n8n-nodes-parakeet

An n8n community node for Parakeet — offline speech-to-text powered by parakeet.cpp (the parakeet-cli binary from the official ghcr.io/ggml-org/whisper.cpp:main image) and the ggml-parakeet-tdt-0.6b-v3-q8_0 model.

Runs fully on-device / offline. No API keys, no cloud.

Install into n8n

n8n add-nodes n8n-nodes-parakeet
# or, in a container:
docker exec -it <n8n> n8n nodes:install n8n-nodes-parakeet

On first run the node self-heals (no postinstall):

1. downloads the ~3 MB parakeet-cli + shared libraries to ~/.n8n/parakeet/bin/<platform>-<arch>/, and
2. downloads the ~638 MB model to ~/.n8n/parakeet/models/ggml-parakeet-tdt-0.6b-v3-q8_0.bin.

Neither the binary nor the model is shipped in the npm package. Both are fetched once and cached under n8n's standard data volume (~/.n8n), so they survive container/pod restarts. To use local copies instead, set PARAKEET_CLI_PATH / PARAKEET_MODEL_PATH.

Node inputs

| Input | Maps to parakeet-cli | Description | |---|---|---| | Input = Binary / URL | -f FILE | Audio source: a binary property from a previous node, or a URL (wav/mp3/flac/ogg). | | Binary Property | — | Name of the binary property holding the audio (default data). | | Audio URL | -f | Used when Input = URL. | | Model Path | -m FILE | Leave empty to use the cached q8_0 model. | | Options → Threads | -t N | CPU threads (default 4). | | Options → Use GPU | (omit -ng) | Enable GPU; requires a CUDA/Metal build. Default off. | | Options → GPU Device | -dev N | GPU device index. | | Options → Print Segments | -ps | Include per-segment / per-token detail. | | Options → Write Text File | -otxt | Also write a .txt (returned as a transcript binary property). | | Options → Output File Base Path | -of FILE | Base path for the .txt output. | | Options → Quiet | -np | Suppress parakeet-cli logging; only the transcript is returned. |

Every parakeet-cli flag is exposed as a node input.

Output

Each input item produces one output item:

{
  "json": {
    "text": "And so, my fellow Americans, ask not what your country can do for you, ask what you can do for your country.",
    "model": "ggml-parakeet-tdt-0.6b-v3-q8_0.bin",
    "input": "data"
  }
}

With Write Text File enabled, a transcript binary property (the .txt) is also attached.

Environment variables

| Variable | Purpose | |---|---| | PARAKEET_MODEL_PATH | Use this exact .bin file; skips the auto-download. | | PARAKEET_CLI_PATH | Use this exact parakeet-cli binary; skips the auto-download. | | PARAKEET_MODEL_URL | Override the model download URL. |

Binary + model (runtime download)

The npm package contains no binaries or weights. On first run the node downloads:

• parakeet-cli + its libparakeet/libggml shared libraries (~3 MB) from a GitHub Release asset, extracted to ~/.n8n/parakeet/bin/<platform>-<arch>/. The binary uses $ORIGIN rpath so it runs with no LD_LIBRARY_PATH.
• the ~638 MB model from Hugging Face to ~/.n8n/parakeet/models/.

Both live under n8n's standard data volume, so they survive container/pod restarts. The binary is extracted from the official ghcr.io/ggml-org/whisper.cpp:main Docker image by .github/workflows/build-and-publish.yml (a new GitHub Release is cut per tag).

Build from source

npm install --ignore-scripts   # n8n-workflow pulls a native dep that needs no build here
npm run build                  # tsc -> dist/nodes/Parakeet/Parakeet.node.js
node smoke.test.js             # end-to-end smoke test (needs model + jfk.wav)

License

MIT. The parakeet-cli binary and the Parakeet model are subject to their respective licenses (whisper.cpp / Parakeet).