nextron-cloud 0.1.0rc1__tar.gz

nextron_cloud-0.1.0rc1/.gitignore

@@ -0,0 +1,9 @@
+__pycache__/
+*.py[cod]
+.venv/
+dist/
+build/
+*.egg-info/
+.pytest_cache/
+.mypy_cache/
+.DS_Store

nextron_cloud-0.1.0rc1/CHANGELOG.md

@@ -0,0 +1,13 @@
+# Changelog
+
+## 0.1.0rc1
+
+First release candidate.
+
+- `NexTron` client: `me`, `models`, `layouts`; datasets (upload with a local format check, Hugging Face import,
+  wait, download link, delete); jobs (estimate with the server's checks, create with name and tags, wait, progress,
+  events, logs, metrics, billing, cancel, checkpoint download with sha256 checks, export).
+- Data helpers: `to_messages`, `write_jsonl`, `check_file`, `prepare_file` for messages, ShareGPT, Alpaca, Dolly,
+  OpenOrca, prompt/completion, question/answer, problem/solution and input/output rows.
+- Experiments: `sweep` (one job per combination, repeat-safe) and `report`.
+- `nextron-cloud` CLI, JSON output; `nextron-cloud guide` prints the agent guide shipped in the package.

nextron_cloud-0.1.0rc1/LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

nextron_cloud-0.1.0rc1/PKG-INFO

@@ -0,0 +1,102 @@
+Metadata-Version: 2.5
+Name: nextron-cloud
+Version: 0.1.0rc1
+Summary: Official Python client for the NexTron Cloud fine-tuning API.
+Project-URL: Homepage, https://nextron.physionlabs.ai
+Project-URL: Issues, https://github.com/Physion-Labs/nextron-python/issues
+Author-email: Physion Labs <support@physionlabs.ai>
+License-Expression: Apache-2.0
+License-File: LICENSE
+Keywords: fine-tuning,nextron,physion,training
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Developers
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Typing :: Typed
+Requires-Python: >=3.10
+Requires-Dist: httpx<1,>=0.27
+Requires-Dist: pydantic<3,>=2.9
+Description-Content-Type: text/markdown
+
+# nextron-cloud
+
+> **Release candidate.** The API may still change before 0.1.0; install with `--prerelease=allow`.
+
+Python SDK and CLI for [NexTron Cloud](https://nextron.physionlabs.ai): fine-tune open models with one request
+(model + dataset + settings) and download the trained checkpoint.
+
+```bash
+uv venv && uv pip install --prerelease=allow nextron-cloud   # or, in a uv project: uv add --prerelease=allow nextron-cloud
+export NEXTRON_API_KEY="nk_live_..."                           # API keys page on nextron.physionlabs.ai
+uv run nextron-cloud models                                    # what can be trained now, at what price
+```
+
+Commands below are written as `nextron-cloud ...`; run them with `uv run` (or activate `.venv`).
+
+```python
+from nextron_cloud import NexTron
+
+client = NexTron()
+dataset = client.datasets.wait(client.datasets.upload("train.jsonl").id)   # checked locally, then by the server
+draft = dict(model="qwen3.8-27b", dataset_id=dataset.id, params={"train_type": "lora", "learning_rate": 1e-4})
+print(client.jobs.estimate(**draft))                                        # cost in US cents: get a yes first
+
+job = client.jobs.wait(client.jobs.create(**draft, name="first try").id)  # only after the cost is approved
+print(job.state, client.jobs.metrics(job.id)[-1])
+
+client.jobs.wait_checkpoint(job.id)                                         # later: weights stay 30 days
+client.jobs.download_checkpoint(job.id, f"checkpoints/{job.id}")
+```
+
+## Start with an agent
+
+Install, set the key, then paste this to your coding agent (Claude Code, Codex, ...) and fill in the brackets:
+
+```text
+Run a fine-tuning experiment for me on NexTron Cloud, a managed training service: one request trains a model on
+its GPUs and leaves a checkpoint to download.
+
+Setup
+- Install the SDK with uv in this project: `uv venv && uv pip install --prerelease=allow nextron-cloud`
+  (in a uv project: `uv add --prerelease=allow nextron-cloud`). Run its commands with `uv run`, e.g.
+  `uv run nextron-cloud guide`.
+- My API key is in the environment variable NEXTRON_API_KEY. Never print it or write it into a file.
+- Before anything else run `uv run nextron-cloud guide` and read all of it. At each step read its topic too:
+  `nextron-cloud guide data`, `... training`, `... monitoring`, `... checkpoints`, `... api`. Follow its rules.
+
+The experiment
+- Goal: [what the model should do better, e.g. "answer our support tickets in our tone and format"]
+- Data: [file path or Hugging Face link, and which fields are the input and which the answer]
+- How I will judge it: [e.g. "held-out tickets, compared with the base model"]
+- Model: [a model id, or "pick one from `nextron-cloud models` and tell me why"]
+- Budget: at most $[N] for this experiment unless I approve more.
+
+How to work
+- Prepare and check the data, hold out a validation set, upload, and plan a small first run.
+- Before anything that costs money, show me the model, settings, estimated cost, steps, time and warnings, and
+  wait for my yes. Never start a job I have not approved.
+- Give every job a name and a tag for this experiment.
+- While jobs run, watch them (`nextron-cloud jobs progress` / `jobs events`) and tell me at once if one fails or
+  stalls.
+- When training ends, report loss and validation loss, what was charged, what you conclude, and what you would try
+  next. Download checkpoints only when I ask.
+- Keep everything you write in this project directory.
+```
+
+## Documentation
+
+The guide is written for people and coding agents alike, in [skills/nextron-cloud](skills/nextron-cloud/SKILL.md):
+
+- [SKILL.md](skills/nextron-cloud/SKILL.md): setup, rules, an experiment end to end
+- [data.md](skills/nextron-cloud/references/data.md): the sample format, what training learns from, checks
+- [training.md](skills/nextron-cloud/references/training.md): models, params, batch, cost and time, sweeps
+- [monitoring.md](skills/nextron-cloud/references/monitoring.md): states, metrics, failures, API errors
+- [checkpoints.md](skills/nextron-cloud/references/checkpoints.md): download, serve, evaluate
+- [api.md](skills/nextron-cloud/references/api.md): every SDK call and CLI command
+- [examples/compare_outputs.py](examples/compare_outputs.py): answer held-out prompts with the fine-tuned and the base
+  model side by side
+
+Claude Code: `/plugin install nextron-cloud --marketplace Physion-Labs/nextron-python` installs it as a skill.
+Other agents: point them at [AGENTS.md](AGENTS.md).

