Storage

audia stores all data in a local SQLite database managed by SQLAlchemy.

Location

Default: ~/.audia/audia.db

Override with:

AUDIA_DATA_DIR=/path/to/your/data

The full directory structure:

~/.audia/
  audia.db          ← SQLite database
  audio/            ← generated .mp3 files
  uploads/          ← PDFs uploaded via the web UI
  debug/            ← per-run text snapshots
    <run_id>/
      1_raw.txt
      2_preprocessed.txt
      3_curated.txt

---

Schema

papers

Stores academic papers (from ArXiv or uploaded manually).

Column	Type	Notes
id	INTEGER PK	Auto-increment
title	VARCHAR(512)	Document title
authors	TEXT	JSON array of strings
abstract	TEXT	Abstract text
arxiv_id	VARCHAR(64)	ArXiv ID (e.g. 2603.12345)
pdf_path	TEXT	Local path to the PDF
pdf_url	TEXT	Remote URL
created_at	DATETIME	UTC timestamp

audio_files

Generated audio files, optionally linked to a paper.

Column	Type	Notes
id	INTEGER PK	Auto-increment
paper_id	INTEGER FK	→ papers.id (SET NULL on delete)
filename	TEXT	File name
file_path	TEXT	Absolute path to .mp3
tts_backend	VARCHAR	edge-tts | kokoro | openai
tts_voice	VARCHAR	Voice name used
duration_seconds	FLOAT	Audio duration
created_at	DATETIME	UTC timestamp

research_sessions

Stores ArXiv query sessions.

Column	Type	Notes
id	INTEGER PK	Auto-increment
query	TEXT	User search query
created_at	DATETIME	UTC timestamp

user_settings

Key-value store for persisted configuration (LLM provider, TTS backend, etc.).

Column	Type	Notes
key	VARCHAR PK	Setting name
value	TEXT	Setting value

---

Exploring the database

Use the bundled explore_db.py script for a full terminal dump:

python scripts/explore_db.py
python scripts/explore_db.py --db /custom/path/audia.db

The script prints table schema, foreign keys, row counts, and all cell values using only the Python standard library (no extra dependencies).