style: apply /style-guide pass to serverless-rl (#2667)

## Summary

This PR applies the `/style-guide` skill (Google Developer Style Guide +
CoreWeave conventions) to the `serverless-rl` section. The run was fully
automated; only style-level edits were made.

## Files edited

- `serverless-rl/api-reference.mdx`
- `serverless-rl/prerequisites.mdx`
- `serverless-rl/sft.mdx`
- `serverless-rl/sft/usage.mdx`
- `serverless-rl/usage-limits.mdx`
- `serverless-rl/usage.mdx`
- `serverless-rl/use-trained-models.mdx`

## Recommendations for technical review

### Prerequisites
- Confirm whether a brief link to `/serverless-rl/prerequisites` should
appear at the top of the Authentication section in `api-reference.mdx`,
not only under Related resources.
- In `prerequisites.mdx`, confirm whether the `ApiKeyCreate` snippet
covers W&B account sign-up, or whether a separate sign-up
link/instruction is needed so the page is self-contained.
- In `sft.mdx`, verify whether ART-specific prerequisites (Python
version, OpenPipe account, ART package install) belong on the page or in
the linked prerequisites doc.
- In `sft/usage.mdx`, consider adding a Prerequisites section listing
the required W&B account, access to Serverless Inference, Models,
Artifacts, optional Weave, and the ART library so the page is
self-contained.
- In `usage.mdx`, confirm whether tutorial-style inline prerequisites
should be present per CoreWeave tutorial guidelines, rather than only
linking out.
- In `use-trained-models.mdx`, consider a labeled Prerequisites section
covering a completed training run, a known saved step, and any
IAM/permissions scope required on the W&B API key; also introduce
`WANDB_API_KEY` before its first use in the cURL example.
- Across pages, confirm that key terms (Serverless RL, LoRA, LoRA
adapter, checkpoint, base model, trajectory, reward function,
post-train, ART framework) are defined on linked prerequisite pages or
add short inline definitions/links.

### Verification steps
- In `api-reference.mdx`, consider a short cURL example against `GET
/v1/health` or `GET /v1/system-check` so readers can confirm credentials
work.
- In `prerequisites.mdx`, consider a verification cue confirming the API
key works and the project was created before "Next steps."
- In `sft/usage.mdx`, confirm whether a brief "what you'll have
accomplished" / expected-outcome sentence is appropriate after training
completes (e.g., a deployed LoRA available via Serverless Inference and
stored as an artifact).
- In `usage.mdx`, consider adding an in-doc "what success looks like"
statement after the linked ART quickstart.
- In `use-trained-models.mdx`, add a short "expected result" note after
each example (response shape, status, or how to confirm the model is
serving).

