agentvibes 5.13.1 → 5.15.0

package/.claude/hooks-windows/play-tts.ps1

@@ -73,7 +73,8 @@ $HooksDir = "$ClaudeDir\hooks-windows"
 $ProviderFile = "$ClaudeDir\tts-provider.txt"
 $MuteFile = "$ClaudeDir\tts-muted.txt"
 
-# Check if TTS is muted
+# Check if TTS is muted (receiver master switch — /agent-vibes:receiver off).
+# Kept as an absolute pre-check so existing behaviour does not regress.
 if (Test-Path $MuteFile) {
     $muteStatus = Get-Content $MuteFile -Raw
     if ($muteStatus.Trim() -eq "true") {
@@ -81,6 +82,35 @@ if (Test-Path $MuteFile) {
     }
 }
 
+# ---------------------------------------------------------------------------
+# MUTE PRECEDENCE (parity with .claude/hooks/play-tts.sh:89-114).
+#
+# Until now this player read ONLY tts-muted.txt above, which /agent-vibes:mute
+# never writes — so `/agent-vibes:mute` did not silence Windows audio at all,
+# and the global `~/.agentvibes-muted` kill-switch was ignored on Windows.
+# Same three levels as bash, in the same order:
+#   1. project agentvibes-unmuted -> speak (overrides a global mute)
+#   2. project agentvibes-muted   -> silent
+#   3. global ~/.agentvibes-muted -> silent
+#
+# The project markers live under the REAL project root (CLAUDE_PROJECT_DIR when
+# set), not $ClaudeDir, which falls back to the user profile for global installs.
+# $HOME is preferred over $env:USERPROFILE so markers written by bash under
+# git-bash resolve to the same file here.
+$HomeDir = if ($env:HOME) { $env:HOME } else { $env:USERPROFILE }
+$MuteScopeDir = if ($env:CLAUDE_PROJECT_DIR -and (Test-Path "$env:CLAUDE_PROJECT_DIR\.claude")) {
+    "$env:CLAUDE_PROJECT_DIR\.claude"
+} else {
+    $ClaudeDir
+}
+if (Test-Path (Join-Path $MuteScopeDir "agentvibes-unmuted")) {
+    # explicit per-project enable — wins over the global kill-switch
+} elseif (Test-Path (Join-Path $MuteScopeDir "agentvibes-muted")) {
+    exit 0
+} elseif (Test-Path (Join-Path $HomeDir ".agentvibes-muted")) {
+    exit 0
+}
+
 # Determine active provider
 $ActiveProvider = "windows-sapi"
 if (Test-Path $ProviderFile) {
@@ -453,6 +483,70 @@ elseif (-not $PlanOk -and $_LlmVoice -and -not $VoiceOverride) {
     $VoiceOverride = $_LlmVoice
 }
 
