osborn 0.9.152 → 0.9.154

package/Dockerfile.sandbox

@@ -45,7 +45,8 @@ RUN apt-get update -qq && \
     git \
     make \
     g++ \
-    python-is-python3 && \
+    python-is-python3 \
+    util-linux && \
     rm -rf /var/lib/apt/lists/*
 
 ARG OSBORN_VERSION=latest
@@ -122,6 +123,25 @@ fi
 echo "[sandbox-d] === boot at $(date -Iseconds) ===" >> "$LOGFILE"
 exec > >(tee -a "$LOGFILE") 2>&1
 
+# === Swap on volume ===
+# 8GB swapfile on /workspace (persistent volume). Created once on first boot,
+# reactivated on every subsequent boot (swapon state is lost across cold starts).
+# Gives 8GB RAM + 8GB swap = 16GB addressable — pressure relief when the dev
+# server or concurrent sub-agents spike memory without killing the voice loop.
+SWAPFILE=/workspace/swapfile
+if [ ! -f "$SWAPFILE" ]; then
+  echo "[sandbox-d] Creating 8GB swapfile at $SWAPFILE..."
+  fallocate -l 8G "$SWAPFILE" 2>/dev/null || dd if=/dev/zero of="$SWAPFILE" bs=1M count=8192 status=none
+  chmod 600 "$SWAPFILE"
+  mkswap "$SWAPFILE"
+  echo "[sandbox-d] Swapfile created"
+fi
+if swapon "$SWAPFILE" 2>/dev/null; then
+  echo "[sandbox-d] Swap active: $(free -h | grep Swap)"
+else
+  echo "[sandbox-d] swapon skipped (already active or kernel restriction)"
+fi
+
 # Onboarding-suppression JSON
 ONBOARDING_JSON='{"numStartups":10,"installMethod":"npm","autoUpdates":false,"hasCompletedOnboarding":true,"hasTrustDialogAccepted":true,"hasTrustDialogHooksAccepted":true,"hasCompletedProjectOnboarding":true,"hasAcknowledgedCostThreshold":true,"effortCalloutV2Dismissed":true,"theme":"dark","projects":{"/workspace":{"hasTrustDialogAccepted":true,"hasTrustDialogHooksAccepted":true,"hasCompletedProjectOnboarding":true}}}'
 

