crewly 1.20.131 → 1.20.133

package/config/skills/agent/pdf-tools/style.css

@@ -0,0 +1,143 @@
+/*
+ * pdf-tools default stylesheet (Crewly, MIT).
+ *
+ * Clean A4 document typography with CJK-first font stacks: PingFang /
+ * Hiragino on macOS, Noto Sans/Serif CJK (fonts-noto-cjk) on Linux,
+ * Microsoft YaHei on Windows. Used for Markdown input; HTML input keeps its
+ * own styling unless `addDefaultStyle` is set.
+ *
+ * Page helpers: <div class="page-break"></div> starts a new page;
+ * class="keep-together" avoids splitting a block across pages.
+ */
+
+@page {
+  size: A4;
+  margin: 18mm 16mm 20mm 16mm;
+}
+
+:root {
+  --text: #1f2328;
+  --muted: #57606a;
+  --accent: #2f6feb;
+  --rule: #d8dee4;
+  --code-bg: #f3f4f6;
+  --table-head: #f0f3f6;
+  --zebra: #fafbfc;
+  --sans: "PingFang SC", "Hiragino Sans GB", "Noto Sans CJK SC", "Noto Sans SC", "Source Han Sans SC",
+    "Microsoft YaHei", "WenQuanYi Micro Hei", -apple-system, "Helvetica Neue", Arial, sans-serif;
+  --serif: "Songti SC", "Noto Serif CJK SC", "Noto Serif SC", "Source Han Serif SC", Georgia, serif;
+  --mono: "SF Mono", Menlo, "JetBrains Mono", "Noto Sans Mono CJK SC", Consolas, monospace;
+}
+
+html {
+  -webkit-print-color-adjust: exact;
+  print-color-adjust: exact;
+}
+
+body {
+  font-family: var(--sans);
+  font-size: 10.5pt;
+  line-height: 1.7;
+  color: var(--text);
+  margin: 0;
+  text-align: justify;
+  word-break: normal;
+  overflow-wrap: anywhere;
+}
+
+main.doc {
+  max-width: 100%;
+}
+
+h1, h2, h3, h4, h5 {
+  font-family: var(--sans);
+  font-weight: 650;
+  line-height: 1.35;
+  color: #111;
+  break-after: avoid;
+  page-break-after: avoid;
+}
+
+h1 {
+  font-size: 20pt;
+  margin: 0 0 10pt;
+  padding-bottom: 6pt;
+  border-bottom: 2px solid var(--text);
+}
+
+h2 {
+  font-size: 14.5pt;
+  margin: 18pt 0 8pt;
+  padding-bottom: 3pt;
+  border-bottom: 1px solid var(--rule);
+}
+
+h3 { font-size: 12pt; margin: 14pt 0 6pt; }
+h4 { font-size: 10.5pt; margin: 12pt 0 4pt; color: var(--muted); }
+
+p { margin: 0 0 8pt; }
+
+a { color: var(--accent); text-decoration: none; }
+
+ul, ol { margin: 0 0 8pt; padding-left: 18pt; }
+li { margin: 2pt 0; }
+
+blockquote {
+  margin: 10pt 0;
+  padding: 6pt 12pt;
+  border-left: 3px solid var(--accent);
+  background: #f5f8ff;
+  color: #333;
+}
+
+blockquote > :last-child { margin-bottom: 0; }
+
+code {
+  font-family: var(--mono);
+  font-size: 0.9em;
+  background: var(--code-bg);
+  padding: 1pt 4pt;
+  border-radius: 3px;
+}
+
+pre {
+  font-family: var(--mono);
+  font-size: 8.5pt;
+  line-height: 1.5;
+  background: var(--code-bg);
+  padding: 8pt 10pt;
+  border-radius: 4px;
+  white-space: pre-wrap;
+  break-inside: avoid;
+}
+
+pre code { background: none; padding: 0; }
+
+table {
+  width: 100%;
+  border-collapse: collapse;
+  margin: 10pt 0;
+  font-size: 9.5pt;
+  break-inside: auto;
+}
+
+th, td {
+  border: 1px solid var(--rule);
+  padding: 4pt 7pt;
+  text-align: left;
+  vertical-align: top;
+}
+
+th { background: var(--table-head); font-weight: 600; }
+tr:nth-child(even) td { background: var(--zebra); }
+thead { display: table-header-group; }
+tr { break-inside: avoid; }
+
+img { max-width: 100%; }
+
+hr { border: none; border-top: 1px solid var(--rule); margin: 16pt 0; }
+
+.page-break { break-after: page; page-break-after: always; }
+.keep-together { break-inside: avoid; page-break-inside: avoid; }
+.muted { color: var(--muted); }
+.serif { font-family: var(--serif); }

