fix(snapshot): reclaim covered WAL bytes

beava-website/project/docs/concepts/streams/index.html

@@ -115,7 +115,7 @@ <h2 id="derived">Derived events</h2>
     return (
         checkout
         .filter(bv.col("amount") > 100)
-        .with_columns(bucket=bv.col("amount").cast(int) // 50)
+        .with_columns(bucket=bv.col("amount").cast("int") // 50)
     )`}
             </Code>
             <p>Push events to the parent (<code>Click</code>); beava routes matching records through the derivation chain and updates downstream tables for free. You never push to a derived event directly.</p>

beava-website/project/docs/get-started/define-a-pipeline/index.html

@@ -125,21 +125,24 @@ <h2 id="good-table">What makes a good beava table?</h2>
             <p>Start with the feature your application actually needs. Add complexity only when the use case requires it.</p>
 
             <h2 id="testing">Test your pipeline</h2>
-            <p>Before you push real traffic at it, drive the pipeline with a fixture and assert what comes out. <code>bv.test.fixture</code> spins up an in-process server, lets you push synthetic events, and exposes the same <code>get</code> / <code>batch_get</code> API you use in production:</p>
+            <p>Before you push real traffic at it, drive the pipeline with a fixture and assert what comes out. <code>bv.test.fixture</code> is a pytest-shaped generator: <code>yield from</code> it inside your own <code>@pytest.fixture</code> and it spins up an in-process server, exposing the same <code>get</code> / <code>batch_get</code> API you use in production:</p>
             <Code lang="python">
-{`import beava as bv
-
-def test_user_clicks_counts_within_window():
-    with bv.test.fixture(Click, UserClicks) as f:
-        f.push("Click", {"user_id": "u_42"})
-        f.push("Click", {"user_id": "u_42"})
-        assert f.get("UserClicks", "u_42") == {"clicks_60s": 2}
-
-        f.advance_time(seconds=120)
-        f.push("Click", {"user_id": "u_42"})
-        assert f.get("UserClicks", "u_42") == {"clicks_60s": 1}`}
+{`import pytest
+import beava as bv
+from beava.test import fixture
+
+@pytest.fixture
+def app():
+    # in-process server; reset_each=True gives every test a clean slate
+    yield from fixture(reset_each=True)
+
+def test_user_clicks_counts_within_window(app):
+    app.register(Click, UserClicks)
+    app.push("Click", {"user_id": "u_42"})
+    app.push("Click", {"user_id": "u_42"})
+    assert app.get("UserClicks", "u_42") == {"clicks_60s": 2}`}
             </Code>
-            <p>The fixture controls the clock with <code>f.advance_time(...)</code>, so you can write deterministic tests for windowed aggregations without sleeping. Run with <code>pytest</code>; no extra config required.</p>
+            <p>The fixture yields a regular <code>bv.App</code>, so the test reads exactly like production code — <code>register</code>, <code>push</code>, <code>get</code>. Run with <code>pytest</code>; no extra config required.</p>
 
             <h2 id="go-deeper">Go deeper</h2>
             <p>Pipelines are written in Python. The engineer who understands the feature can also ship it; the runtime is a single binary that you talk to over HTTP.</p>

beava-website/project/field-guide-ch1.html

@@ -130,13 +130,13 @@ <h2>Try it — 3 curl commands</h2>
 
         <pre className="code-well">
 <span className="cmt"># 1. Register the pipeline</span>{'\n'}
-<span className="kw">curl</span> -X POST localhost:8080/pipelines -d @pipeline.py{'\n'}{'\n'}
+<span className="kw">curl</span> -X POST localhost:8080/register -d @pipeline.json{'\n'}{'\n'}
 <span className="cmt"># 2. Send a watch</span>{'\n'}
-<span className="kw">curl</span> -X POST localhost:8080/events/Watch \<br/>
+<span className="kw">curl</span> -X POST localhost:8080/push \<br/>
 {'  '}-H <span className="str">'content-type: application/json'</span> \<br/>
-{'  '}-d <span className="str">{`'{"user_id":"u1","video_id":"v_abc","watch_seconds":42}'`}</span>{'\n'}{'\n'}
+{'  '}-d <span className="str">{`'{"event":"Watch","data":{"user_id":"u1","video_id":"v_abc","watch_seconds":42}}'`}</span>{'\n'}{'\n'}
 <span className="cmt"># 3. Ask what this user cares about in the last hour</span>{'\n'}
-<span className="kw">curl</span> localhost:8080/features/UserTagAffinity?user_id=u1{'\n'}
+<span className="kw">curl</span> localhost:8080/get/UserTagAffinity/u1{'\n'}
 <span className="cmt">{`# => [{"tag":"woodworking","weight_1h":42.0,...}, ...]`}</span>
         </pre>
 

