lsegkd 0.1.2

A Python library for retrieving financial data using the LSEG Knowledge Direct API

LSEG Knowledge Direct Python API

A Python client for the LSEG Knowledge Direct REST API. It authenticates, downloads event headlines + transcripts, parses transcript XML into typed Pydantic models, and ships a Typer-based CLI for bulk download.

Features

• Authentication and short-lived service token issuance
• Paginated event-headline retrieval (GetEventHeadlines) with country and publication-status filters
• Sync and async transcript document fetching with automatic public-URL → signed-URL fallback (GetDocument)
• Pydantic models for headlines and parsed transcripts
• Typer CLI with progress bar, idempotent re-runs, and pagination loop

Installation

From PyPI

uv add lsegkd

or

pip install lsegkd

From source (development)

git clone https://github.com/yurukatsu/lseg-knowledge-direct.git
cd lseg-knowledge-direct
uv sync --all-groups

Authentication

Set the three environment variables (e.g. via a .env file — see .env.sample):

LSEG_KNOWLEDGE_DIRECT_USERNAME=your_username
LSEG_KNOWLEDGE_DIRECT_APP_ID=your_app_id
LSEG_KNOWLEDGE_DIRECT_PASSWORD=your_password

Or pass them directly when constructing Credentials:

from lsegkd import Credentials

credentials = Credentials(
    username="your_username",
    app_id="your_app_id",
    password="your_password",
)

CLI usage

# Minimum: a date range
uv run lsegkd -d .env fetch-transcripts --from-date 2025-11-22 --to-date 2025-11-23

# Multiple countries (comma-separated or repeated)
uv run lsegkd -d .env fetch-transcripts \
    --from-date 2025-11-22 --to-date 2025-11-23 \
    --countries US,UK,JP

# All transcript statuses, force re-download, custom output dir
uv run lsegkd -d .env fetch-transcripts \
    --from-date 2025-11-22 --to-date 2025-11-23 \
    --transcript-status any \
    --no-skip-existing \
    --output-dir ./data/raw

Run uv run lsegkd fetch-transcripts --help for the full option list.

Output layout

The fetch-transcripts command writes three files per event under output_dir:

<output_dir>/
├── xml/<EventId>.xml             raw transcript XML
├── headlines/<EventId>.json      EventHeadline metadata
└── transcripts/<EventId>.json    parsed EventTranscript

Re-runs skip events whose xml/ and transcripts/ JSON already exist (--skip-existing is on by default; pass --no-skip-existing to force).

Python API usage

import asyncio
from lsegkd import (
    Credentials,
    EventTranscriptParser,
    StreetEventsServiceClient,
    TokenManagementServiceClient,
)


async def main() -> None:
    credentials = Credentials()  # reads env vars

    # Exchange username/password for a short-lived service token
    token = TokenManagementServiceClient(credentials).create_service_token()

    client = StreetEventsServiceClient(
        credentials=credentials,
        token=token.token,
        async_mode=True,
    )

    try:
        response = client.get_event_headlines(
            from_date=...,  # datetime
            to_date=...,    # datetime
            countries=["US", "UK"],
            transcript_status="Final",
        )

        for headline in response.extract_event_headlines():
            if headline.Transcript is None:
                continue
            xml = await client.aget_document(
                transcript_id=headline.Transcript.TranscriptId
            )
            event = EventTranscriptParser.from_string(xml).parse()
            # event is a Pydantic EventTranscript — use it however you need
    finally:
        await client.aclose()


asyncio.run(main())

Pydantic data types (EventHeadline, Duration, Organization, Symbol, PaginationResult, ...) live in lsegkd.api.types and are also re-exported from lsegkd.api. Parsed transcript types (EventTranscript, Speaker, Episode, Section, Turn, ...) live in lsegkd.xml.transcript.

Development

uv sync --all-groups
uv run poe check       # ruff format + ruff check + ty + pytest
uv run poe test        # tests only

See CLAUDE.md for architecture notes and the conventions enforced by the codebase (HTTP / endpoint base classes, type-naming rules, the LSEG-API quirks pinned by tests).

License

MIT — see LICENSE.