### Technical accuracy
- **`api-reference.mdx`** — `/v1/chat/completions` and
`/v1/chat/completions/` both link to "Create Chat Completion" (slugs
`create-chat-completion-1` and `create-chat-completion`); likely a
duplicate from the OpenAPI spec — needs SME input on whether one should
be hidden.
- **`api-reference.mdx`** — Auto-generated bullet titles render as
"Create Sft Training Job" / "Create Rl Training Job"; should be "Create
SFT Training Job" / "Create RL Training Job" (upstream OpenAPI fix).
- **`api-reference.mdx`** — The bullet for `POST
/v1/preview/models/{model_id}/log` is just "Log", which is ambiguous;
the OpenAPI summary should be clarified.
- **`api-reference.mdx`** — Endpoint paths mix `/v1/...` (health, chat)
with `/v1/preview/...` (models, training-jobs); consider whether
"preview" status should be called out.
- **`prerequisites.mdx`** — Confirm the link
`https://docs.wandb.ai/models/track/project-page` is the canonical
"Projects guide" URL (path suggests "Project page," which may not match
the displayed link text). Also confirm `/serverless-rl/api-reference`
resolves to the intended destination.
- **`prerequisites.mdx` (line 6)** — "Serverless RL **currently**
supports the following foundation models for training." Confirm intent:
a list as of publication (remove "currently") or a live catalog (remove
as redundant with the auto-generation note).
- **`sft.mdx`** — Confirm
`https://art.openpipe.ai/fundamentals/sft-training` is the canonical,
current Serverless SFT entry point, and that
`https://art.openpipe.ai/getting-started/about` is the correct ART
"about" landing.
- **`sft/usage.mdx`** — "Save them locally or to a third party for
backup" — "third party" is vague; name supported options or link to a
reference.
- **`usage-limits.mdx`** — Verify "Training is free during the public
preview period" (line 20) is still accurate and plan how the page should
be updated when the preview ends.
- **`usage-limits.mdx`** — Confirm the 2,000 / 6,000 concurrency numbers
(line 30) are current, and that `support@wandb.com` is still the right
contact for limit-increase requests.
- **`usage-limits.mdx`** — Verify the "5 GB / ~30 LoRAs" model-storage
estimate (line 24) is current.
- **`usage.mdx`** — Verify the Google Colab notebook URL
(`openpipe/art-notebooks/.../2048/2048.ipynb`) still resolves to the
intended 2048 example. Confirm whether "Serverless RL" should remain as
a product name or get a formal first-use expansion ("Serverless
reinforcement learning (RL)").
- **`use-trained-models.mdx`** — Slash inconsistency in
`wandb-artifact://` scheme: the cURL example (line 41) uses
`wandb-artifact://[ENTITY]/...` (two slashes), while the schema (line
16) and Python example (line 66) use `wandb-artifact:///` (three
slashes). Confirm the accepted form and standardize.
- **`use-trained-models.mdx`** — Step prefix ambiguity: schema uses
`:step25`, Python uses `:step100`, cURL uses `[STEP]` without clarifying
whether `step` is part of the value. Confirm and document.
- **`use-trained-models.mdx` (line 21)** — "Your W&B entity's (team)
name" reads awkwardly; confirm whether "entity" and "team" are
synonymous in this context and rephrase consistently.

### Missing content
- **`api-reference.mdx`** — No mention of rate limits, pagination, error
response format, versioning policy, or SDKs/client libraries; the
"OpenAI-compatible" claim for chat completions isn't elaborated (which
OpenAI client configurations work and any deviations).
- **`prerequisites.mdx`** — Clarify whether `mailto:support@wandb.ai` is
for (a) requesting unlisted models, (b) requesting access/onboarding for
listed models, or (c) both. Consider a brief legend explaining "Type,"
"Context Window," and "Parameters" (especially `Active-Total` MoE
notation). The HTML comment about auto-generation is invisible to
readers — consider a visible "Last updated" or "Source" note.
- **`prerequisites.mdx` ("Create a project in W&B")** — Consider noting
whether the project name must match anything referenced later in ART or
API configuration, and whether the section should be an explicit
single-step procedure or merged with the API-key step.
- **`prerequisites.mdx` ("Next steps")** — Only two items; consider
additional follow-on resources (tutorial, examples repo,
pricing/limits).
- **`sft.mdx`** — Indicate which models or model families are supported
via Serverless SFT and any scale/quota considerations distinct from
Serverless RL; add a brief statement of how Serverless SFT relates to
Serverless RL for readers landing from search.
- **`sft/usage.mdx`** — Consider a high-level overview of the workflow
(what the user does in ART, what comes back into W&B) so readers know
what they're committing to before leaving the page. Define or link
"checkpoint" on first use.
- **`usage-limits.mdx`** — Geographic restrictions bullet (line 32)
links to the Terms of Service without indicating whether an in-docs list
of supported/unsupported regions exists; add one if available. Consider
whether `429` concurrency-limit behavior warrants a `<Note>` or
`<Warning>` callout rather than inline prose.
- **`usage.mdx`** — No "Resources" or "Related" section; if more
references are added later (Colab notebooks, API reference, blog posts),
consider reintroducing it, provided the inline links in the intro are
restructured to avoid duplication. "Post-train" and "ART framework" are
referenced without on-page definition (ART is linked; "post-train" may
warrant a glossary or concept link).
- **`use-trained-models.mdx`** — No mention of how to find the entity,
project, model name, or step values from the W&B UI or API; link to the
relevant reference. Consider adding a parent H2 (such as "Make an
inference request") with the two examples as H3 children, plus
procedural subheadings ("Construct the endpoint", "Send a request") for
scannability. Standardize the Python example to bracket-style uppercase
placeholders so both examples teach the same naming convention.