nextron_cloud-0.1.0rc1/README.md

@@ -0,0 +1,80 @@
+# nextron-cloud
+
+> **Release candidate.** The API may still change before 0.1.0; install with `--prerelease=allow`.
+
+Python SDK and CLI for [NexTron Cloud](https://nextron.physionlabs.ai): fine-tune open models with one request
+(model + dataset + settings) and download the trained checkpoint.
+
+```bash
+uv venv && uv pip install --prerelease=allow nextron-cloud   # or, in a uv project: uv add --prerelease=allow nextron-cloud
+export NEXTRON_API_KEY="nk_live_..."                           # API keys page on nextron.physionlabs.ai
+uv run nextron-cloud models                                    # what can be trained now, at what price
+```
+
+Commands below are written as `nextron-cloud ...`; run them with `uv run` (or activate `.venv`).
+
+```python
+from nextron_cloud import NexTron
+
+client = NexTron()
+dataset = client.datasets.wait(client.datasets.upload("train.jsonl").id)   # checked locally, then by the server
+draft = dict(model="qwen3.8-27b", dataset_id=dataset.id, params={"train_type": "lora", "learning_rate": 1e-4})
+print(client.jobs.estimate(**draft))                                        # cost in US cents: get a yes first
+
+job = client.jobs.wait(client.jobs.create(**draft, name="first try").id)  # only after the cost is approved
+print(job.state, client.jobs.metrics(job.id)[-1])
+
+client.jobs.wait_checkpoint(job.id)                                         # later: weights stay 30 days
+client.jobs.download_checkpoint(job.id, f"checkpoints/{job.id}")
+```
+
+## Start with an agent
+
+Install, set the key, then paste this to your coding agent (Claude Code, Codex, ...) and fill in the brackets:
+
+```text
+Run a fine-tuning experiment for me on NexTron Cloud, a managed training service: one request trains a model on
+its GPUs and leaves a checkpoint to download.
+
+Setup
+- Install the SDK with uv in this project: `uv venv && uv pip install --prerelease=allow nextron-cloud`
+  (in a uv project: `uv add --prerelease=allow nextron-cloud`). Run its commands with `uv run`, e.g.
+  `uv run nextron-cloud guide`.
+- My API key is in the environment variable NEXTRON_API_KEY. Never print it or write it into a file.
+- Before anything else run `uv run nextron-cloud guide` and read all of it. At each step read its topic too:
+  `nextron-cloud guide data`, `... training`, `... monitoring`, `... checkpoints`, `... api`. Follow its rules.
+
+The experiment
+- Goal: [what the model should do better, e.g. "answer our support tickets in our tone and format"]
+- Data: [file path or Hugging Face link, and which fields are the input and which the answer]
+- How I will judge it: [e.g. "held-out tickets, compared with the base model"]
+- Model: [a model id, or "pick one from `nextron-cloud models` and tell me why"]
+- Budget: at most $[N] for this experiment unless I approve more.
+
+How to work
+- Prepare and check the data, hold out a validation set, upload, and plan a small first run.
+- Before anything that costs money, show me the model, settings, estimated cost, steps, time and warnings, and
+  wait for my yes. Never start a job I have not approved.
+- Give every job a name and a tag for this experiment.
