To mount the routes on an existing FastAPI app, app.include_router(...). HTTP serving. To skip endpoints, run pxt schema update, insert from Python, then export_sql. Self-hosting.

The links above install the skill and MCP. How the skill writes app.py: AI coding agents.

npx skills add pixeltable/pixeltable-skill

The skill writes a TableModel in app.py. If the agent writes create_table in application code, names the file schema.py, or writes the removed command pxt serve, the installed skill is stale: reinstall npx skills add pixeltable/pixeltable-skill.

Notebooks and tests still use pxt.create_table(). An app puts tables in app.py and creates them with pxt schema update.

Apache 2.0. Contributing · Discord