## How to review

- Each file's changes are style edits only. Compare side-by-side and
flag any that change technical meaning.
- Approve and merge to accept the edits, or close to reject them.

Author: johndmulhausen

serverless-rl/api-reference.mdx

@@ -1,52 +1,56 @@
 ---
 title: API overview
 description: "Browse the Serverless RL API endpoints for chat completions, models, training jobs, and health checks with authentication details."
+keywords: ["OpenAI-compatible", "create training job", "list model checkpoints", "SFT training job", "reinforcement learning API"]
 ---
 
 <Note>
-The Serverless RL API provides endpoints for managing and interacting with training jobs, including serverless reinforcement learning (RL) and supervised fine-tuning (SFT). The API is OpenAI-compatible for chat completions.
+The Serverless RL API provides endpoints for managing training jobs, including serverless reinforcement learning (RL) and supervised fine-tuning (SFT). The API is OpenAI-compatible for chat completions.
 </Note>
 
+Use this reference to look up authentication requirements, the base URL, and the available endpoints for chat completions, models, training jobs, and health checks.
+
 ## Authentication
 
 All API requests require authentication using your W&B API key. Create an API key at [wandb.ai/settings](https://wandb.ai/settings).
 
 Include your API key in the `Authorization` header:
 
-```
+```http
 Authorization: Bearer YOUR_API_KEY
 ```
 
 ## Base URL
 
-```
+```text
 https://api.training.wandb.ai/v1
 ```
 
 ## Available endpoints
 
+The following sections describe the available endpoints, grouped by resource type.
 
-### chat-completions
+### Chat completions
 
 - **[POST /v1/chat/completions](https://docs.wandb.ai/serverless-rl/api-reference/chat-completions/create-chat-completion-1)** - Create Chat Completion
 - **[POST /v1/chat/completions/](https://docs.wandb.ai/serverless-rl/api-reference/chat-completions/create-chat-completion)** - Create Chat Completion
 
-### models
+### Models
 
 - **[POST /v1/preview/models](https://docs.wandb.ai/serverless-rl/api-reference/models/create-model)** - Create Model
 - **[DELETE /v1/preview/models/{model_id}](https://docs.wandb.ai/serverless-rl/api-reference/models/delete-model)** - Delete Model
 - **[DELETE /v1/preview/models/{model_id}/checkpoints](https://docs.wandb.ai/serverless-rl/api-reference/models/delete-model-checkpoints)** - Delete Model Checkpoints
 - **[GET /v1/preview/models/{model_id}/checkpoints](https://docs.wandb.ai/serverless-rl/api-reference/models/list-model-checkpoints)** - List Model Checkpoints
 - **[POST /v1/preview/models/{model_id}/log](https://docs.wandb.ai/serverless-rl/api-reference/models/log)** - Log
 
-### training-jobs
+### Training jobs
 
 - **[POST /v1/preview/sft-training-jobs](https://docs.wandb.ai/serverless-rl/api-reference/training-jobs/create-sft-training-job)** - Create Sft Training Job
 - **[POST /v1/preview/training-jobs](https://docs.wandb.ai/serverless-rl/api-reference/training-jobs/create-rl-training-job)** - Create Rl Training Job
 - **[GET /v1/preview/training-jobs/{training_job_id}](https://docs.wandb.ai/serverless-rl/api-reference/training-jobs/get-training-job)** - Get Training Job
 - **[GET /v1/preview/training-jobs/{training_job_id}/events](https://docs.wandb.ai/serverless-rl/api-reference/training-jobs/get-training-job-events)** - Get Training Job Events
 