+- While jobs run, watch them (`nextron-cloud jobs progress` / `jobs events`) and tell me at once if one fails or
+  stalls.
+- When training ends, report loss and validation loss, what was charged, what you conclude, and what you would try
+  next. Download checkpoints only when I ask.
+- Keep everything you write in this project directory.
+```
+
+## Documentation
+
+The guide is written for people and coding agents alike, in [skills/nextron-cloud](skills/nextron-cloud/SKILL.md):
+
+- [SKILL.md](skills/nextron-cloud/SKILL.md): setup, rules, an experiment end to end
+- [data.md](skills/nextron-cloud/references/data.md): the sample format, what training learns from, checks
+- [training.md](skills/nextron-cloud/references/training.md): models, params, batch, cost and time, sweeps
+- [monitoring.md](skills/nextron-cloud/references/monitoring.md): states, metrics, failures, API errors
+- [checkpoints.md](skills/nextron-cloud/references/checkpoints.md): download, serve, evaluate
+- [api.md](skills/nextron-cloud/references/api.md): every SDK call and CLI command
+- [examples/compare_outputs.py](examples/compare_outputs.py): answer held-out prompts with the fine-tuned and the base
+  model side by side
+
+Claude Code: `/plugin install nextron-cloud --marketplace Physion-Labs/nextron-python` installs it as a skill.
+Other agents: point them at [AGENTS.md](AGENTS.md).

nextron_cloud-0.1.0rc1/pyproject.toml

@@ -0,0 +1,57 @@
+[project]
+name = "nextron-cloud"
+dynamic = ["version"]
+description = "Official Python client for the NexTron Cloud fine-tuning API."
+readme = "README.md"
+license = "Apache-2.0"
+license-files = ["LICENSE"]
+requires-python = ">=3.10"
+authors = [{ name = "Physion Labs", email = "support@physionlabs.ai" }]
+keywords = ["nextron", "physion", "fine-tuning", "training"]
+classifiers = [
+  "Development Status :: 3 - Alpha",
+  "Intended Audience :: Developers",
+  "Programming Language :: Python :: 3.10",
+  "Programming Language :: Python :: 3.11",
+  "Programming Language :: Python :: 3.12",
+  "Programming Language :: Python :: 3.13",
+  "Typing :: Typed",
+]
+# Upper bounds at the next major: an unbounded httpx once resolved to a 1.0 dev release that broke Client().
+dependencies = ["httpx>=0.27,<1", "pydantic>=2.9,<3"]
+
+[project.scripts]
+nextron-cloud = "nextron_cloud.cli:main"
+
+[project.urls]
+Homepage = "https://nextron.physionlabs.ai"
+Issues = "https://github.com/Physion-Labs/nextron-python/issues"
+
+[dependency-groups]
+dev = ["pytest>=8.0", "mypy>=1.13"]
+
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[tool.hatch.version]
+path = "src/nextron_cloud/_version.py"
+
+[tool.hatch.build.targets.sdist]
+only-include = ["src", "skills", "README.md", "LICENSE", "CHANGELOG.md", "pyproject.toml"]
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/nextron_cloud"]
+
+# The agent guide ships inside the package: `nextron-cloud guide` prints it after a plain pip install.
+[tool.hatch.build.targets.wheel.force-include]
+"skills/nextron-cloud" = "nextron_cloud/guide"
+
+[tool.pytest.ini_options]
+testpaths = ["tests"]
+addopts = "-q"
+
+[tool.mypy]
+python_version = "3.10"
+strict = true
+files = ["src/nextron_cloud"]