beava-website/project/field-guide-ch2.html

@@ -134,13 +134,13 @@ <h2>Try it — 3 curl commands</h2>
 
         <pre className="code-well">
 <span className="cmt"># 1. Register the pipeline</span>{'\n'}
-<span className="kw">curl</span> -X POST localhost:8080/pipelines -d @pipeline.py{'\n'}{'\n'}
+<span className="kw">curl</span> -X POST localhost:8080/register -d @pipeline.json{'\n'}{'\n'}
 <span className="cmt"># 2. Send a watch</span>{'\n'}
-<span className="kw">curl</span> -X POST localhost:8080/events/Watch \<br/>
+<span className="kw">curl</span> -X POST localhost:8080/push \<br/>
 {'  '}-H <span className="str">'content-type: application/json'</span> \<br/>
-{'  '}-d <span className="str">{`'{"user_id":"u1","video_id":"v_abc","watch_seconds":42}'`}</span>{'\n'}{'\n'}
+{'  '}-d <span className="str">{`'{"event":"Watch","data":{"user_id":"u1","video_id":"v_abc","watch_seconds":42}}'`}</span>{'\n'}{'\n'}
 <span className="cmt"># 3. Ask what this user cares about in the last hour</span>{'\n'}
-<span className="kw">curl</span> localhost:8080/features/UserTagAffinity?user_id=u1{'\n'}
+<span className="kw">curl</span> localhost:8080/get/UserTagAffinity/u1{'\n'}
 <span className="cmt">{`# => [{"tag":"woodworking","weight_1h":42.0,...}, ...]`}</span>
         </pre>
 

beava-website/project/guide/chapter-1/index.html

@@ -473,9 +473,12 @@ <h2 style={{ fontFamily: 'var(--font-serif)', fontWeight: 600, fontSize: 'clamp(
               borderRadius: 8, padding: '12px 14px', margin: 0,
               overflowX: 'auto',
             }}>
-{'# curl (recommended) — fetches the platform wheel from latest GH Release\n'}
+{'# pip (recommended) — fetches the platform wheel + bundled server binary\n'}
 {'$ pip install beava\n'}
 {'\n'}
+{'# brew (macOS / Linux) — server binary on PATH the Homebrew way\n'}
+{'$ brew install beava-dev/beava/beava\n'}
+{'\n'}
 {'# docker (zero deps on host — :edge tag rebuilt from main on every push)\n'}
 {'$ docker run -p 8080:8080 -p 8081:8081 beavadev/beava:edge\n'}
 {'\n'}
@@ -690,14 +693,15 @@ <h2 style={{ fontFamily: 'var(--font-serif)', fontWeight: 600, fontSize: 'clamp(
     // ---------------- Query panel (live curl + JSON response) ---------------- //
 
     // Renders a curl POST request body and its JSON response based on the
-    // actual rows in state. Teaches: your app calls POST /batch/<Table>
-    // with the user_ids visible on screen, gets back the metrics, renders.
+    // actual rows in state. Teaches: your app calls POST /batch_get with
+    // {requests:[{table,key},...]}, gets back {results:[...]}, renders.
     const QueryPanel = ({ table, userIds, rowsList, metricKeys, empty, onFirstRun, autoRefreshMs }) => {
       // Show all visible users (up to 6 to keep the payload readable).
       const ids = userIds.slice(0, 6);
-      const curl = `curl -X POST https://your-beava/batch/${table} \\
+      const reqBody = JSON.stringify({ requests: ids.map(k => ({ table, key: k })) });
+      const curl = `curl -X POST https://your-beava/batch_get \\
   -H 'Content-Type: application/json' \\
-  -d '${JSON.stringify(ids)}'`;
+  -d '${reqBody}'`;
 
