fix: harden browser attach and detect virtual mic

browser_controller.py

@@ -19,6 +19,7 @@ Implements the interface expected by hermes_node_agent.py.
 import asyncio
 import base64
 import inspect
+import json
 import logging
 from typing import Dict, Any, Optional
 from pathlib import Path
@@ -42,7 +43,118 @@ class BrowserController:
         self.default_context: Optional[BrowserContext] = None
         self.pages: Dict[str, Page] = {}  # page_id -> page
         self.lock = asyncio.Lock()
+
+    def _ingest_attached_browser_state(self) -> None:
+        """Populate internal context/page maps from an attached browser session."""
+        self.contexts = {}
+        self.pages = {}
+        self.default_context = None
+
+        if not self.browser:
+            return
+
+        all_contexts = list(getattr(self.browser, "contexts", []) or [])
+        if all_contexts and not self.default_context:
+            self.default_context = all_contexts[0]
+
+        page_index = 0
+        for idx, ctx in enumerate(all_contexts):
+            ctx_name = "default" if idx == 0 else f"attached_ctx_{idx}"
+            self.contexts[ctx_name] = ctx
+            if idx == 0:
+                self.default_context = ctx
+            for page in list(getattr(ctx, "pages", []) or []):
+                self.pages[f"page_{page_index}"] = page
+                page_index += 1
         
+    async def _fetch_cdp_target_counts(self, cdp_url: str) -> Dict[str, Any]:
+        """Inspect the raw CDP target list for attach diagnostics."""
+        json_list_url = cdp_url.rstrip("/") + "/json/list"
+        try:
+            proc = await asyncio.create_subprocess_exec(
+                "curl",
+                "-fsS",
+                json_list_url,
+                stdout=asyncio.subprocess.PIPE,
+                stderr=asyncio.subprocess.PIPE,
+            )
+            stdout, stderr = await asyncio.wait_for(proc.communicate(), timeout=5)
+            if proc.returncode != 0:
+                err = (stderr or b"").decode("utf-8", errors="replace").strip()
+                return {"ok": False, "error": err or f"curl exit {proc.returncode}"}
+
+            targets = json.loads((stdout or b"[]").decode("utf-8", errors="replace"))
+            counts: Dict[str, int] = {}
+            for target in targets:
+                kind = str(target.get("type", "unknown"))
+                counts[kind] = counts.get(kind, 0) + 1
+
+            return {
+                "ok": True,
+                "total": len(targets),
+                "counts": counts,
+            }
+        except Exception as e:
+            return {"ok": False, "error": str(e)}
+
+    async def _connect_over_cdp_with_retries(self, browser_type: str, endpoint: str, cdp_url: str, config: Dict[str, Any]):
+        """Attach to a CDP endpoint with progressive timeout backoff and diagnostics."""
+        base_timeout_ms = int(config.get("connect_timeout_ms", 5000))
+        max_retries = max(1, min(5, int(config.get("connect_retries", 5))))
+        timeout_step_ms = int(config.get("connect_timeout_step_ms", 5000))
+
+        browser_launcher = {
+            "chromium": self.playwright.chromium,
+            "firefox": self.playwright.firefox,
+            "webkit": self.playwright.webkit,
+        }.get(browser_type, self.playwright.chromium)
+
+        diagnostics = await self._fetch_cdp_target_counts(cdp_url)
+        if diagnostics.get("ok"):
+            logger.info(
+                "CDP pre-attach targets at %s: total=%s counts=%s",
+                cdp_url,
+                diagnostics.get("total"),
+                diagnostics.get("counts"),
+            )
+        else:
+            logger.warning("CDP pre-attach target inspection failed at %s: %s", cdp_url, diagnostics.get("error"))
+
+        last_error = None
+        for attempt in range(1, max_retries + 1):
+            timeout_ms = base_timeout_ms + ((attempt - 1) * timeout_step_ms)
+            logger.info(
+                "CDP attach attempt %s/%s to %s (browser=%s timeout_ms=%s)",
+                attempt,
+                max_retries,
+                endpoint,
+                browser_type,
+                timeout_ms,
+            )
+            try:
+                browser = await browser_launcher.connect_over_cdp(endpoint, timeout=timeout_ms)
+                logger.info(
+                    "CDP attach succeeded on attempt %s/%s to %s",
+                    attempt,
+                    max_retries,
+                    endpoint,
+                )
+                return browser, {
+                    "attempts": attempt,
+                    "timeout_ms": timeout_ms,
+                    "target_diagnostics": diagnostics,
+                }
+            except Exception as e:
+                last_error = e
+                logger.warning("CDP attach attempt %s/%s failed: %s", attempt, max_retries, e)
+                if attempt < max_retries:
+                    await asyncio.sleep(min(1.5 * attempt, 5.0))
+
+        raise RuntimeError(
+            f"connect_over_cdp failed after {max_retries} attempts "
+            f"(start_timeout_ms={base_timeout_ms}, step_ms={timeout_step_ms}): {last_error}"
+        )
+
     async def initialize(self):
         """Initialize Playwright."""
         if not HAS_PLAYWRIGHT:
@@ -81,6 +193,7 @@ class BrowserController:
         attach = config.get("attach", False)
         cdp_url = config.get("cdp_url", "http://localhost:9222")
         browser_type = config.get("browser_type", "chromium")
+        attach_meta: Dict[str, Any] = {}
         
