opensanctions-pravda 0.1.2

Durable web evidence capture with a remote browser, Postgres, and content-addressed storage

Pravda

Pravda is a Python library for durable web evidence capture. It drives a remote Playwright browser to preserve rendered HTML, plaintext, full-page screenshots, metadata, and HAR recordings with response bodies. Snapshots are recorded in Postgres and on any fsspec-compatible backend for later inspection or comparison.

Pravda is a library, not a service: it connects directly from the caller's process to the browser, database, and storage backend. Applications own that infrastructure (see Infrastructure).

• Python 3.13+
• Browser: a remote Playwright Chromium WebSocket endpoint (headed Chrome under xvfb). The browser is a client connection; Pravda does not launch one.
• Database: PostgreSQL, upgraded to Pravda's schema with the migration helper.
• Storage: any fsspec URL (local path, s3://, gs://, …) for content-addressed artifacts.

Installation

pip install opensanctions-pravda

Quick start

Construct a PravdaConfig, build a long-lived Pravda, and use it as an async context manager so its database engine is disposed on teardown. Reuse a single instance across captures — it owns the pooled engine, session factory, and storage backend, while each capture opens its own browser connection.

from pravda import Pravda, PravdaConfig

config = PravdaConfig(
    database_url="postgresql+asyncpg://pravda:pravda@localhost:5432/pravda",
    browser_ws_url="ws://localhost:3000",
    storage_base_path="./data",
)

async def capture_example():
    async with Pravda(config) as pravda:
        snapshot = await pravda.snapshot("https://example.com")
        print(snapshot.id, snapshot.http_status, snapshot.rendered_html)

Configuration

PravdaConfig takes three settings, supplied explicitly per instance:

• database_url — async SQLAlchemy Postgres URL (postgresql+asyncpg://user:pass@host/db).
• browser_ws_url — remote Playwright WebSocket URL.
• storage_base_path — fsspec storage URL, such as ./data, s3://bucket, or gs://bucket.

Usage

Capture a page

By default, Pravda navigates to the URL, waits for the normal load state, captures the evidence, and persists the result. The complete pipeline is bounded by a wall-clock timeout:

async def capture_example():
    async with Pravda(config) as pravda:
        snapshot = await pravda.snapshot("https://example.com")

For custom navigation or interaction, pass an async drive(page, url) callback. The callback owns the initial navigation; Pravda owns capture, persistence, and cleanup, and records the state it leaves behind.

async def drive(page, url):
    await page.goto(url, wait_until="commit")
    await page.wait_for_selector(".results")

async def capture_results():
    async with Pravda(config) as pravda:
        snapshot = await pravda.snapshot("https://example.com", drive=drive)

page is a real playwright.async_api.Page, so selectors, clicks, form fills, and other Playwright operations are available. Playwright errors and timeouts from drive are persisted as failed snapshots. Other callback exceptions propagate and persist nothing.

Chrome is configured to download PDFs instead of opening its viewer. In custom callbacks, page.goto() may therefore raise Download is starting; catch that Playwright error if the download is expected. Pravda recovers the downloaded body into the HAR.

Query history

The configured instance returns all snapshots for an exact URL, newest first:

async def print_history():
    async with Pravda(config) as pravda:
        history = await pravda.snapshots("https://example.com")
        for snapshot in history:
            print(snapshot.captured_at, snapshot.http_status)

Database migrations

Alembic owns the Pravda schema; the migration scripts ship inside the distribution. Bring a database up to the current schema head from application startup — the database URL is passed explicitly and no DATABASE_URL environment variable is required:

import pravda

async def setup():
    await pravda.migrate("postgresql+asyncpg://user:pass@host/db")

migrate() runs the packaged revisions through Alembic (not metadata.create_all), is safe to call repeatedly (a database already at head is a no-op), works from inside a running event loop, and disposes the engine it creates. Database and migration failures propagate. There is no downgrade or automatic-startup behavior: call migrate() where and when you want the schema applied.

Storage

Artifacts are content-addressed files under Snapshot.prefix. The plaintext, rendered_html, and screenshot fields contain filenames; HAR response.content._file fields refer to stored response bodies. Consumers read these files directly from the shared fsspec backend.

Infrastructure

Applications own the external infrastructure Pravda talks to; Pravda does not launch or manage it:

• Browser — a remote Playwright Chromium server exposed over WebSocket. Release images are published to GitHub Container Registry as ghcr.io/opensanctions/pravda-browser:<version> (and latest for the newest non-prerelease). The image runs headed Chrome under xvfb and accepts launch options through the x-playwright-launch-options WebSocket header. Run it locally with:

  docker run --rm --init -p 3000:3000 \
    ghcr.io/opensanctions/pravda-browser:latest
  

• Postgres — a database the application provisions and migrates.

• Storage — an fsspec backend the application points at via storage_base_path.

Development

Requires uv and Docker.

# Start the remote browser and Postgres
docker compose up -d

# Install dependencies
uv sync

# Validate
uv run pytest
uv run ruff check .
uv run ruff format --check .

The migration scripts live inside the package at pravda/migrations. After changing models in pravda/db.py, the developer alembic command reads DATABASE_URL and points at the packaged scripts via alembic.ini:

DATABASE_URL=postgresql+asyncpg://pravda:pravda@localhost:5432/pravda \
  uv run alembic upgrade head
DATABASE_URL=postgresql+asyncpg://pravda:pravda@localhost:5432/pravda \
  uv run alembic revision --autogenerate -m "describe the change"

License

MIT — see LICENSE.