package/config/skills/agent/transcribe-audio/SKILL.md

@@ -1,19 +1,34 @@
 ---
 name: transcribe-audio
-description: Transcribe a local audio (or video) file to text with timestamps using Whisper. Defaults to local whisper.cpp (free, offline, word/segment-level timestamps) and automatically falls back to the OpenAI Whisper API (whisper-1) when the local engine is not installed. Any agent can call it for meetings, podcasts, voice notes, interviews, or video audio tracks.
-category: content
+description: Transcribe a local audio (or video) file — Slack voice messages / audio clips, meetings, podcasts, interviews — to text with timestamps using Whisper. Defaults to local whisper.cpp (free, offline) and falls back to the OpenAI Whisper API (whisper-1) when the local engine is not installed. When a dependency is missing it returns `needsSetup:true` — run `install-skill --id transcribe-audio` instead of telling the user it cannot be done.
+category: content-creation
 assignableRoles:
   - "*"
-version: "1.0.0"
+version: "1.1.0"
+author: Crewly Team
 tags:
   - audio
+  - voice
+  - voice-message
+  - speech-to-text
   - transcribe
   - transcription
   - whisper
   - whisper.cpp
   - openai
-  - speech-to-text
   - timestamps
+  - m4a
+  - mp3
+  - video
+  - 语音
+  - 录音
+  - 转文字
+triggers:
+  - transcribe audio
+  - transcribe voice message
+  - speech to text
+  - what does this recording say
+  - audio clip
 ---
 
 # Transcribe Audio (Whisper)
@@ -91,10 +106,32 @@ downstream consumers have a stable shape).
 
 ## Dependencies & Setup
 
+**Automatic:** everything below is declared in `skill.json` → `setup`. Run
+`install-skill --id transcribe-audio` (agents) or `crewly skills setup transcribe-audio`
+(terminal) and Crewly installs what is missing — ffmpeg, whisper.cpp and the
+547 MB model (sha256-verified) — and reports "already satisfied" for the rest.
+On Linux, whisper.cpp comes from the official prebuilt Ubuntu release binaries
+(`install-whisper-cpp.sh`, pinned + sha256-checked) with a cmake source build as
+the fallback; it lands in `~/.crewly/bin/whisper-cli`.
+
+**When something is missing** the skill fails with machine-readable JSON:
+
+```json
+{"success":false,"error":"ffmpeg is required but not installed","needsSetup":true,
+ "skill":"transcribe-audio","missing":["ffmpeg"],"hint":"Run install-skill --id transcribe-audio …"}
+```
+
+On `needsSetup:true`, tell the user in one line that you are installing the
+transcription skill (a few minutes), run `install-skill --id transcribe-audio`,
+and continue when the `[SKILL INSTALLED]` message arrives. Never just reply
+"whisper.cpp is not installed".
+
+Manual details:
+
 - **`ffmpeg`** — required for both engines (audio is normalized to 16 kHz mono WAV).
   Install: `brew install ffmpeg`.
 - **Local engine (`whisper.cpp`)** — needs the `whisper-cli` binary and a model file:
-  - Binary: `brew install whisper-cpp` (provides `whisper-cli`). Override with `FLOPOST_WHISPER_BIN`.
+  - Binary: `brew install whisper-cpp` (provides `whisper-cli`); on Linux `~/.crewly/bin/whisper-cli`. Override with `FLOPOST_WHISPER_BIN`.
   - Model: `ggml-large-v3-turbo-q5_0.bin` in `~/.flopost/whisper/` or `~/.cache/whisper-models/`. Override with `FLOPOST_WHISPER_MODEL`.
   - If the binary or model is missing, the skill falls back to OpenAI (or reports a clear hint when `engine:"local"` is forced).
 - **OpenAI engine** — needs an OpenAI API key. Resolution order:

package/config/skills/agent/transcribe-audio/execute.sh

@@ -15,6 +15,10 @@
 #   bash execute.sh '{"audioFile":"/x.wav","engine":"openai"}'
 #
 # Requires: ffmpeg, jq, curl. Local engine also needs whisper-cli + a model file.
+# All of it is declared in skill.json's `setup` block. When something is
+# missing the error JSON carries `"needsSetup": true, "skill": "transcribe-audio"`
+# so an agent can go straight to `install-skill --id transcribe-audio`
+# instead of telling the user "whisper.cpp is not installed".
 
 set -euo pipefail
 
@@ -25,14 +29,29 @@ WHISPER_CACHE_DIR="${HOME}/.cache/whisper-models"
 OPENAI_TRANSCRIBE_URL="https://api.openai.com/v1/audio/transcriptions"
 OPENAI_MODEL="whisper-1"
 SETTINGS_URL="${CREWLY_API_URL:-http://localhost:${WEB_PORT:-8787}}/api/settings"
+SKILL_ID="transcribe-audio"
+# Crewly-managed binaries (the Linux whisper.cpp install lands here).
+CREWLY_BIN_DIR="${CREWLY_HOME:-${HOME}/.crewly}/bin"
 
 err_json() { printf '{"success":false,"error":%s}\n' "$(printf '%s' "$1" | jq -Rsa .)"; exit 1; }
 
+# needs_setup_json <message> <missing...> — a failure an install can fix.
+# Works without jq (jq itself may be the missing piece).
+needs_setup_json() {
+  local msg="$1"; shift
+  local missing="" item esc
+  for item in "$@"; do missing="${missing:+${missing},}\"${item}\""; done
+  esc=$(printf '%s' "$msg" | sed -e 's/\\/\\\\/g' -e 's/"/\\"/g' | tr '\n' ' ')
+  printf '{"success":false,"error":"%s","needsSetup":true,"skill":"%s","missing":[%s],"hint":"Run install-skill --id %s (it installs the missing pieces in the background and messages you when done), then retry."}\n' \
+    "$esc" "$SKILL_ID" "$missing" "$SKILL_ID"
+  exit 1
+}
+
 # ── Input parsing ───────────────────────────────────────────────────────────────
 INPUT="${1:-}"
 [ -z "$INPUT" ] && err_json "Usage: execute.sh '{\"audioFile\":\"/path/to/audio.m4a\"}'"
 
-command -v jq >/dev/null 2>&1 || { echo '{"success":false,"error":"jq is required but not installed (brew install jq)"}'; exit 1; }
+command -v jq >/dev/null 2>&1 || needs_setup_json "jq is required but not installed" jq
 
 AUDIO_FILE=$(printf '%s' "$INPUT" | jq -r '.audioFile // empty')
 OUTPUT_FILE=$(printf '%s' "$INPUT" | jq -r '.outputFile // empty')
@@ -43,14 +62,18 @@ ENGINE=$(printf '%s' "$INPUT" | jq -r '.engine // "auto"')
 [ -f "$AUDIO_FILE" ] || err_json "audioFile not found: ${AUDIO_FILE}"
 case "$ENGINE" in auto|local|openai) ;; *) err_json "engine must be one of: auto, local, openai (got: ${ENGINE})";; esac
 
-command -v ffmpeg >/dev/null 2>&1 || err_json "ffmpeg is required but not installed (brew install ffmpeg)"
+command -v ffmpeg >/dev/null 2>&1 || needs_setup_json "ffmpeg is required but not installed" ffmpeg
 