         try:
             if attach:
@@ -91,17 +204,18 @@ class BrowserController:
                     "webkit": f"{cdp_url.replace(':9222', ':9223') if ':9222' in cdp_url else cdp_url}"
                 }
                 endpoint = endpoints.get(browser_type, cdp_url)
-                
-                if browser_type == "chromium":
-                    self.browser = await self.playwright.chromium.connect_over_cdp(endpoint)
-                elif browser_type == "firefox":
-                    self.browser = await self.playwright.firefox.connect_over_cdp(endpoint)
-                elif browser_type == "webkit":
-                    self.browser = await self.playwright.webkit.connect_over_cdp(endpoint)
-                else:
-                    self.browser = await self.playwright.chromium.connect_over_cdp(endpoint)
-                
-                logger.info(f"Attached to existing {browser_type} browser at {endpoint}")
+                self.browser, attach_meta = await self._connect_over_cdp_with_retries(
+                    browser_type,
+                    endpoint,
+                    cdp_url,
+                    config,
+                )
+                self._ingest_attached_browser_state()
+
+                logger.info(
+                    f"Attached to existing {browser_type} browser at {endpoint} "
+                    f"(contexts={len(self.contexts)}, pages={len(self.pages)})"
+                )
             else:
                 # Launch new browser
                 browser_types = {
@@ -124,7 +238,14 @@ class BrowserController:
                 )
                 logger.info(f"Launched {browser_type} browser (headless={headless})")
             
-            return {"success": True, "browser": browser_type, "headless": headless, "mode": "attach" if attach else "launch"}
+            result = {"success": True, "browser": browser_type, "headless": headless, "mode": "attach" if attach else "launch"}
+            if attach:
+                result["contexts"] = list(self.contexts.keys())
+                result["page_ids"] = list(self.pages.keys())
+                result["attach_attempts"] = attach_meta.get("attempts")
+                result["attach_timeout_ms"] = attach_meta.get("timeout_ms")
+                result["target_diagnostics"] = attach_meta.get("target_diagnostics")
+            return result
             
         except Exception as e:
             logger.error(f"Failed to launch/attach browser: {e}")

hermes_node_agent.py

@@ -634,6 +634,7 @@ class PosixAudioController(AudioControllerBase):
     def capability_info(self) -> Dict[str, Any]:
         monitor_ready, monitor_name = self._default_monitor_source()
         input_ready, input_source = self._default_input_source()
+        inject_ready, inject_source, inject_sink = self._virtual_mic_info()
         return {
             'platform': 'linux',
             'backend': self.backend,
@@ -641,11 +642,13 @@ class PosixAudioController(AudioControllerBase):
             'can_capture_output': monitor_ready,
             'can_capture_input': input_ready,
             'can_play_audio': bool(self.ffplay or self.aplay or self.ffmpeg),
-            'can_inject_mic': False,
+            'can_inject_mic': inject_ready,
             'capture_output_ready': monitor_ready,
             'capture_output_backend': 'pulseaudio-monitor' if monitor_ready else None,
             'default_output_monitor': monitor_name,
             'default_input_source': input_source,
+            'virtual_mic_source': inject_source,
+            'virtual_mic_sink': inject_sink,
             'ffmpeg': bool(self.ffmpeg),
             'ffplay': bool(self.ffplay),
             'pactl': bool(self.pactl),
@@ -690,6 +693,26 @@ class PosixAudioController(AudioControllerBase):
             return True, f'{sink}.monitor'
         return False, None
 
+    def _virtual_mic_info(self) -> tuple[bool, Optional[str], Optional[str]]:
+        if not self.pactl:
+            return False, None, None
+        sinks = self._run_quiet([self.pactl, 'list', 'short', 'sinks'])
+        sources = self._run_quiet([self.pactl, 'list', 'short', 'sources'])
+        sink_names = []
+        source_names = []
+        if sinks['success']:
+            sink_names = [line.split('\t')[1] for line in sinks.get('stdout', '').splitlines() if '\t' in line]
+        if sources['success']:
+            source_names = [line.split('\t')[1] for line in sources.get('stdout', '').splitlines() if '\t' in line]
+        preferred_pairs = [
+            ('hermes_virtual_mic', 'hermes_mic_sink'),
+            ('hermes_mic_sink.monitor', 'hermes_mic_sink'),
+        ]
+        for source_name, sink_name in preferred_pairs:
+            if source_name in source_names and sink_name in sink_names:
+                return True, source_name, sink_name
+        return False, None, None
+
     def _expand_output_path(self, path: Optional[str], suffix: str) -> str:
         if path:
             return str(Path(path).expanduser())
@@ -741,6 +764,7 @@ class PosixAudioController(AudioControllerBase):
     def get_audio_status(self) -> Dict[str, Any]:
         monitor_ready, monitor_name = self._default_monitor_source()
         input_ready, input_source = self._default_input_source()
+        inject_ready, inject_source, inject_sink = self._virtual_mic_info()
         status = {
             'success': True,
             'backend': self.backend,
@@ -750,6 +774,9 @@ class PosixAudioController(AudioControllerBase):
             'capture_output_ready': monitor_ready,
             'capture_input_ready': input_ready,
             'can_play_audio': bool(self.ffplay or self.aplay or self.ffmpeg),
+            'can_inject_mic': inject_ready,
+            'virtual_mic_source': inject_source,
+            'virtual_mic_sink': inject_sink,
         }
         if self.pactl:
             info = self._run_quiet([self.pactl, 'info'])