-### health
+### Health
 
 - **[GET /v1/health](https://docs.wandb.ai/serverless-rl/api-reference/health/health-check)** - Health Check
 - **[GET /v1/system-check](https://docs.wandb.ai/serverless-rl/api-reference/health/system-check)** - System Check

serverless-rl/prerequisites.mdx

@@ -1,6 +1,7 @@
 ---
 title: Prerequisites
 description: "Set up your environment for Serverless RL by creating an account, generating an API key, and configuring a project."
+keywords: ["OpenPipe ART", "wandb login", "install wandb", "ART framework setup"]
 ---
 
 import ApiKeyCreate from "/snippets/_includes/api-key-create.mdx";
@@ -13,17 +14,17 @@ Before starting, review the [usage information and limits](/serverless-rl/usage-
 
 ## Sign up and create an API key
 
-To authenticate your machine with W&B, you must first generate an API key.
+If you don't already have a W&B account, sign up before continuing. To authenticate your machine with W&B, you must first generate an API key.
 
 <ApiKeyCreate/>
 
 ## Create a project in W&B
 
-Create a project in your W&B account to track usage, record training metrics, and save trained models. See the [Projects guide](https://docs.wandb.ai/models/track/project-page) for more information.
+Create a project in your W&B account to track usage, record training metrics, and save trained models. For more information, see the [Projects guide](https://docs.wandb.ai/models/track/project-page).
 
 ## Next steps
 
-After completing the prerequisites:
+After you complete the prerequisites, try the following resources:
 
-* Check the [API reference](/serverless-rl/api-reference) to learn about available endpoints
-* Try the [ART quickstart](https://art.openpipe.ai/getting-started/quick-start)
+* Check the [API reference](/serverless-rl/api-reference) to learn about available endpoints.
+* Try the [ART quickstart](https://art.openpipe.ai/getting-started/quick-start).

serverless-rl/sft.mdx

@@ -1,34 +1,37 @@
 ---
 title: Serverless SFT
 description: Learn how to fine-tune models using supervised fine-tuning (SFT) on W&B
+keywords: ["LoRA adapter", "distillation", "warmup before RL", "ART framework", "managed training cluster"]
 ---
 
-Use Serverless SFT to fine-tune LLMs with supervised learning on curated datasets. Serverless SFT is now in public preview. W&B provisions the training infrastructure ([on CoreWeave](https://docs.coreweave.com/docs/platform)) for you while allowing full flexibility in your environment's setup. You get instant access to a managed training cluster that elastically auto-scales to handle your training workloads.
+Use Serverless SFT to fine-tune LLMs with supervised learning on curated datasets. Serverless SFT is in public preview. W&B provisions the training infrastructure ([on CoreWeave](https://docs.coreweave.com/docs/platform)) for you and gives you full flexibility to set up your environment. You get instant access to a managed training cluster that auto-scales to handle your training workloads.
 
-Serverless SFT is ideal for tasks like:
-* **Distillation**: Transferring knowledge from a larger, more capable model into a smaller, faster one
-* **Teaching output style and format**: Training a model to follow specific response formats, tone, or structure
-* **Warmup before RL**: Pre-training a model with supervised examples before applying reinforcement learning for further refinement
+Serverless SFT is ideal for tasks such as:
+* **Distillation**: Transferring knowledge from a larger, more capable model into a smaller, faster one.
+* **Teaching output style and format**: Training a model to follow specific response formats, tone, or structure.
+* **Warmup before RL**: Pre-training a model with supervised examples before applying reinforcement learning for further refinement.
 
 Serverless SFT trains low-rank adapters (LoRAs) to specialize a model for your specific task. W&B automatically stores the LoRAs you train as artifacts in your account. You can also save them locally or to a third party for backup. Serverless Inference also automatically hosts models that you train through Serverless SFT.
 