 # ── Local whisper.cpp detection (lifted from whisperModule.ts) ───────────────────
 resolve_whisper_binary() {
   if [ -n "${FLOPOST_WHISPER_BIN:-}" ] && [ -x "${FLOPOST_WHISPER_BIN}" ]; then echo "${FLOPOST_WHISPER_BIN}"; return; fi
   if [ -x "${FLOPOST_WHISPER_DIR}/whisper-cli" ]; then echo "${FLOPOST_WHISPER_DIR}/whisper-cli"; return; fi
+  if [ -x "${CREWLY_BIN_DIR}/whisper-cli" ]; then echo "${CREWLY_BIN_DIR}/whisper-cli"; return; fi
   if command -v whisper-cli >/dev/null 2>&1; then command -v whisper-cli; return; fi
-  for p in /opt/homebrew/bin/whisper-cli /usr/local/bin/whisper-cli; do
+  # Homebrew locations (a backend started by launchd may lack them on PATH).
+  # TRANSCRIBE_WHISPER_BIN_CANDIDATES overrides the list (tests).
+  local candidates="${TRANSCRIBE_WHISPER_BIN_CANDIDATES-/opt/homebrew/bin/whisper-cli /usr/local/bin/whisper-cli}"
+  for p in $candidates; do
     [ -x "$p" ] && { echo "$p"; return; }
   done
   echo ""
@@ -61,11 +84,10 @@ resolve_whisper_model() {
   if [ -f "${WHISPER_CACHE_DIR}/${MODEL_FILENAME}" ]; then echo "${WHISPER_CACHE_DIR}/${MODEL_FILENAME}"; return; fi
   echo ""
 }
-binary_hint() {
-  case "$(uname -s)" in
-    Darwin) echo "brew install whisper-cpp" ;;
-    *) echo "Build whisper.cpp from https://github.com/ggerganov/whisper.cpp and put whisper-cli on PATH" ;;
-  esac
+# Names of the local-engine pieces that are missing (for needsSetup JSON).
+missing_local_parts() {
+  [ -n "$WHISPER_BIN" ] || printf '%s\n' whisper-cli
+  [ -n "$WHISPER_MODEL" ] || printf '%s\n' whisper-model
 }
 
 WHISPER_BIN="$(resolve_whisper_binary)"
@@ -82,7 +104,10 @@ resolve_openai_key() {
 # ── Choose engine ────────────────────────────────────────────────────────────────
 CHOSEN=""
 if [ "$ENGINE" = "local" ]; then
-  [ "$LOCAL_AVAILABLE" = true ] || err_json "engine=local requested but whisper.cpp is not installed. Binary: $( [ -n "$WHISPER_BIN" ] && echo found || echo "missing ($(binary_hint))" ); Model: $( [ -n "$WHISPER_MODEL" ] && echo found || echo "missing ${MODEL_FILENAME} (put it in ${FLOPOST_WHISPER_DIR}/ or set FLOPOST_WHISPER_MODEL)" )."
+  if [ "$LOCAL_AVAILABLE" != true ]; then
+    # shellcheck disable=SC2046 # word-splitting the list of missing parts is intended
+    needs_setup_json "engine=local requested but the local whisper.cpp engine is not set up (binary: $( [ -n "$WHISPER_BIN" ] && echo found || echo missing ); model ${MODEL_FILENAME}: $( [ -n "$WHISPER_MODEL" ] && echo found || echo missing ))" $(missing_local_parts)
+  fi
   CHOSEN="local"
 elif [ "$ENGINE" = "openai" ]; then
   CHOSEN="openai"
@@ -138,7 +163,7 @@ run_local() {
 # ── Engine: OpenAI Whisper API ───────────────────────────────────────────────────
 run_openai() {
   local key; key="$(resolve_openai_key)"
-  [ -z "$key" ] && { fail "OpenAI engine selected but no OpenAI API key found. Set OPENAI_API_KEY (Crewly secrets) or configure Settings > API Keys."; return 1; }
+  [ -z "$key" ] && { fail "OpenAI engine selected but no OpenAI API key found. Set OPENAI_API_KEY (Crewly secrets) or configure Settings > API Keys."; : > "${WORK}/no-openai-key"; return 1; }
   echo "{\"status\":\"transcribing\",\"engine\":\"openai-whisper-1\",\"message\":\"Uploading to OpenAI Whisper API...\"}" >&2
   local resp="${WORK}/openai.json"
   local lang_args=(); [ "$LANGUAGE" != "auto" ] && lang_args=(-F "language=${LANGUAGE}")
@@ -179,7 +204,15 @@ if [ "$CHOSEN" = "local" ]; then
     err_json "$(cat "${WORK}/engine.err" 2>/dev/null)"
   fi
 else
-  run_openai || err_json "$(cat "${WORK}/engine.err" 2>/dev/null)"
+  if ! run_openai; then
+    # auto mode with no local engine and no OpenAI key: installing the local
+    # engine fixes it, so say so in a way an agent can act on.
+    if [ "$ENGINE" = "auto" ] && [ -f "${WORK}/no-openai-key" ]; then
+      # shellcheck disable=SC2046
+      needs_setup_json "No transcription engine is available: the local whisper.cpp engine is not set up and there is no OpenAI API key" $(missing_local_parts)
+    fi
+    err_json "$(cat "${WORK}/engine.err" 2>/dev/null)"
+  fi
 fi
 