+# --- Dynamic session self-ID ({{session}} token) ------------------------------
+# Parity with play-tts.sh:250-296. A per-LLM PRETEXT containing {{session}}
+# expands to "<LLM> on <Project> in <Terminal>" so multi-session users can tell
+# which window just spoke. Announced ONCE per session; on later utterances the
+# token is dropped rather than prefixing every line. Opt-in — this whole block
+# is inert unless the configured pretext actually contains the token.
+if ($_LlmPretext -like "*{{session}}*") {
+    # Session key. There is no true Claude-session id available to a hook, so
+    # this is best-effort: a terminal-session id when the emulator exports one
+    # (stable for that tab's life), else this process's parent PID. Worst case
+    # the self-ID repeats or is skipped — cosmetic, and only when opted in.
+    $_SidRaw = if ($env:WT_SESSION) { $env:WT_SESSION }
+               elseif ($env:TERM_SESSION_ID) { $env:TERM_SESSION_ID }
+               elseif ($env:TMUX) { $env:TMUX }
+               else {
+                   try { (Get-CimInstance Win32_Process -Filter "ProcessId=$PID").ParentProcessId } catch { $PID }
+               }
+    $_SidProj = ""
+    $_SidProjDirRaw = if ($env:CLAUDE_PROJECT_DIR) { $env:CLAUDE_PROJECT_DIR } else { $ClaudeDir }
+    try { $_SidProj = Split-Path -Leaf ($_SidProjDirRaw.TrimEnd('\', '/')) } catch { }
+    $_SidKey = (("$llm|$_SidProj|$_SidRaw") -replace '[^A-Za-z0-9]', '_')
+
+    # Announce-marker dir under the per-user temp path (already user-scoped on
+    # Windows: %LOCALAPPDATA%\Temp), so no cross-user marker collisions.
+    $_AnnDir  = Join-Path ([System.IO.Path]::GetTempPath()) "agentvibes-session"
+    $_AnnOk   = $false
+    try { $null = New-Item -ItemType Directory -Path $_AnnDir -Force -ErrorAction Stop; $_AnnOk = $true } catch { }
+    $_AnnFile = Join-Path $_AnnDir "announced-$_SidKey"
+
+    # First utterance -> replacement is the self-ID; later utterances -> empty.
+    $_SessionId = ""
+    if ($_AnnOk -and (Test-Path $_AnnFile)) {
+        $_SessionId = ""
+    } else {
+        $_SidScript = Join-Path $HooksDir "agentvibes-session-id.ps1"
+        if (Test-Path $_SidScript) {
+            $_SidProjDir = if ($env:CLAUDE_PROJECT_DIR) { $env:CLAUDE_PROJECT_DIR } else { $ClaudeDir }
+            try { $_SessionId = & $_SidScript -LlmKey $llm -ProjectDir $_SidProjDir } catch { $_SessionId = "" }
+        }
+        if ($_AnnOk) {
+            try {
+                if (-not (Test-Path $_AnnFile)) { $null = New-Item -ItemType File -Path $_AnnFile -ErrorAction Stop }
+                # Bound growth: drop announce-markers older than a day.
+                Get-ChildItem -Path $_AnnDir -Filter 'announced-*' -ErrorAction SilentlyContinue |
+                    Where-Object { $_.LastWriteTime -lt (Get-Date).AddDays(-1) } |
+                    Remove-Item -Force -ErrorAction SilentlyContinue
+            } catch { }
+        }
+    }
+
+    if ($_SessionId) {
+        $_LlmPretext = $_LlmPretext -replace [regex]::Escape('{{session}}'), $_SessionId
+    } else {
+        # Empty replacement: also swallow one adjacent ", " so "Hey {{session}}, ready"
+        # becomes "Hey ready", not "Hey , ready"; then drop any bare token.
+        $_LlmPretext = $_LlmPretext -replace [regex]::Escape('{{session}}, '), ''
+        $_LlmPretext = $_LlmPretext -replace [regex]::Escape(', {{session}}'), ''
+        $_LlmPretext = $_LlmPretext -replace [regex]::Escape('{{session}}'), ''
+    }
+    # Final tidy: collapse doubled/edge commas and runs of whitespace.
+    $_LlmPretext = $_LlmPretext -replace ',\s*,', ',' -replace '^\s*,\s*', '' -replace '\s*,\s*$', '' -replace '\s{2,}', ' '
+    $_LlmPretext = $_LlmPretext.Trim()
+}
+
 # --- Apply LLM-specific pretext ----------------------------------------------
 # Prepend the configured pretext (e.g. "Agent Vibes Here") to the speech
 # text.  Guard against double-prefixing on re-entrant or looped calls by
@@ -509,6 +603,23 @@ if ($OverrideEffects -ne "" -and $OverrideEffects -in @("off", "light", "medium"
     }
 }
 
+# Fail loudly when mixing was requested but ffmpeg is unavailable. Without this,
+# background music and reverb are dropped silently and the voice plays "fine" —
+# indistinguishable from a config problem. A long-running tts-watcher.ps1 that
+# started before ffmpeg was installed keeps a stale PATH snapshot and hits this,
+# so the message names the restart explicitly. Mirrors the bash-side warning in
+# audio-processor.sh ("ffmpeg not installed, skipping background mix").
+if (-not $HasFfmpeg -and ($BgEnabled -or $HasReverb)) {
+    $_dropped = @()
+    if ($BgEnabled) { $_dropped += "background music" }
+    if ($HasReverb) { $_dropped += "reverb ($ReverbLevel)" }
+    Write-Host ("[WARNING] play-tts.ps1: ffmpeg not found on PATH — dropping " +
+        ($_dropped -join " and ") + "; voice will play unmixed.") -ForegroundColor Yellow
+    Write-Host ("[WARNING] Install ffmpeg (e.g. 'scoop install ffmpeg'). If it IS " +
+        "installed, a background tts-watcher started before it will have a stale " +
+        "PATH — restart the watcher.") -ForegroundColor Yellow
+}
+
 # --- Apply LLM-specific engine override --------------------------------------
 # Allowed local Windows engines: windows-sapi, windows-piper, soprano.
 # Transport providers (ssh-remote etc.) are not listed because they forward

package/.claude/hooks-windows/session-start-tts.ps1

@@ -33,6 +33,42 @@ if ($env:CLAUDE_PROJECT_DIR -and (Test-Path "$env:CLAUDE_PROJECT_DIR\.claude"))
     $ProjectClaudeDir = Join-Path (Split-Path -Parent (Split-Path -Parent $ScriptDir)) ".claude"
 }
 
