docs: update AGENTS.md and CLAUDE.md for current architecture

Both were merged in from the original history and still described the
old three-module, MP3-only tool. Bring them in line with the actual
video-mode/x_api/models architecture.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
This commit is contained in:
2026-08-24 03:48:18 +02:00
parent 23bf673c12
commit 93eb346a00
2 changed files with 24 additions and 16 deletions

View File

@@ -8,8 +8,9 @@ This file provides guidance to Claude Code (claude.ai/code) when working with co
# Install (editable)
pip install -e .
# Run the CLI
# Run the CLI (video mode by default)
x-space-dl <URL>
x-space-dl --mode audio <URL>
python -m src.main <URL>
# Run all tests
@@ -27,14 +28,17 @@ pytest -m integration
## Architecture
Three-module CLI tool that wraps yt-dlp to download X/Twitter Space recordings as MP3:
CLI tool that downloads X/Twitter Space recordings and renders them as an MP4 replay with a participant view, or as plain MP3 audio:
- **`src/main.py`** — Click CLI entry point (`cli`). Validates URL via validator, then delegates to downloader. Registered as `x-space-dl` console script.
- **`src/validator.py`** — `validate_space_url()` returns `(is_valid, space_id, error_message)` tuple. Accepts both `x.com` and `twitter.com` domains.
- **`src/downloader.py`** — `download_space()` configures yt-dlp with `bestaudio/best` format and FFmpeg MP3 postprocessor. Output files are named `space_{SPACE_ID}.mp3`. Raises `DownloadError` on failure.
- **`src/main.py`** — Click CLI entry point (`cli`). Validates the URL, then delegates to `download_space` (audio mode) or `download_space_video` (video mode, default). Registered as `x-space-dl` console script.
- **`src/validator.py`** — `validate_space_url()` returns `(is_valid, space_id, error_message)`. Accepts both `x.com` and `twitter.com` domains.
- **`src/downloader.py`** — `download_space()` wraps yt-dlp for audio extraction. `download_space_video()` fetches metadata/timeline through `XApiClient`, downloads the audio, and renders the MP4 via `video.render_space_video()`.
- **`src/x_api.py`** — `XApiClient` calls X's web GraphQL (`AudioSpaceById`) and the Periscope replay chat endpoints to build participant metadata and a speaker timeline. Requires the `X_BEARER_TOKEN` environment variable (see README).
- **`src/video.py`** — Renders MP4 frames (Pillow) of the participant view and pipes them into ffmpeg alongside the audio track.
- **`src/models.py`** — Shared dataclasses: `Participant`, `SpaceMetadata`, `SpeakerInterval`.
Tests mock `yt_dlp.YoutubeDL` — no network calls in unit tests. Integration tests use the `@pytest.mark.integration` marker.
Tests mock `yt_dlp.YoutubeDL` and outbound HTTP calls — no real network access in unit tests. Integration tests use the `@pytest.mark.integration` marker.
## Dependencies
Runtime: `yt-dlp`, `click`. System: `ffmpeg`. Dev: `pytest`, `pytest-mock`.
Runtime: `yt-dlp`, `click`, `requests`, `Pillow`. System: `ffmpeg`. Dev: `pytest`, `pytest-mock`.