danvics/openreader
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
openreader/.env.example
Richard R 874e5ef359 refactor(whisper): migrate word alignment to ONNX backend and remove whisper.cpp integration 
Replace the previous whisper.cpp-based word alignment with a fully ONNX-based
implementation using onnxruntime-node and @huggingface/tokenizers. Add new
Whisper ONNX model management, alignment mapping, and spectral analysis modules.
Remove all code and documentation referencing whisper.cpp, update environment
variables, Dockerfile, and docs to reflect ONNX-only alignment. Add unit tests
for alignment and ONNX model logic.
2026-05-19 13:00:21 -06:00

124 lines
6.9 KiB
Text
Raw Blame History

# Local / OpenAI TTS API Configuration (default)
# Suggest using https://github.com/remsky/Kokoro-FastAPI
#
# NOTE: On first boot, the server auto-seeds these values into a "default-openai"
# admin-managed shared provider (DB-backed, encrypted at rest). After that, the
# admin UI is authoritative and changing these env vars has no effect. You may
# remove them once the seed has run. See Settings → Admin → Shared providers.
API_BASE=http://localhost:8880/v1
API_KEY=api_key_optional
# (Optional) TTS request/cache tuning (leave unset to use defaults)
# TTS_CACHE_MAX_SIZE_BYTES=268435456 # 256MB
# TTS_CACHE_TTL_MS=1800000 # 30 minutes
# TTS_MAX_RETRIES=2
# TTS_RETRY_INITIAL_MS=250
# TTS_RETRY_MAX_MS=2000
# TTS_RETRY_BACKOFF=2
# TTS_UPSTREAM_TIMEOUT_MS=285000 # 285 seconds
# (Optional) Enable TTS character rate limiting (default is `false`)
# TTS_ENABLE_RATE_LIMIT=true
# (Optional) TTS per-user daily limits (leave unset to use defaults)
# TTS_DAILY_LIMIT_ANONYMOUS=50000
# TTS_DAILY_LIMIT_AUTHENTICATED=500000
# (Optional) TTS IP backstop daily limits (leave unset to use defaults)
# TTS_IP_DAILY_LIMIT_ANONYMOUS=100000
# TTS_IP_DAILY_LIMIT_AUTHENTICATED=1000000
# Auth configuration (recommended for contributors and public instances)
# (Optional) Auth is only enabled when AUTH_SECRET and BASE_URL are set
BASE_URL=http://localhost:3003 # Externally facing URL for this app (set to LAN IP for access from other devices on the network)
AUTH_SECRET=some_random_secret_key # Generate with `openssl rand -hex 32`
AUTH_TRUSTED_ORIGINS=http://localhost:3003,http://127.0.0.1:3003 # Additional trusted origins (BASE_URL is always trusted)