+# ---------------------------------------------------------------------------
+# OPT-IN INJECTION GATE (Issue: global-install cacophony).
+#
+# Mirror of the gate in .claude/hooks/session-start-tts.sh — see the long
+# rationale there. Previously this hook injected the ~250-token TTS protocol
+# into EVERY Windows session where AgentVibes was found on disk, so a global
+# install made every open session start speaking at once, and idle sessions
+# still paid the token cost. Injection is now OPT-IN on Windows too.
+#
+# Precedence (identical to bash, so injection and audio agree):
+#   1. project agentvibes-unmuted  -> inject   (explicit per-project ON)
+#   2. project agentvibes-muted    -> silent
+#   3. global ~/.agentvibes-muted  -> silent   (the global kill-switch)
+#   4. project agentvibes-enabled, OR global opt-in
+#      (~/.claude/agentvibes-enabled or ~/.claude/agentvibes-unmuted) -> inject
+#   5. otherwise                   -> silent   (opt-in default; zero tokens)
+#
+# $HOME is honoured ahead of $env:USERPROFILE so the markers resolve the same
+# way under git-bash (where bash writes them) as under native PowerShell.
+$HomeDir = if ($env:HOME) { $env:HOME } else { $env:USERPROFILE }
+if (Test-Path (Join-Path $ProjectClaudeDir "agentvibes-unmuted")) {
+    # explicit per-project enable — wins over a global mute, as in play-tts.sh
+} elseif (Test-Path (Join-Path $ProjectClaudeDir "agentvibes-muted")) {
+    exit 0
+} elseif (Test-Path (Join-Path $HomeDir ".agentvibes-muted")) {
+    exit 0   # global kill-switch, no per-project override
+} elseif ((Test-Path (Join-Path $ProjectClaudeDir "agentvibes-enabled")) -or
+          (Test-Path (Join-Path $HomeDir ".claude\agentvibes-enabled"))  -or
+          (Test-Path (Join-Path $HomeDir ".claude\agentvibes-unmuted"))) {
+    # opted in (project install marker, or a deliberate global opt-in)
+} else {
+    # Not enabled for this project -> inject nothing (zero tokens).
+    # Enable for this project with:  /agent-vibes:unmute
+    exit 0
+}
+
 # Build the -ProjectDir flag to inject into TTS commands (empty string = omit flag).
 # Sanitize: strip any embedded quotes that would break PowerShell argument quoting.
 $ProjectDirFlag = ""