-See the ART [Serverless SFT docs](https://art.openpipe.ai/fundamentals/sft-training) to get started.
+To begin training a model with Serverless SFT, see the ART [Serverless SFT docs](https://art.openpipe.ai/fundamentals/sft-training).
 
-## Why Serverless SFT?
+## Why Serverless SFT
 
 Supervised fine-tuning (SFT) is a training technique where a model learns from curated input-output examples. Serverless SFT on W&B provides the following advantages:
 
-* **Lower training costs**: By multiplexing shared infrastructure across many users, skipping the setup process for each job, and scaling your GPU costs down to 0 when you're not actively training, Serverless SFT reduces training costs significantly.
-* **Faster training time**: By immediately provisioning training infrastructure when you need it, Serverless SFT speeds up your training jobs and lets you iterate faster.
-* **Automatic deployment**: Serverless SFT automatically deploys every checkpoint you train, so you do not need to manually set up hosting infrastructure. You can access and test trained models immediately in local, staging, or production environments.
+* **Lower training costs**: Serverless SFT multiplexes shared infrastructure across many users, skips the setup process for each job, and scales your GPU costs down to zero when you aren't actively training. This reduces training costs.
+* **Faster training time**: Serverless SFT immediately provisions training infrastructure when you need it. This speeds up your training jobs and lets you iterate faster.
+* **Automatic deployment**: Serverless SFT automatically deploys every checkpoint you train, so you don't need to manually set up hosting infrastructure. You can access and test trained models immediately in local, staging, or production environments.
 
 ## How Serverless SFT uses W&B services
 
 Serverless SFT uses a combination of the following W&B components to operate:
 
-* [Inference](/inference): To run your models
-* [Models](/models): To track performance metrics during the LoRA adapter's training
-* [Artifacts](/models/artifacts): To store and version the LoRA adapters
-* [Weave (optional)](/weave): To gain observability into how the model responds at each step of the training loop
+* [Inference](/inference): To run your models.
+* [Models](/models): To track performance metrics during the LoRA adapter's training.
+* [Artifacts](/models/artifacts): To store and version the LoRA adapters.
+* [Weave](/weave) (Optional): To gain observability into how the model responds at each step of the training loop.
 
-Serverless SFT is in public preview. During the preview, W&B charges you only for inference usage and artifact storage. W&B does not charge for adapter training during the preview period.
+<Note>
+Serverless SFT is in public preview. During the preview, W&B charges you only for inference usage and artifact storage. W&B doesn't charge for adapter training during the preview period.
+</Note>

serverless-rl/sft/usage.mdx

@@ -1,9 +1,11 @@
 ---
 title: How to use Serverless SFT
 description: "Fine-tune models with Serverless SFT using OpenPipe's ART framework and the Serverless RL API for supervised learning."
+keywords: ["run SFT job", "ART SFT training", "SFT tutorial"]
 ---
 
-Use Serverless SFT through [OpenPipe's ART framework](https://art.openpipe.ai/getting-started/about) and the [Serverless RL API](/serverless-rl/api-reference).
+Use Serverless supervised fine-tuning (SFT) through [OpenPipe's ART framework](https://art.openpipe.ai/getting-started/about) and the Serverless RL API.
 
-To start using Serverless SFT, satisfy the [prerequisites](/serverless-rl/prerequisites) to use W&B tools, and then go through the ART [Serverless SFT docs](https://art.openpipe.ai/fundamentals/sft-training).
-- To learn about Serverless SFT's API endpoints, see the [Serverless RL API reference](/serverless-rl/api-reference).
+To start using Serverless SFT, satisfy the [prerequisites](/serverless-rl/prerequisites) for W&B tools, and then see the [ART Serverless SFT documentation](https://art.openpipe.ai/fundamentals/sft-training).
+
+To learn about Serverless SFT's API endpoints, see the [Serverless RL API reference](/serverless-rl/api-reference).

