chore: pin bun version and update lockfile (#126)

* chore: pin bun version and update lockfile

* docs(proxies): clarify bandwidth measurement language

* docs: update magnitude url and replit documentation link

Author: fukouda

.github/workflows/changelog-draft.yml

@@ -68,7 +68,9 @@ jobs:
         if: steps.gate.outputs.should_run == 'true'
         uses: oven-sh/setup-bun@v2
         with:
-          bun-version: latest
+          # Keep in step with ci.yml: an unpinned bun changes the lockfile
+          # version and breaks --frozen-lockfile with no repo change.
+          bun-version: 1.4.2
 
       - name: Install dependencies
         if: steps.gate.outputs.should_run == 'true'

.github/workflows/ci.yml

@@ -23,7 +23,9 @@ jobs:
       - name: Set up Bun
         uses: oven-sh/setup-bun@v2
         with:
-          bun-version: latest
+          # Pinned: `latest` silently moved to a bun that writes a different
+          # lockfile version, breaking --frozen-lockfile with no repo change.
+          bun-version: 1.4.2
 
       - name: Install dependencies
         run: bun install --frozen-lockfile

bun.lock

@@ -115,7 +115,7 @@ A full run takes ~45 seconds. The `finally` block stops the agent first, then re
 
 ## Related
 
-[Magnitude docs](https://docs.magnitude.run)
+[Magnitude docs](https://docs.magnitude.dev)
 
 ## Related recipes
 

content/docs/cookbook/magnitude.mdx

@@ -61,6 +61,6 @@ Anthropic is Magnitude's default planning provider — the example configures `l
 
 ### Resources
 
-*   [Magnitude documentation](https://docs.magnitude.run/) – Concepts, agent APIs, and examples
+*   [Magnitude documentation](https://docs.magnitude.dev/) – Concepts, agent APIs, and examples
 *   [Steel Sessions API reference](/api-reference) – Programmatic session control for Steel browsers
 *   [Steel Discord](https://discord.gg/steel-dev) – Get help and share what you build

content/docs/integrations/magnitude.mdx

@@ -13,7 +13,7 @@ A good fit for prototyping, scheduled jobs, and sharing runnable examples.
 
 *   **Steel API Key**: Any plan; get a free key [here](https://app.steel.dev/settings/api-keys)
 *   **Replit account**: Free tier available
-*   **Languages**: Python and Node.js (full list of supported languages [here](https://replit.com/templates/languages))
+*   **Languages**: Python and Node.js (see the [Replit docs](https://docs.replit.com/) for everything else Replit runs)
 
 ### Starter Templates
 

content/docs/integrations/replit.mdx

@@ -262,6 +262,10 @@ Reach out to us on the <span className="font-bold">#help</span> channel on [Disc
 
 Steel-managed residential proxies are billed per GB of usage, while the default behavior (no proxy, Steel's datacenter IPs) is free and included in all plans. Bring Your Own Proxy carries no charge from Steel — you only pay your own proxy provider.
 
+### Why is my proxy bandwidth higher than the content I downloaded?
+
+Bandwidth is measured on the wire at the proxy, so it covers every byte the session moves in both directions: TLS handshakes, request and response headers, and every file the page loads, not just the response bodies your code reads. That overhead is normally a small percentage. The bigger difference is usually traffic the browser asked for but never used. If a download is still in flight when the session is released, or a page starts fetching a large asset before you navigate away, those bytes have already crossed the proxy and count even though nothing consumed them. Block what you don't need with [`optimizeBandwidth`](/overview/sessions-api/configuration#resource-control) or `blockAds`, and release sessions once you have what you came for.
+
 ### Do Steel sessions use a proxy by default?
 
 No — proxies are disabled by default (`useProxy: false` is the implicit setting), so traffic originates from Steel's own datacenter IPs at no proxy-bandwidth cost. Set `useProxy: true` to route through a Steel-managed residential IP (US-based by default).