package/.claude/hooks-windows/soprano-gradio-synth.py

@@ -1,153 +1,153 @@
-#!/usr/bin/env python3
-#
-# File: .claude/hooks-windows/soprano-gradio-synth.py
-#
-# AgentVibes - Finally, your AI Agents can Talk Back!
-# Website: https://agentvibes.org
-# Repository: https://github.com/paulpreibisch/AgentVibes
-#
-# Co-created by Paul Preibisch with Claude AI
-# Copyright (c) 2025 Paul Preibisch
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-#
-# ---
-#
-# Soprano Gradio WebUI synthesizer helper.
-# Calls the Soprano WebUI's Gradio API and saves the result as a WAV file.
-# Uses only Python stdlib (json, sys, urllib) — no extra dependencies.
-#
-# Usage: python3 soprano-gradio-synth.py "text to speak" output.wav [port]
-#
-"""
-Soprano Gradio WebUI synthesizer helper for AgentVibes.
-
-Calls the Soprano WebUI's Gradio API via the Server-Sent Events (SSE) protocol
-and downloads the generated audio as a WAV file.
-
-Flow:
-  1. Submit generation request -> get event_id
-  2. Poll SSE stream for audio file URL
-  3. Download WAV file to output path
-
-See: https://github.com/ekwek1/soprano
-"""
-import json
-import sys
-import urllib.request
-import urllib.error
-
-
-def synth(text: str, output_path: str, port: int = 7860) -> None:
-    base = f"http://127.0.0.1:{port}"
-
-    # Step 1: Submit generation request
-    # Args: text, temperature, top_p, repetition_penalty, chunk_size, streaming
-    payload = json.dumps({
-        "data": [text, 0.0, 0.95, 1.2, 1, False]
-    }).encode()
-
-    event_id = submit_request(base, payload)
-
-    # Step 2: Poll SSE stream for audio file URL
-    audio_url = poll_for_result(base, event_id)
-
-    # Step 3: Download the audio file (validate localhost to prevent SSRF)
-    download_file(audio_url, output_path, port)
-
-
-def submit_request(base: str, payload: bytes) -> str:
-    """Submit generation request to Gradio API, return event_id."""
-    for api_base in ["/gradio_api/call", "/call"]:
-        url = f"{base}{api_base}/generate_speech"
-        req = urllib.request.Request(
-            url,
-            data=payload,
-            headers={"Content-Type": "application/json"},
-        )
-        try:
-            with urllib.request.urlopen(req, timeout=30) as resp:
-                return json.loads(resp.read())["event_id"]
-        except urllib.error.URLError:
-            continue
-
-    raise RuntimeError("Could not reach Soprano WebUI API")
-
-
-def poll_for_result(base: str, event_id: str) -> str:
-    """Poll SSE endpoint until audio file URL is returned."""
-    for api_base in ["/gradio_api/call", "/call"]:
-        url = f"{base}{api_base}/generate_speech/{event_id}"
-        req = urllib.request.Request(url)
-        try:
-            with urllib.request.urlopen(req, timeout=120) as resp:
-                for raw_line in resp:
-                    line = raw_line.decode("utf-8", errors="replace").strip()
-                    if not line.startswith("data: "):
-                        continue
-                    try:
-                        parsed = json.loads(line[6:])
-                    except json.JSONDecodeError:
-                        continue
-                    # Response format: [{"path": "...", "url": "...", ...}, "status string"]
-                    if isinstance(parsed, list) and len(parsed) >= 1:
-                        audio = parsed[0]
-                        if isinstance(audio, dict) and "url" in audio:
-                            return audio["url"]
-            break
-        except urllib.error.URLError:
-            continue
-
-    raise RuntimeError("No audio URL in Soprano response")
-
-
-def download_file(url: str, output_path: str, port: int = 7860) -> None:
-    """Download audio file from Gradio file server.
-
-    Security: Validates URL points to localhost to prevent SSRF.
-    """
-    # Validate URL is localhost to prevent SSRF via crafted Gradio responses
-    allowed_prefixes = [
-        f"http://127.0.0.1:{port}/",
-        f"http://localhost:{port}/",
-    ]
-    if not any(url.startswith(prefix) for prefix in allowed_prefixes):
-        raise RuntimeError(f"Refused download from non-localhost URL: {url[:80]}")
-
-    req = urllib.request.Request(url)
-    with urllib.request.urlopen(req, timeout=30) as resp:
-        with open(output_path, "wb") as f:
-            while True:
-                chunk = resp.read(8192)
-                if not chunk:
-                    break
-                f.write(chunk)
-
-
-if __name__ == "__main__":
-    if len(sys.argv) < 3:
-        print(f"Usage: {sys.argv[0]} \"text\" output.wav [port]", file=sys.stderr)
-        sys.exit(1)
-
-    text = sys.argv[1]
-    output = sys.argv[2]
-    port = int(sys.argv[3]) if len(sys.argv) > 3 else 7860
-    if not (1 <= port <= 65535):
-        print(f"Error: Invalid port number: {port}", file=sys.stderr)
-        sys.exit(1)
-
-    try:
-        synth(text, output, port)
-    except Exception as e:
-        print(f"Error: {e}", file=sys.stderr)