serverless-rl/usage-limits.mdx

@@ -1,15 +1,18 @@
 ---
 title: Usage information and limits
 description: Understand pricing, usage limits, and account restrictions for W&B Serverless RL
+keywords: ["concurrency limits", "geographic restrictions", "LoRA checkpoint storage", "rate limit"]
 ---
 
+This page describes the pricing model, concurrency limits, and geographic restrictions that apply to W&B Serverless RL. Review this information to estimate costs and to understand the constraints that affect how you run training and inference workloads.
+
 ## Pricing
 
-Pricing has three components: inference, training, and storage. For specific billing rates, visit our [pricing page](https://site.wandb.ai/pricing/training).
+Pricing has three components: inference, training, and storage. For specific billing rates, visit our [pricing page](https://site.wandb.ai/pricing/training). The following sections describe each component.
 
 ### Inference
 
-Pricing for Serverless RL inference requests matches Serverless Inference pricing. See [model-specific costs](https://site.wandb.ai/pricing/reinforcement-learning) for more details. Learn more about purchasing credits, account tiers, and usage caps in the [Serverless Inference docs](/inference/usage-limits#purchase-more-credits).
+Pricing for Serverless RL inference requests matches Serverless Inference pricing. See [model-specific costs](https://site.wandb.ai/pricing/reinforcement-learning). Learn more about purchasing credits, account tiers, and usage caps in the [Serverless Inference docs](/inference/usage-limits#purchase-more-credits).
 
 ### Training
 
@@ -19,10 +22,12 @@ Training is free during the public preview period.
 
 ### Model storage
 
-Serverless RL stores checkpoints of your trained LoRAs so you can evaluate, serve, or continue training them at any time. W&B bills storage monthly based on total checkpoint size and your [pricing plan](https://wandb.ai/site/pricing). Every plan includes at least 5GB of free storage, which is enough for roughly 30 LoRAs. Delete low-performing LoRAs to save space. See the [ART SDK](https://art.openpipe.ai/features/checkpoint-deletion) for instructions.
+Serverless RL stores checkpoints of your trained LoRAs so you can evaluate, serve, or continue training them at any time. W&B bills storage monthly based on total checkpoint size and your [pricing plan](https://wandb.ai/site/pricing). Every plan includes at least 5 GB of free storage, which is enough for roughly 30 LoRAs. To save space, delete low-performing LoRAs. See the [ART SDK](https://art.openpipe.ai/features/checkpoint-deletion) for instructions.
 
 ## Limits
 
-* **Inference concurrency limits**: By default, Serverless RL currently supports up to 2000 concurrent requests per user and 6000 per project. If you exceed your rate limit, the Inference API returns a `429 Concurrency limit reached for requests` response. To avoid this error, reduce the number of concurrent requests your training job or production workload makes at once. If you need a higher rate limit, you can request one at support@wandb.com.
+The following limits apply to Serverless RL usage. Review them when sizing workloads or when planning to use the service from a new region.
+
+* **Inference concurrency limits**: By default, Serverless RL supports up to 2,000 concurrent requests per user and 6,000 per project. If you exceed your rate limit, the Inference API returns a `429 Concurrency limit reached for requests` response. To avoid this error, reduce the number of concurrent requests your training job or production workload makes at once. If you need a higher rate limit, request one at support@wandb.com.
 
 * **Geographic restrictions**: Serverless RL is only available in supported geographic locations. For more information, see the [Terms of Service](https://site.wandb.ai/terms/).

serverless-rl/usage.mdx

@@ -1,10 +1,11 @@
 ---
 title: How to use Serverless RL
 description: "Post-train models with Serverless RL using OpenPipe's ART framework and the Serverless RL API for reinforcement learning."
+keywords: ["Colab notebook", "ART quickstart", "post-training tutorial"]
 ---
 
 Use Serverless RL through [OpenPipe's ART framework](https://art.openpipe.ai/getting-started/about) and the [Serverless RL API](/serverless-rl/api-reference).
 