nextron_cloud-0.1.0rc1/skills/nextron-cloud/SKILL.md

@@ -0,0 +1,115 @@
+---
+name: nextron-cloud
+description: Run fine-tuning experiments on NexTron Cloud with the `nextron-cloud` Python SDK or CLI — choose a model, prepare and check JSONL training data, upload or import datasets from Hugging Face, estimate cost, start and watch training jobs, compare runs, download checkpoints and evaluate them. Use whenever the user wants to fine-tune, train, SFT or LoRA a model on NexTron, prepare data for it, check on a NexTron job, or run a training experiment.
+---
+
+# NexTron Cloud
+
+NexTron Cloud fine-tunes open models on its own H100s. One request = one whole training run: you send a model, a
+dataset and a few settings, the service trains and leaves a checkpoint (a Hugging Face model directory) to download.
+There is no per-step API, and nothing heavy runs on the user's machine. Text-only supervised fine-tuning (full or
+LoRA) today.
+
+Everything below works from Python (`from nextron_cloud import NexTron`) or the shell (`nextron-cloud ...`, every
+command prints JSON). Read the reference for each step when you reach it (`nextron-cloud guide <topic>` prints it,
+e.g. `nextron-cloud guide data`; this page is `nextron-cloud guide`):
+
+| step | read |
+|---|---|
+| choose a model, settings, cost and time; run a sweep | [references/training.md](references/training.md) |
+| build, convert and check data; what is trained on | [references/data.md](references/data.md) |
+| watch a job, read its metrics, fix a failure | [references/monitoring.md](references/monitoring.md) |
+| download, serve and evaluate the checkpoint | [references/checkpoints.md](references/checkpoints.md) |
+| every SDK call and CLI command | [references/api.md](references/api.md) |
+
+## Setup
+
+```bash
+uv venv && uv pip install --prerelease=allow nextron-cloud   # in a uv project: uv add --prerelease=allow nextron-cloud
+export NEXTRON_API_KEY="nk_live_..."
+uv run nextron-cloud me              # works = ready; prints the organisation and its balance in US cents
+```
+
+Use uv, not plain pip. Commands in this guide are written `nextron-cloud ...`: run them as `uv run nextron-cloud ...`
+(Python as `uv run python ...`), or activate `.venv` first.
+
+No key yet: accounts are by invitation (ask Physion Labs, support@physionlabs.ai). The owner signs in at
+https://nextron.physionlabs.ai, opens **API keys** (only an organisation owner can create one) and adds credit under
+**Billing**. Never print the key or write it into files you commit.
+
+## Rules
+
+1. **Jobs cost the owner money. Ask before starting one.** Run `jobs.estimate` with the exact arguments you will
+   create with, tell the owner the cost (`credits` / 100 = US dollars), the model and the plan, and wait for a yes.
+   The CLI enforces this: `nextron-cloud jobs create ...` without `--yes` prints the estimate and starts nothing.
+2. **Take models and limits from `client.models()`**, never from memory or this file: model ids, which methods each
+   trains (`full`, `lora`), the longest sample per method, prices, GPU setups and whether one is free now.
+   `jobs.estimate` runs the same checks as creating the job (it raises with the reason when the job would be refused)
+   and returns the settings, about how many steps, and warnings: show those to the owner with the price.
+3. **Never start the same job twice.** `jobs.create` retries safely with one idempotency key. If your own process
+   died mid-create, pass the same `idempotency_key=` (CLI `--idempotency-key`) or check `jobs.list()` first.
+4. **Check data before uploading** (`check_file`, CLI `datasets check`) and read what the server reports after its
+   check (`dataset.errors`, `dataset.tokens`).
+5. **Label every job**: `name=` says what the run tests, `tags=` groups runs (a sweep tags its jobs `sweep:<name>`).