       // User-triggered query: the response only appears after "Run query"
       // is clicked. Once clicked, auto-refresh kicks in so the snapshot
@@ -720,20 +724,23 @@ <h2 style={{ fontFamily: 'var(--font-serif)', fontWeight: 600, fontSize: 'clamp(
 
       const buildResponse = React.useCallback(() => {
         const rowByUid = new Map(rowsRef.current);
-        const obj = {};
+        const results = [];
         for (const uid of idsRef.current) {
           const r = rowByUid.get(uid);
-          if (!r) { obj[uid] = null; continue; }
+          // Server returns a flat feature map per result, in request order
+          // (no table/entity_id/key wrapping).
           const o = {};
-          for (const k of metricKeys) {
-            if (k === 'views_total')           o.views_total = r.views_total;
-            else if (k === 'views_24h')        o.views_24h   = r.views_24h;
-            else if (k === 'distinct_cats_7d') o.distinct_cats_7d = r.cats.size;
-            else if (k === 'last_seen')        o.last_seen   = new Date(r.lastTsMs).toISOString().replace(/\.\d{3}Z$/, 'Z');
+          if (r) {
+            for (const k of metricKeys) {
+              if (k === 'views_total')           o.views_total = r.views_total;
+              else if (k === 'views_24h')        o.views_24h   = r.views_24h;
+              else if (k === 'distinct_cats_7d') o.distinct_cats_7d = r.cats.size;
+              else if (k === 'last_seen')        o.last_seen   = new Date(r.lastTsMs).toISOString().replace(/\.\d{3}Z$/, 'Z');
+            }
           }
-          obj[uid] = o;
+          results.push(o);
         }
-        return JSON.stringify(obj, null, 2);
+        return JSON.stringify({ results }, null, 2);
       }, [metricKeys]);
 
       const doFetch = React.useCallback((opts = {}) => {
@@ -777,7 +784,7 @@ <h2 style={{ fontFamily: 'var(--font-serif)', fontWeight: 600, fontSize: 'clamp(
               color: 'var(--fg3)', letterSpacing: '0.04em',
               textTransform: 'lowercase',
             }}>
-              ↓ how your app reads it: <span style={{ color: 'var(--accent)', fontWeight: 700 }}>POST /batch/{table}</span>
+              ↓ how your app reads it: <span style={{ color: 'var(--accent)', fontWeight: 700 }}>POST /batch_get</span>
             </div>
             <div style={{ display: 'flex', alignItems: 'center', gap: 10 }}>
               {state === 'done' && latency != null && (
@@ -806,9 +813,9 @@ <h2 style={{ fontFamily: 'var(--font-serif)', fontWeight: 600, fontSize: 'clamp(
               <div style={{ fontFamily: 'var(--font-mono)', fontSize: 10.5, color: 'var(--fg3)', letterSpacing: '0.04em', textTransform: 'lowercase', marginBottom: 8 }}>1. request — copy-paste this into your terminal</div>
               <div style={{ position: 'relative' }}>
                 <pre className="code-well" style={{ fontSize: 12.5, padding: '18px 20px', lineHeight: 1.65, margin: 0, whiteSpace: 'pre', minHeight: 140 }}>
-                  <span className="cmt">$ </span>curl -X POST https://your-beava/batch/<span className="fn">{table}</span> \{'\n'}
+                  <span className="cmt">$ </span>curl -X POST https://your-beava/batch_get \{'\n'}
                   {'    '}-H <span className="str">'Content-Type: application/json'</span> \{'\n'}
-                  {'    '}-d <span className="str">'{JSON.stringify(ids)}'</span>
+                  {'    '}-d <span className="str">'{reqBody}'</span>
                 </pre>
                 <CopyBtn text={curl}/>
               </div>

beava-website/project/index.html

@@ -320,7 +320,7 @@
     };
 
     const FeedRow = ({ row, now, columnWidths }) => (
-      <div style={{
+      <div className="beava-feed-row" style={{
         display: 'grid', gridTemplateColumns: columnWidths,
         alignItems: 'center', columnGap: 24,
         padding: '11px 22px', borderTop: '1px solid var(--border)',
@@ -569,6 +569,7 @@
           @media (max-width: 760px) {
             section { padding-left: 18px !important; padding-right: 18px !important; }
             .beava-feed-headers { display: none !important; }
+            .beava-feed-row { grid-template-columns: 1fr !important; align-items: start !important; row-gap: 5px !important; }
             .beava-pillars-grid { grid-template-columns: 1fr !important; }
             .beava-finalcta-grid { grid-template-columns: 1fr !important; }
             footer .bv-foot-grid { grid-template-columns: 1fr !important; gap: 32px !important; }
@@ -1000,7 +1001,7 @@
         {'        '}<span style={{ background: PIPE_HL.blue, borderRadius: 3, padding: '0 2px' }}>top_issue_30m</span>       = bv.<span style={PIPE_CS.fn}>top_k</span>(<span style={PIPE_CS.str}>"topic"</span>, k=<span style={PIPE_CS.num}>1</span>, window=<span style={PIPE_CS.str}>"30m"</span>,{'\n'}
         {'                                    '}where=(bv.<span style={PIPE_CS.fn}>col</span>(<span style={PIPE_CS.str}>"event"</span>) == <span style={PIPE_CS.str}>"api_error"</span>) | (bv.<span style={PIPE_CS.fn}>col</span>(<span style={PIPE_CS.str}>"event"</span>) == <span style={PIPE_CS.str}>"docs_view"</span>)),{'\n'}
         {'    '}){'\n\n'}
-        bv.<span style={PIPE_CS.fn}>App</span>(<span style={PIPE_CS.str}>"0.0.0.0:6400"</span>).<span style={PIPE_CS.fn}>register</span>(<span style={PIPE_CS.ty}>CustomerEvent</span>, <span style={PIPE_CS.fn}>CustomerState</span>).<span style={PIPE_CS.fn}>serve</span>()
+        bv.<span style={PIPE_CS.fn}>App</span>(<span style={PIPE_CS.str}>"http://localhost:8080"</span>).<span style={PIPE_CS.fn}>register</span>(<span style={PIPE_CS.ty}>CustomerEvent</span>, <span style={PIPE_CS.fn}>CustomerState</span>)
       </>
     );
 
