TAO—Block—Subnets85 / 128active
Inference Optimization by Economic Kernel Search
what the project says about itself — the input for semantic labels
Inference Optimization by Economic Kernel Search
scope of the data
Taonets fetches the repo's public metadata (description, topics, language, stars, last push) and the head of its README. Commit-count histories, contributor lists and release notes require the GitHub commits/releases endpoints, which the worker doesn't consume yet — those panels are coming with the ingest upgrade and are not simulated in the meantime.
first lines of the default branch README, unedited
# Cacheon Cacheon is an inference-speed competition built around SGLang and designed to operate as a Bittensor subnet. Miners submit inspectable contributions at registered boundaries of a validator-owned engine: modules of the served model, or the scheduler's prefix cache. Validators admit those contributions through a typed, isolated pipeline and reward only improvements measured end to end on the registered agent workload under its speed and quality policy. This repository—published as [`latent-to/cacheon`](https://github.com/latent-to/cacheon)—contains the miner SDK and examples, validator and chain control plane, evaluation runtime, and settlement and incentive machinery. > [!IMPORTANT] > Cacheon is pre-release software. Implemented paths, retained empirical evidence, > and production readiness are separate claims. Start with the > [validator first hour](docs/validator-guide/first-hour.md) for what a deployment must prove. ## Start here | Goal | Documentation | |---|---| | Understand the system | [Concepts](docs/get-started/concepts.md) and the [product model](docs/architecture/product-model.md) | | Understand why miners participate and how rewards work | [How miners earn rewards](docs/miner-guide/incentives.md) | | Build a miner contribution | [Miner guide](docs/miner-guide/overview.md) | | Operate a validator | [Validator guide](docs/validator-guide/overview.md) | | Review trust boundaries | [Security model](docs/security/threat-model.md) | | Contribute to the repository | [Contributing](CONTRIBUTING.md) | The canonical documentation source lives under [`docs/`](docs/) in this repository and is rendered at [cacheon.ai/docs](https://cacheon.ai/docs/). ## Local correctness loop Python 3.11 or newer is required. ```bash python3 -m venv .venv source .venv/bin/activate python -m pip install -e ".[cpu,dev]" python -m cacheon.cli scan examples/miner_node_identity python -m cacheon.cli verify examples/miner_node_identity ``` `scan` and `verify` are development checks; `check` runs the node audit and graph checks in the published arena image. They do not establish serving throughput, end-to-end quality, settlement eligibility, or a production release. Those decisions belong to the validator-owned qualification path described in the documentation. ## Design boundaries - The validator owns the model, workload, references, timing, outputs, and reward policy. A miner contribution owns only its registered target. - Candidate build and execution run in validator-owned, no-egress OCI workers; wallet and chain-signing authority remain outside candidate lifetimes. - One complete audited qualification PASS makes a contribution eligible for settlement, which reopens its retained evidence before recording a crown. - Evaluation acceptance and serving are different decisions. Nothing in the repository turns a crown into a release; that is a maintainer decision made outside it. See the [product model](docs/architecture/product-model.md) and [slot contract](docs/architecture/slot-contract.md) for the normative invariants. ## Tests ```bash python -m pytest -q tests ``` Documentation checks are described in [Contributing](CONTRIBUTING.md#documentation). ## License The repository is licensed under [Apache-2.0](LICENSE). Miner submissions are governed separately by the [draft submission terms](docs/legal/submission-terms.md); those terms require legal review before production use.