+   `jobs.list()` returns the whole organisation's jobs, other members' included; `jobs.list(tag=...)` finds yours.
+   `nextron-cloud report --tag ...` compares them.
+6. **Training and the checkpoint are separate steps.** `jobs.wait` returns when training is over; read the results
+   then. The checkpoint uploads afterwards (minutes to an hour by model size) and stays downloadable for 30 days:
+   `jobs.wait_checkpoint` / `download_checkpoint` only when the owner needs the weights.
+7. **Poll slowly** (jobs every 30 s or more) and do not leave a job unwatched without telling the owner its id.
+8. Text only: samples with images, audio or video are refused.
+
+## An experiment, end to end
+
+```python
+from nextron_cloud import NexTron
+from nextron_cloud.data import prepare_file
+
+client = NexTron()
+
+# 1. Model: pick one that trains the method and length the data needs, and is free or has a short queue.
+for m in client.models():
+    print(m.id, {k: (v.max_seq_len, v.price_per_mtok) for k, v in m.methods.items()},
+          {l: (x.available, x.queued) for l, x in m.layouts.items()})
+
+# 2. Data: convert, check, hold out a validation set, upload.
+prepare_file("tickets.csv", "data", prompt_column="ticket", response_column="reply")   # data/train.jsonl, data/val.jsonl
+train = client.datasets.wait(client.datasets.upload("data/train.jsonl").id)
+val = client.datasets.wait(client.datasets.upload("data/val.jsonl").id)
+assert train.state == val.state == "ready", (train.errors, val.errors)
+
+# 3. Cost, then the owner's yes.
+draft = dict(model="qwen3.8-27b", dataset_id=train.id,
+             params={"train_type": "lora", "learning_rate": 1e-4, "num_train_epochs": 2, "val_dataset_id": val.id})
+quote = client.jobs.estimate(**draft)                 # quote.credits / 100 dollars; quote.enough
+
+# 4. Train, watch, read the results. client.jobs.progress(job.id) says at any time where it is and how long is left.
+job = client.jobs.create(**draft, name="support lora lr=1e-4 2ep", tags=["support-v1"])
+job = client.jobs.wait(job.id)                        # or: for line in client.jobs.logs(job.id, follow=True)
+print(job.state, job.error, job.notices, client.jobs.metrics(job.id)[-1], client.jobs.billing(job.id).charged_credits)
+
+# 5. Later, when the owner wants the weights (kept 30 days).
+client.jobs.wait_checkpoint(job.id)
+client.jobs.download_checkpoint(job.id, f"checkpoints/{job.id}")
+```
+
+The same from the shell:
+
+```bash
+nextron-cloud models
+nextron-cloud datasets prepare tickets.csv --out-dir data --prompt-column ticket --response-column reply
+nextron-cloud datasets upload data/train.jsonl            # waits for the server's check
+nextron-cloud jobs create --model qwen3.8-27b --dataset ds_... --param train_type=lora --param learning_rate=1e-4 \
+    --name "support lora" --tag support-v1
+# -> prints the estimate (exit code 3); after the owner agrees, the same command with --yes
+nextron-cloud jobs progress job_...                       # phase, queue, ETA, upload bytes
+nextron-cloud jobs events job_... --follow                # state changes, steps, validation, checkpoint stages
+nextron-cloud sweep --name lr --model qwen3.8-27b --dataset ds_... --param train_type=lora \
+    --grid learning_rate=5e-5,1e-4,2e-4                    # estimates; --yes starts all of them
+nextron-cloud report --tag sweep:lr --table
+nextron-cloud jobs wait-checkpoint job_... && nextron-cloud jobs download job_... checkpoints/job_...
+```
+
+A first run on new data should be a small one: `max_steps: 3` on the smallest GPU setup proves the data and settings
+work for a fraction of the cost. Datasets of a few thousand short samples are smaller than one training step; read
+"Small datasets" in training.md before planning them.