@@ -1024,7 +1025,7 @@
         {'    '}<span style={PIPE_CS.kw}>return</span> e.<span style={PIPE_CS.fn}>group_by</span>(<span style={PIPE_CS.str}>"shopper_id"</span>).<span style={PIPE_CS.fn}>agg</span>({'\n'}
         {'        '}<span style={{ background: PIPE_HL.blue, borderRadius: 3, padding: '0 2px' }}>view_price_30m</span> = bv.<span style={PIPE_CS.fn}>mean</span>(<span style={PIPE_CS.str}>"price"</span>, window=<span style={PIPE_CS.str}>"30m"</span>, where=bv.<span style={PIPE_CS.fn}>col</span>(<span style={PIPE_CS.str}>"event"</span>) == <span style={PIPE_CS.str}>"view"</span>),{'\n'}
         {'    '}){'\n\n'}
-        bv.<span style={PIPE_CS.fn}>App</span>(<span style={PIPE_CS.str}>"0.0.0.0:6400"</span>).<span style={PIPE_CS.fn}>register</span>(<span style={PIPE_CS.ty}>CommerceEvent</span>, <span style={PIPE_CS.fn}>SkuMomentum</span>, <span style={PIPE_CS.fn}>ShopperIntent</span>).<span style={PIPE_CS.fn}>serve</span>()
+        bv.<span style={PIPE_CS.fn}>App</span>(<span style={PIPE_CS.str}>"http://localhost:8080"</span>).<span style={PIPE_CS.fn}>register</span>(<span style={PIPE_CS.ty}>CommerceEvent</span>, <span style={PIPE_CS.fn}>SkuMomentum</span>, <span style={PIPE_CS.fn}>ShopperIntent</span>)
       </>
     );
 
@@ -1050,7 +1051,7 @@
         {'    '}<span style={PIPE_CS.kw}>return</span> e.<span style={PIPE_CS.fn}>group_by</span>(<span style={PIPE_CS.str}>"zone_id"</span>).<span style={PIPE_CS.fn}>agg</span>({'\n'}
         {'        '}<span style={{ background: PIPE_HL.blue, borderRadius: 3, padding: '0 2px' }}>jobs_5m</span> = bv.<span style={PIPE_CS.fn}>count</span>(window=<span style={PIPE_CS.str}>"5m"</span>, where=bv.<span style={PIPE_CS.fn}>col</span>(<span style={PIPE_CS.str}>"event"</span>) == <span style={PIPE_CS.str}>"job_created"</span>),{'\n'}
         {'    '}){'\n\n'}
-        bv.<span style={PIPE_CS.fn}>App</span>(<span style={PIPE_CS.str}>"0.0.0.0:6400"</span>).<span style={PIPE_CS.fn}>register</span>(<span style={PIPE_CS.ty}>OpsEvent</span>, <span style={PIPE_CS.fn}>VendorHealth</span>, <span style={PIPE_CS.fn}>ZonePressure</span>).<span style={PIPE_CS.fn}>serve</span>()
+        bv.<span style={PIPE_CS.fn}>App</span>(<span style={PIPE_CS.str}>"http://localhost:8080"</span>).<span style={PIPE_CS.fn}>register</span>(<span style={PIPE_CS.ty}>OpsEvent</span>, <span style={PIPE_CS.fn}>VendorHealth</span>, <span style={PIPE_CS.fn}>ZonePressure</span>)
       </>
     );
 