-        sys.exit(1)
+#!/usr/bin/env python3
+#
+# File: .claude/hooks-windows/soprano-gradio-synth.py
+#
+# AgentVibes - Finally, your AI Agents can Talk Back!
+# Website: https://agentvibes.org
+# Repository: https://github.com/paulpreibisch/AgentVibes
+#
+# Co-created by Paul Preibisch with Claude AI
+# Copyright (c) 2025 Paul Preibisch
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+# ---
+#
+# Soprano Gradio WebUI synthesizer helper.
+# Calls the Soprano WebUI's Gradio API and saves the result as a WAV file.
+# Uses only Python stdlib (json, sys, urllib) — no extra dependencies.
+#
+# Usage: python3 soprano-gradio-synth.py "text to speak" output.wav [port]
+#
+"""
+Soprano Gradio WebUI synthesizer helper for AgentVibes.
+
+Calls the Soprano WebUI's Gradio API via the Server-Sent Events (SSE) protocol
+and downloads the generated audio as a WAV file.
+
+Flow:
+  1. Submit generation request -> get event_id
+  2. Poll SSE stream for audio file URL
+  3. Download WAV file to output path
+
+See: https://github.com/ekwek1/soprano
+"""
+import json
+import sys
+import urllib.request
+import urllib.error
+
+
+def synth(text: str, output_path: str, port: int = 7860) -> None:
+    base = f"http://127.0.0.1:{port}"
+
+    # Step 1: Submit generation request
+    # Args: text, temperature, top_p, repetition_penalty, chunk_size, streaming
+    payload = json.dumps({
+        "data": [text, 0.0, 0.95, 1.2, 1, False]
+    }).encode()
+
+    event_id = submit_request(base, payload)
+
+    # Step 2: Poll SSE stream for audio file URL
+    audio_url = poll_for_result(base, event_id)
+
+    # Step 3: Download the audio file (validate localhost to prevent SSRF)
+    download_file(audio_url, output_path, port)
+
+
+def submit_request(base: str, payload: bytes) -> str:
+    """Submit generation request to Gradio API, return event_id."""
+    for api_base in ["/gradio_api/call", "/call"]:
+        url = f"{base}{api_base}/generate_speech"
+        req = urllib.request.Request(
+            url,
+            data=payload,
+            headers={"Content-Type": "application/json"},
+        )
+        try:
+            with urllib.request.urlopen(req, timeout=30) as resp:
+                return json.loads(resp.read())["event_id"]
+        except urllib.error.URLError:
+            continue
+
+    raise RuntimeError("Could not reach Soprano WebUI API")
+
+
+def poll_for_result(base: str, event_id: str) -> str:
+    """Poll SSE endpoint until audio file URL is returned."""
+    for api_base in ["/gradio_api/call", "/call"]:
+        url = f"{base}{api_base}/generate_speech/{event_id}"
+        req = urllib.request.Request(url)
+        try:
+            with urllib.request.urlopen(req, timeout=120) as resp:
+                for raw_line in resp:
+                    line = raw_line.decode("utf-8", errors="replace").strip()
+                    if not line.startswith("data: "):
+                        continue
+                    try:
+                        parsed = json.loads(line[6:])
+                    except json.JSONDecodeError:
+                        continue
+                    # Response format: [{"path": "...", "url": "...", ...}, "status string"]
+                    if isinstance(parsed, list) and len(parsed) >= 1:
+                        audio = parsed[0]
+                        if isinstance(audio, dict) and "url" in audio:
+                            return audio["url"]
+            break
+        except urllib.error.URLError:
+            continue
+
+    raise RuntimeError("No audio URL in Soprano response")
+
+
+def download_file(url: str, output_path: str, port: int = 7860) -> None:
+    """Download audio file from Gradio file server.
+
+    Security: Validates URL points to localhost to prevent SSRF.
+    """
+    # Validate URL is localhost to prevent SSRF via crafted Gradio responses
+    allowed_prefixes = [
+        f"http://127.0.0.1:{port}/",
+        f"http://localhost:{port}/",
+    ]
+    if not any(url.startswith(prefix) for prefix in allowed_prefixes):
+        raise RuntimeError(f"Refused download from non-localhost URL: {url[:80]}")
+
+    req = urllib.request.Request(url)
+    with urllib.request.urlopen(req, timeout=30) as resp:
+        with open(output_path, "wb") as f:
+            while True:
+                chunk = resp.read(8192)
+                if not chunk:
+                    break
+                f.write(chunk)
+
+
+if __name__ == "__main__":
+    if len(sys.argv) < 3:
+        print(f"Usage: {sys.argv[0]} \"text\" output.wav [port]", file=sys.stderr)
+        sys.exit(1)
+
+    text = sys.argv[1]
+    output = sys.argv[2]
+    port = int(sys.argv[3]) if len(sys.argv) > 3 else 7860
+    if not (1 <= port <= 65535):
+        print(f"Error: Invalid port number: {port}", file=sys.stderr)
+        sys.exit(1)
+
+    try:
+        synth(text, output, port)
+    except Exception as e:
+        print(f"Error: {e}", file=sys.stderr)
+        sys.exit(1)

package/.claude/verbosity.txt

@@ -1 +1 @@
-high
+high