-To start using Serverless RL, satisfy the [prerequisites](/serverless-rl/prerequisites) to use W&B tools, and then go through the ART [quickstart](https://art.openpipe.ai/getting-started/quick-start).
-- For code examples and workflows, see the [Google Colab notebook](https://colab.research.google.com/github/openpipe/art-notebooks/blob/main/examples/2048/2048.ipynb).
-- To learn about Serverless RL's API endpoints, see the [Serverless RL API reference](/serverless-rl/api-reference).
+To start using Serverless RL, you must satisfy the [prerequisites](/serverless-rl/prerequisites) to use W&B tools, and then go through the ART [quickstart](https://art.openpipe.ai/getting-started/quick-start).
+
+For code examples and workflows, see the [Google Colab notebook](https://colab.research.google.com/github/openpipe/art-notebooks/blob/main/examples/2048/2048.ipynb).

serverless-rl/use-trained-models.mdx

@@ -1,44 +1,45 @@
 ---
 title: Use your trained models
 description: Make inference requests to the models you've trained
+keywords: ["wandb-artifact", "model endpoint", "OpenAI SDK", "cURL", "checkpoint step"]
 ---
 
-After you train a model with Serverless RL, it is automatically available for inference.
+After you train a model with Serverless RL, it's automatically available for inference. This page shows you how to construct the endpoint for a trained model and send inference requests to it. Use this endpoint to integrate your model into your application or evaluation workflows.
 
-To send requests to your trained model, you need:
+To send requests to your trained model, you need the following:
 * Your [W&B API key](https://wandb.ai/settings)
 * The [Serverless RL API's](/serverless-rl/api-reference) base URL, `https://api.training.wandb.ai/v1/`
 * Your model's endpoint
 
 The model's endpoint uses the following schema:
 
-```
-wandb-artifact:///<entity>/<project>/<model-name>:<step>
+```text
+wandb-artifact:///[ENTITY]/[PROJECT]/[MODEL-NAME]:[STEP]
 ```
 
 The schema consists of:
 
 * Your W&B entity's (team) name
 * The name of the project associated with your model
 * The trained model's name
-* The training step of the model you want to deploy (this is usually the step where the model performed best in your evaluations)
+* The training step of the model you want to deploy. This is usually the step where the model performed best in your evaluations.
 
-For example, if your W&B team is named `email-specialists`, your project is called `mail-search`, your trained model is named `agent-001`, and you wanted to deploy it on step 25, the endpoint looks like this:
+For example, if your W&B team is named `email-specialists`, your project is called `mail-search`, your trained model is named `agent-001`, and you want to deploy it on step 25, the endpoint looks like this:
 
-```
+```text
 wandb-artifact:///email-specialists/mail-search/agent-001:step25
 ```
 
-Once you have your endpoint, you can integrate it into your normal inference workflows. The following examples show how to make inference requests to your trained model using a cURL request or the [Python OpenAI SDK](https://github.com/openai/openai-python).
+After you have your endpoint, you can integrate it into your normal inference workflows. The following examples show how to make inference requests to your trained model using a cURL request or the [Python OpenAI SDK](https://github.com/openai/openai-python). Choose the example that matches your environment.
 
-### cURL
+## cURL
 
-```shell
+```bash
 curl https://api.training.wandb.ai/v1/chat/completions \
     -H "Authorization: Bearer $WANDB_API_KEY" \
     -H "Content-Type: application/json" \
     -d '{
-            "model": "wandb-artifact://<entity>/<project>/<model-name>:<step>",
+            "model": "wandb-artifact://[ENTITY]/[PROJECT]/[MODEL-NAME]:[STEP]",
             "messages": [
             {"role": "system", "content": "You are a helpful assistant."},
             {"role": "user", "content": "Summarize our training run."}
@@ -48,7 +49,7 @@ curl https://api.training.wandb.ai/v1/chat/completions \
         }'
 ```
 
-### OpenAI SDK
+## OpenAI SDK
 
 ```python
 from openai import OpenAI