NeuroShard
Contribute neural computation and use native NEURO to pay for inference. NeuroShard runs its own blockchain. Your keys stay on your machine; joining requires no website account or registration token.
Experimental testnet. This release uses a small pretrained instruction model with a trainable adapter. Validators replay the computation before accepting work or paying providers. Test NEURO has no promised monetary value or mainnet conversion.
Install and join
python3 -m venv ~/.venvs/neuroshard
source ~/.venvs/neuroshard/bin/activate
python -m pip install --upgrade pip setuptools neuroshard-ai
neuroshard doctor
neuroshard join
The Python package is a lightweight client. join installs a separate, pinned CPU
runtime, downloads and checks the model and network data, creates a local key,
starts a full node, and contributes training stage 1 when sponsored work is available.
It displays its progress. Press Ctrl+C to stop; your key and history remain on disk.
Full nodes require Linux x86_64 and Python 3.10–3.12. Allow at least 8 GiB RAM and 5 GiB free disk for the runtime, model, and initial history. Supported workers use the pinned CPU profile; arbitrary GPU execution is not accepted by this release. Wallet and remote inference commands do not download the neural runtime.
neuroshard status
neuroshard wallet export ./neuroshard-key.json
Keep the exported key private. It controls the account. Rewards are paid only when the complete task is accepted by native validators; returning work alone is not payment.
Use the model
After earning NEURO or receiving a native transfer:
neuroshard chat "Explain what a blockchain is in one sentence." --max-price 0.1
The client shows the total budget and request ID. The chain locks the inference
budget, checks the provider's output, and settles payment. A timed-out request
unlocks its budget; the submission fee is spent. Use neuroshard request REQUEST_ID
to inspect an uncertain or pending result before submitting another request.
Prompts and responses are public. This small model can make mistakes and is not suitable for sensitive information. The current deployment has one operator across two hosts; independent operators and economical large-model verification remain open work.
Choose how to participate
neuroshard join --role observer # Verify the ledger without contributing training
neuroshard join --role provider # Run a node and serve requests addressed to your key
neuroshard --help # Wallet, transfer, request, and connection options
An existing participant can use neuroshard start to resume. Use --home for a
separate identity and --network-file for another reviewed native network. A remote
RPC supplies a view of the ledger; running a full node independently replays it.
Upgrading from 0.2.x replaces the old neuroshard --token ... client. Website
registration tokens and the earlier observer balances do not migrate to the native
chain. Version 0.3's tiny reference chain remains a separate research network.
Website · Operator guide · Source and issues
Apache License 2.0. Pretrained model and dataset provenance are documented in the repository.
1 maintainer
Metadata
1 maintainer
Release files for neuroshard-ai 0.4.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| neuroshard_ai-0.4.0.tar.gz | 1.0 MB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| neuroshard_ai-0.4.0-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 2.1 MB
Release files / neuroshard_ai-0.4.0.tar.gz
| Download URL | neuroshard_ai-0.4.0.tar.gz |
|---|---|
| Size | 1.0 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
dbc6518bec1860200edc73e8e4133fc0890e871cd7ed3e71861a085bf24d190a
|
|
BLAKE2b-256 checksum How to use checksums |
d8af7ea6866e9d0a357aa85181b217968e9a3da9efe11e5620a309593262f7de
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.10.12
|
Release files / neuroshard_ai-0.4.0-py3-none-any.whl
| Download URL | neuroshard_ai-0.4.0-py3-none-any.whl |
|---|---|
| Size | 1.1 MB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
df547b21ba148b4e6b83142c2635bb3c74a5b6dd54673aba042745e78c614e32
|
|
BLAKE2b-256 checksum How to use checksums |
8546e5ea403ee929066a42f65da85c030ea7fbd633dbf65fc4f73a3b9bf2550a
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.10.12
|
