Configuration¶

Precedence¶

CLI flags
Environment: provider secrets (OPENROUTER_API_KEY, OPENAI_API_KEY, ELEVENLABS_API_KEY, XAI_API_KEY), OpenRouter base URL, and TTS (AURUM_TTS_*)
Config file
Built-in defaults

A provider is never selected merely because its API key is present. Omitted STT/TTS provider remains local.

Environment¶

Variable	Purpose
`OPENROUTER_API_KEY`	OpenRouter auth (preferred over file `api_key`)
`OPENROUTER_BASE_URL`	Override OpenRouter API base (tests / proxies)
`OPENAI_API_KEY`	OpenAI provider-scoped secret
`ELEVENLABS_API_KEY`	ElevenLabs provider-scoped secret
`XAI_API_KEY`	xAI provider-scoped secret
`AURUM_TTS_MODEL`	Override `[tts].model`
`AURUM_TTS_VOICE`	Override `[tts].voice`
`AURUM_TTS_LANGUAGE`	Override `[tts].language`
`RUST_LOG`	Tracing filters

Secrets are stored as redacting SecretString values. Effective-config / doctor / support output shows presence only (***), never plaintext.

Config file¶

Resolved via the directories crate (app name aurum):

Platform	Typical path
macOS	`~/Library/Application Support/aurum/config.toml`
Linux	`~/.config/aurum/config.toml`
Windows	`%APPDATA%\aurum\config.toml`

Canonical schema¶

[stt]
provider = "local"          # local | openrouter | openai | xai
model = "base"
language = "auto"
# output = "txt"            # txt | srt | json

[tts]
provider = "local"          # local | openrouter | openai | elevenlabs | xai
model = "kitten-nano-int8"
voice = "Luna"
language = "en"
speaking_rate = 1.0
max_chars = 5000
timeout_ms = 120000
# Optional local pack override (directory with aurum-tts-manifest.json — not a bare .onnx)
# pack_dir = "/path/to/pack"
# allow_unverified = false

# Optional custom catalogue entries (JOE-1620). Never shadow built-in ids.
# [[tts.custom_models]]
# id = "my-tone"
# adapter = "fake-sine-v1"
# pack_dir = "/path/to/pack"
# trust = "verified"   # verified | local_unverified (never builtin)
# license = "CC0"

[cleanup]
style = "raw"              # raw | clean | bullets | professional | summary
provider = "rules"         # rules | openrouter
# openrouter_model = "google/gemini-2.5-flash-lite"

# Named provider options + optional file secrets (prefer env vars for keys).
# Unknown [providers.*] keys fail closed.

# [providers.openrouter]
# stt_mode = "auto"          # auto | chat | transcriptions
# model = "google/gemini-2.5-flash"
# base_url = "https://openrouter.ai/api/v1"
# allow_custom_endpoint = false
# use_system_proxy = false

# [providers.openai]
# base_url = "https://api.openai.com/v1"

# [providers.elevenlabs]
# [providers.xai]

Only canonical sections are accepted: [stt], [cleanup], [tts], [providers.*]. Unknown top-level sections (including old [default] / [openrouter]) fail closed.

`local_only`¶

When local_only is set on the runtime/validated config (CLI offline flag and library builders), validation rejects a remote STT or TTS provider before encoding, upload, or request construction.

Safety limits¶

Limit	Default
Max duration (STT)	~2.25 hours
Max decoded PCM (STT)	~500 MB (enforced during decode)
Max remote upload (STT)	~24 MB compressed
Max TTS characters	5000 (`[tts].max_chars`)
TTS timeout	120000 ms (`[tts].timeout_ms`)
TTS speaking rate	1.0, allowed range `0.5..=2.0` (CLI clamps the same)

Whisper special tokens such as [BLANK_AUDIO] are stripped. Segment timestamps are clamped to audio duration.