Notex

To start your Phoenix server:

• Run mix setup to install and setup dependencies
• Start Phoenix endpoint with mix phx.server or inside IEx with iex -S mix phx.server

Now you can visit localhost:4000 from your browser.

Ready to run in production? Please check our deployment guides.

Local media dependencies

Video generation uses ffmpeg for MP4 rendering and a local TTS command for narration. Install these before using the Video media action:

sudo apt-get update
sudo apt-get install -y ffmpeg open-jtalk open-jtalk-mecab-naist-jdic hts-voice-nitech-jp-atr503-m001 fonts-noto-cjk

Japanese narration uses Open JTalk when it is installed. You can tune its rate with:

export NOTEX_OPEN_JTALK_RATE="1.0"

If Open JTalk is not installed, Notex falls back to espeak-ng/espeak when available. That fallback is mainly for basic narration and is not recommended for Japanese.

sudo apt-get install -y espeak-ng

Learn more

• Official website: https://www.phoenixframework.org/
• Guides: https://hexdocs.pm/phoenix/overview.html
• Docs: https://hexdocs.pm/phoenix
• Forum: https://elixirforum.com/c/phoenix-forum
• Source: https://github.com/phoenixframework/phoenix

LLM synthesis

Notex uses local retrieval first, then can synthesize cited answers with Codex app-server using GPT-5.5 low reasoning. This is the default provider when the codex command is available:

export NOTEX_LLM_PROVIDER="codex_app_server"
export NOTEX_LLM_MODEL="gpt-5.5"
export NOTEX_LLM_REASONING_EFFORT="low"

You can also use the OpenAI Responses API-compatible provider:

export NOTEX_LLM_PROVIDER="openai"
export OPENAI_API_KEY="..."
export NOTEX_LLM_MODEL="gpt-5.5"
export NOTEX_LLM_REASONING_EFFORT="low"
export NOTEX_LLM_BASE_URL="https://api.openai.com/v1"

If Codex app-server or API credentials are unavailable, Notex returns an explicit LLM error instead of generating a deterministic fallback answer.