package/dist/index.js

@@ -298,6 +298,12 @@ wsProxy.on('error', (err, _req, socket) => {
 // valid cookie are forwarded, so a forgotten agent route can never be silently
 // swallowed by the IDE proxy. Cleared in stopIde().
 let ideSessionToken = null;
+// Cancellation-generation guard for startCodeServer(). Incremented by stopIde()
+// every time the IDE is torn down. startCodeServer() captures myGen at entry;
+// if ideStartGeneration !== myGen when the poll resolves, the start has been
+// superseded (stopIde ran mid-flight) and the result is discarded instead of
+// re-enabling the proxy with no participant present.
+let ideStartGeneration = 0;
 // Helper: parse the osborn_ide cookie from a request and return true iff it
 // matches the current ideSessionToken (and a token exists).
 function hasValidIdeCookie(req) {
@@ -1539,7 +1545,7 @@ function startApiServer(workingDir, port) {
                 port: 8300,
                 path: targetUrl.pathname + (targetUrl.search || ''),
                 method: req.method,
-                headers: { ...req.headers, host: '127.0.0.1:8300' },
+                headers: { ...req.headers, host: '127.0.0.1:8300', 'x-forwarded-host': req.headers.host },
             };
             const proxyReq = httpRequest(options, (proxyRes) => {
                 res.writeHead(proxyRes.statusCode || 502, proxyRes.headers);
@@ -1602,6 +1608,13 @@ function startApiServer(workingDir, port) {
         const isAgentRoute = AGENT_ROUTE_PREFIXES.some(p => upgradePath === p || upgradePath.startsWith(p + '/'));
         if (ideProxyEnabled && hasValidIdeCookie(req) && !isAgentRoute) {
             ideLastProxiedActivity = Date.now();
+            // code-server's authenticateOrigin() compares Origin against the effective host
+            // (X-Forwarded-Host first). changeOrigin:true rewrites Host→127.0.0.1:8300 while the
+            // browser's Origin stays the public hostname, causing a 403 → WS 1006. Preserve the
+            // real public host so the origin check matches.
+            if (!req.headers['x-forwarded-host'] && req.headers.host) {
+                req.headers['x-forwarded-host'] = req.headers.host;
+            }
             wsProxy.ws(req, socket, head, { target: IDE_TARGET });
             return;
         }
@@ -2339,6 +2352,9 @@ async function main() {
     };
     // Kill code-server, disable proxy, clear interval, reset IDE state.
     const stopIde = (reason) => {
+        // Bump the generation counter so any in-flight startCodeServer() knows it
+        // has been cancelled and must not re-enable the proxy or mint a new token.
+        ideStartGeneration++;
         ideProxyEnabled = false;
         ideSessionToken = null;
         ideStartInProgress = null;
@@ -2412,6 +2428,11 @@ async function main() {
         // Fresh start: kill any stale proc (shouldn't normally exist here, but be safe).
         if (ideCodeServerProc)
             stopIde();
+        // Capture the current generation AFTER any preceding stopIde() call so that
+        // if stopIde() already ran (above), myGen reflects the incremented counter.
+        // A concurrent stopIde() after this point will increment ideStartGeneration
+        // beyond myGen, signalling cancellation to the poll loop and success branch.
+        const myGen = ideStartGeneration;
         let resolveFn;
         let rejectFn;
         ideStartInProgress = new Promise((res, rej) => { resolveFn = res; rejectFn = rej; });
@@ -2462,11 +2483,13 @@ async function main() {
             await new Promise((res, rej) => {
                 let attempts = 0;
                 const poll = setInterval(() => {
-                    // Early-exit: the proc we spawned is no longer the active one (stopIde
-                    // was called, or a concurrent path replaced it).
-                    if (ideCodeServerProc !== codeServerProc) {
+                    // Early-exit: stopIde() was called while we were waiting (generation bumped),
+                    // OR the proc we spawned is no longer the active one for another reason.
+                    // Either way, cancel the poll immediately so we don't keep hitting 8300
+                    // for a boot that has already been invalidated.
+                    if (ideStartGeneration !== myGen || ideCodeServerProc !== codeServerProc) {
                         clearInterval(poll);
-                        rej(new Error('code-server process was replaced or stopped mid-poll'));
+                        rej(new Error('code-server start cancelled (stopIde ran mid-poll)'));
                         return;
                     }
                     attempts++;
@@ -2487,6 +2510,25 @@ async function main() {
                 }, 500);
             });
             console.log('✅ code-server ready on 8300');
+            // Generation guard: check again right before enabling the proxy. The poll
+            // already checks this on each tick, but there is a narrow window between the
+            // last tick resolving and reaching here. If stopIde() ran in that window,
+            // ideStartGeneration will have been incremented beyond myGen — we must NOT
+            // enable the proxy or mint a token, and we must kill the orphan process so
+            // no zombie code-server is left running with no participant present.
+            if (ideStartGeneration !== myGen) {
+                console.log('🖥️ IDE start cancelled (stopIde ran while code-server was booting) — killing orphan proc');
+                try {
+                    codeServerProc.kill();
+                }
+                catch { }
+                // ideCodeServerProc may already be null (stopIde cleared it); only null it
+                // if it still points to our proc to avoid clobbering a newer start.
+                if (ideCodeServerProc === codeServerProc)
+                    ideCodeServerProc = null;
+                resolveFn(); // resolve (not reject) so the promise chain exits cleanly
+                return;
+            }
             // Mint the session cookie token and enable the proxy. From this point the
             // /editor route will 302+Set-Cookie and the fall-through proxy will only
             // forward requests that carry the valid osborn_ide cookie.
@@ -2667,7 +2709,7 @@ async function main() {
                 const { readSessionHistory } = await import('./session-access.js');
                 const history = readSessionHistory(sessionId, workingDir, {
                     lastN: 10,
-                    types: ['assistant'],
+                    types: ['assistant', 'tool_use'], // tool_use entries also carry Claude's spoken text
                 });
                 recentMessages = history
                     .filter((m) => m.text)
@@ -2683,8 +2725,14 @@ async function main() {
         // BEFORE the previous TTS completed, and we may have already suppressed in-flight
         // tts_say events that arrived during that overlap).
         const carriedSuppressed = lastInterruption?.suppressedText ?? '';
-        lastInterruption = { spokenText: fullText, recentMessages, suppressedText: carriedSuppressed, timestamp: Date.now() };
-        console.log(`📋 Interruption context stored (text: ${fullText.length} chars, JSONL: ${recentMessages.length} chars, suppressed carried: ${carriedSuppressed.length} chars)`);
+        lastInterruption = {
+            spokenText: fullText,
+            recentMessages,
+            suppressedText: carriedSuppressed,
+            fullBlockText: fullBlockText && fullBlockText !== fullText ? fullBlockText : undefined,
+            timestamp: Date.now(),
+        };
+        console.log(`📋 Interruption context stored (text: ${fullText.length} chars, JSONL: ${recentMessages.length} chars, block: ${fullBlockText?.length ?? 0} chars, suppressed carried: ${carriedSuppressed.length} chars)`);
     }
     /**
      * Append text the agent tried to say while the user was speaking, but which we
@@ -2724,6 +2772,7 @@ async function main() {
             spokenText: lastInterruption.spokenText,
             recentMessages: lastInterruption.recentMessages,
             suppressedText: lastInterruption.suppressedText,
+            fullBlockText: lastInterruption.fullBlockText,
         };
         lastInterruption = null;
         return ctx;

package/dist/pipeline-direct-llm.d.ts

@@ -21,6 +21,8 @@ export interface InterruptionContext {
      * and can re-articulate the relevant bits in its next response.
      */
     suppressedText: string;
+    /** Full text of the interrupted TTS block — fallback when JSONL read returns empty */
+    fullBlockText?: string;
 }
 export interface PipelineDirectOptions extends ClaudeLLMOptions {
     onFastBrainResult?: (result: FastBrainPanelResult) => void;

package/dist/pipeline-direct-llm.js

@@ -204,7 +204,10 @@ export class PipelineDirectLLM extends llm.LLM {
                     `Anything in "Your recent messages" below that appears AFTER the quoted heard text is content the user did not hear. The user has no memory of it.`,
                     ``,
                     `Your recent messages (full untruncated — you wrote these):`,
-                    interruptCtx.recentMessages || '(no recent messages found)',
+                    interruptCtx.recentMessages
+                        || (interruptCtx.fullBlockText
+                            ? `(JSONL unavailable — full TTS block text follows)\n"${interruptCtx.fullBlockText}"`
+                            : '(no recent messages found)'),
                     suppressedBlock,
                     ``,
                     `User's message: "${userText}"`,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "osborn",
-  "version": "0.9.152",
+  "version": "0.9.154",
   "description": "Voice AI coding assistant - local agent that connects to Osborn frontend",
   "type": "module",
   "bin": {