# (Optional) Allow anonymous auth sessions when auth is enabled (default: `false`)
USE_ANONYMOUS_AUTH_SESSIONS=
# (Optional) Sign in w/ GitHub Configuration
GITHUB_CLIENT_ID=
GITHUB_CLIENT_SECRET=
# (Optional) Disable Better Auth built-in rate limiting (useful for testing)
DISABLE_AUTH_RATE_LIMIT=
# (Optional) Comma-separated list of emails that are auto-promoted to admin.
# Admins see the "Admin" tab in Settings (TTS shared providers + site features).
# Demotion is automatic: removing an email here demotes the user on next login.
# Requires auth to be enabled (AUTH_SECRET + BASE_URL).
ADMIN_EMAILS=
# (Optional) Backend DB used for server-side metadata (documents/audiobooks) and, when auth is enabled, auth tables.
# Defaults to SQLite at docstore/sqlite3.db when not set.
POSTGRES_URL=
# Embedded SeaweedFS weed mini config
# (Optional) Enable embedded weed mini for local S3-compatible storage (default: `true`)
USE_EMBEDDED_WEED_MINI=
WEED_MINI_DIR=
WEED_MINI_WAIT_SEC=
# S3 storage config (use with embedded weed mini or external S3-compatible storage)
# (Optional) For embedded weed mini, set explicit keys if you want stable credentials across restarts.
S3_ACCESS_KEY_ID=
S3_SECRET_ACCESS_KEY=
S3_BUCKET=
S3_REGION=
# (Optional) If empty in embedded mode, OpenReader uses BASE_URL host (when set) or detected LAN host.
S3_ENDPOINT=
S3_FORCE_PATH_STYLE=
S3_PREFIX=
# Migrations configuration
# (Optional) Skip automatic startup migrations when set to `false` (default: `true`)
RUN_DRIZZLE_MIGRATIONS=
# (Optional) Skip automatic filesystem->S3/DB migration pass when set to `false` (default: `true`)
RUN_FS_MIGRATIONS=
# (Optional) Server library import roots (uses /docstore/library by default)
IMPORT_LIBRARY_DIR=
IMPORT_LIBRARY_DIRS=
# Heavy compute backend mode for ONNX whisper alignment + PDF layout parsing.
# local = run compute in-process (default)
# none = disable both capabilities (good for preview/serverless)
# worker = reserved for future external worker mode (not implemented in v1)
OPENREADER_COMPUTE_MODE=local
# OPENREADER_COMPUTE_WORKER_URL=
# OPENREADER_COMPUTE_WORKER_TOKEN=
# Optional overrides for Whisper ONNX artifacts
# Defaults target: onnx-community/whisper-base_timestamped int8
# OPENREADER_WHISPER_MODEL_CONFIG_URL=https://huggingface.co/onnx-community/whisper-base_timestamped/resolve/main/config.json
# OPENREADER_WHISPER_MODEL_GENERATION_CONFIG_URL=https://huggingface.co/onnx-community/whisper-base_timestamped/resolve/main/generation_config.json
# OPENREADER_WHISPER_MODEL_TOKENIZER_URL=https://huggingface.co/onnx-community/whisper-base_timestamped/resolve/main/tokenizer.json
# OPENREADER_WHISPER_MODEL_TOKENIZER_CONFIG_URL=https://huggingface.co/onnx-community/whisper-base_timestamped/resolve/main/tokenizer_config.json
# OPENREADER_WHISPER_MODEL_MERGES_URL=https://huggingface.co/onnx-community/whisper-base_timestamped/resolve/main/merges.txt
# OPENREADER_WHISPER_MODEL_VOCAB_URL=https://huggingface.co/onnx-community/whisper-base_timestamped/resolve/main/vocab.json
# OPENREADER_WHISPER_MODEL_NORMALIZER_URL=https://huggingface.co/onnx-community/whisper-base_timestamped/resolve/main/normalizer.json
# OPENREADER_WHISPER_MODEL_ADDED_TOKENS_URL=https://huggingface.co/onnx-community/whisper-base_timestamped/resolve/main/added_tokens.json
# OPENREADER_WHISPER_MODEL_PREPROCESSOR_URL=https://huggingface.co/onnx-community/whisper-base_timestamped/resolve/main/preprocessor_config.json
# OPENREADER_WHISPER_MODEL_SPECIAL_TOKENS_MAP_URL=https://huggingface.co/onnx-community/whisper-base_timestamped/resolve/main/special_tokens_map.json
# OPENREADER_WHISPER_MODEL_ENCODER_URL=https://huggingface.co/onnx-community/whisper-base_timestamped/resolve/main/onnx/encoder_model_int8.onnx
# OPENREADER_WHISPER_MODEL_DECODER_MERGED_URL=https://huggingface.co/onnx-community/whisper-base_timestamped/resolve/main/onnx/decoder_model_merged_int8.onnx
# OPENREADER_WHISPER_MODEL_DECODER_WITH_PAST_URL=https://huggingface.co/onnx-community/whisper-base_timestamped/resolve/main/onnx/decoder_with_past_model_int8.onnx
# Optional overrides for PDF layout model artifacts
# OPENREADER_PDF_LAYOUT_MODEL_URL=https://huggingface.co/Bei0001/PP-DocLayoutV3-ONNX/resolve/main/PP-DocLayoutV3.onnx
# OPENREADER_PDF_LAYOUT_MODEL_DATA_URL=https://huggingface.co/Bei0001/PP-DocLayoutV3-ONNX/resolve/main/PP-DocLayoutV3.onnx.data
# OPENREADER_PDF_LAYOUT_CONFIG_URL=https://huggingface.co/Bei0001/PP-DocLayoutV3-ONNX/resolve/main/config.json
# OPENREADER_PDF_LAYOUT_PREPROCESSOR_URL=https://huggingface.co/Bei0001/PP-DocLayoutV3-ONNX/resolve/main/preprocessor_config.json
# (Optional) Override ffmpeg binary path used for audiobook processing
FFMPEG_BIN=
# (Optional) Client feature flags — seeded into the admin-managed runtime
# config on first boot, then ignored. Edit values from Settings → Admin →
# Site features instead of redeploying. SSR-injected so they take effect
# without rebuilding (unlike the old NEXT_PUBLIC_* build-time pattern).
# NEXT_PUBLIC_ENABLE_DOCX_CONVERSION=true
# NEXT_PUBLIC_ENABLE_DESTRUCTIVE_DELETE_ACTIONS=true
# NEXT_PUBLIC_ENABLE_TTS_PROVIDERS_TAB=true
# NEXT_PUBLIC_ENABLE_USER_SIGNUPS=true
# NEXT_PUBLIC_RESTRICT_USER_API_KEYS=true
# NEXT_PUBLIC_DEFAULT_TTS_PROVIDER=custom-openai
# NEXT_PUBLIC_CHANGELOG_FEED_URL=https://docs.openreader.richardr.dev/changelog/manifest.json
# NEXT_PUBLIC_ENABLE_AUDIOBOOK_EXPORT=true
Reference in a new issue View git blame Copy permalink