@@ -1339,7 +1340,8 @@
               to   { opacity: 1; transform: translateY(0); }
             }
             @media (max-width: 920px) {
-              .beava-pipelines-grid { grid-template-columns: 1fr !important; }
+              .beava-pipelines-grid { grid-template-columns: minmax(0, 1fr) !important; }
+              .beava-pipelines-grid > * { min-width: 0 !important; }
             }
           `}</style>
         </section>

beava-website/project/js/_shared/SiteFooter.jsx

@@ -55,12 +55,12 @@ const SiteFooter = ({ maxWidth = 1200 }) => (
         {SITE_FOOTER_COLS.map(col => (
           <div key={col.title}>
             <div style={{ fontFamily: 'var(--font-sans)', fontSize: 12, fontWeight: 700, textTransform: 'uppercase', letterSpacing: '0.08em', color: 'var(--fg3)', marginBottom: 12 }}>{col.title}</div>
-            <ul style={{ listStyle: 'none', padding: 0, margin: 0, display: 'flex', flexDirection: 'column', gap: 8 }}>
+            <ul style={{ listStyle: 'none', padding: 0, margin: 0, display: 'flex', flexDirection: 'column', gap: 2 }}>
               {col.links.map(l => (
                 <li key={l.label}>
                   <a href={l.href}
                      {...(l.external ? { target: '_blank', rel: 'noopener' } : {})}
-                     style={{ fontFamily: 'var(--font-sans)', fontSize: 14, color: 'var(--fg2)', textDecoration: 'none' }}>
+                     style={{ fontFamily: 'var(--font-sans)', fontSize: 14, color: 'var(--fg2)', textDecoration: 'none', display: 'inline-block', padding: '5px 0' }}>
                     {l.label}
                   </a>
                 </li>

beava-website/project/sdk/python/index.html

@@ -140,11 +140,12 @@ <h2 id="pick-a-path">Pick a path</h2>
 
       <!-- =============== INSTALL =============== -->
       <h2 id="install"><span class="step-num">01</span> Install beava</h2>
-      <p><code>pip install beava</code> drops a platform wheel from PyPI. The wheel ships the SDK <em>and</em> the Rust server binary together (~4 MB, polars / ruff / uv pattern). If you'd rather run the server in a container, Docker is the alternative.</p>
+      <p>Three ways in, pick whichever fits your stack. <code>pip install beava</code> drops a platform wheel from PyPI that ships the SDK <em>and</em> the Rust server binary together (~4 MB, polars / ruff / uv pattern). <code>brew install</code> puts the server binary on your PATH the Homebrew way. <code>docker run</code> hands you a container on <code>:8080</code> with a persistent volume.</p>
 
       <div class="tabs" data-tabs="server">
         <div class="tabs-strip">
           <button class="active" data-target="t-pip">pip</button>
+          <button data-target="t-brew">brew</button>
           <button data-target="t-docker">docker</button>
         </div>
 
@@ -164,6 +165,22 @@ <h2 id="install"><span class="step-num">01</span> Install beava</h2>
           <p>The wheel comes from PyPI. The bundled <code>beava</code> binary lands in your Python user-scripts dir (<code>~/.local/bin/beava</code> on Linux, <code>~/Library/Python/&lt;ver&gt;/bin/beava</code> on macOS), so embed mode (<code>bv.App()</code> with no URL) finds it without extra setup. Pin a specific version with <code>pip install beava==0.0.0</code>. Add <code>--memory-only</code> to skip the WAL entirely.</p>
         </div>
 
+        <div class="tabs-panel" id="t-brew">
+          <div class="codeblock">
+            <div class="codeblock-head">
+              <span class="filename">terminal</span>
+              <button class="copy-btn">Copy</button>
+            </div>
+<pre><span class="tk-k">brew</span> install beava-dev/beava/beava
+
+<span class="tk-c"># server binary on PATH; `beava` is ready to run</span>
+<span class="tk-k">beava</span> --data-dir ./.beava/
+<span class="tk-c"># → HTTP listen   : 127.0.0.1:8080</span>
+<span class="tk-c"># → TCP listen    : 127.0.0.1:8081 (enabled=true)</span></pre>
+          </div>
+          <p>The formula installs the <code>beava</code> server binary into your Homebrew prefix (<code>/opt/homebrew/bin/beava</code> on Apple Silicon, <code>/usr/local/bin/beava</code> on Intel). Roll forward with <code>brew upgrade beava</code>. The Python SDK still comes from <code>pip install beava</code>; install both if you want embed mode and the client in the same env.</p>
+        </div>
+
         <div class="tabs-panel" id="t-docker">
           <div class="codeblock">
             <div class="codeblock-head">