 [ -f "$RESULT_FILE" ] || err_json "Transcription produced no result"

package/config/skills/agent/transcribe-audio/execute.test.sh

@@ -0,0 +1,90 @@
+#!/bin/bash
+# Tests for transcribe-audio — run with: bash execute.test.sh </dev/null
+# Uses fake ffmpeg / ffprobe / whisper-cli on an isolated PATH and HOME, so it
+# never touches a real model or the network. Exit 0 on pass.
+set -u
+HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+EXEC="$HERE/execute.sh"
+PASS=0; FAIL=0
+check() {
+  local name="$1"; local got="$2"; local want="$3"
+  if [ "$got" = "$want" ]; then PASS=$((PASS+1)); else FAIL=$((FAIL+1)); echo "FAIL: $name"; echo "  want: $want"; echo "  got:  $got"; fi
+}
+
+JQ="$(command -v jq)"
+BASH_BIN="${TEST_BASH:-$(command -v bash)}"
+[ -n "$JQ" ] || { echo "jq is required to run these tests"; exit 1; }
+T="$(mktemp -d)"
+trap 'rm -rf "$T"' EXIT
+mkdir -p "$T/bin" "$T/home"
+# Minimal PATH: basic tools + jq, but no ffmpeg / whisper-cli.
+for tool in bash jq mktemp cat rm tail dirname mkdir sed tr sysctl nproc uname basename; do
+  src="$(command -v "$tool" 2>/dev/null)" && ln -sf "$src" "$T/bin/$tool"
+done
+echo "fake audio" > "$T/clip.m4a"
+
+run() {
+  env -i HOME="$T/home" PATH="$1" CREWLY_HOME="$T/home/.crewly" CREWLY_API_URL="http://127.0.0.1:9" \
+    TRANSCRIBE_WHISPER_BIN_CANDIDATES="" "$BASH_BIN" "$EXEC" "$2" 2>/dev/null </dev/null
+}
+
+# 1. ffmpeg missing → needsSetup, machine-readable.
+OUT=$(run "$T/bin" "{\"audioFile\":\"$T/clip.m4a\"}"); RC=$?
+check "no ffmpeg: exit 1" "$RC" "1"
+check "no ffmpeg: needsSetup" "$(printf '%s' "$OUT" | "$JQ" -r .needsSetup)" "true"
+check "no ffmpeg: skill" "$(printf '%s' "$OUT" | "$JQ" -r .skill)" "transcribe-audio"
+check "no ffmpeg: missing" "$(printf '%s' "$OUT" | "$JQ" -c .missing)" '["ffmpeg"]'
+check "no ffmpeg: hint names install-skill" "$(printf '%s' "$OUT" | "$JQ" -r '.hint | test("install-skill --id transcribe-audio")')" "true"
+
+# Fake ffmpeg (writes the last argument, the WAV) and ffprobe.
+cat > "$T/bin/ffmpeg" <<'EOF'
+#!/bin/bash
+for a in "$@"; do last="$a"; done
+echo "RIFF" > "$last"
+EOF
+printf '#!/bin/bash\necho 1.5\n' > "$T/bin/ffprobe"
+chmod 755 "$T/bin/ffmpeg" "$T/bin/ffprobe"
+
+# 2. engine=local without whisper-cli or model → needsSetup listing both.
+OUT=$(run "$T/bin" "{\"audioFile\":\"$T/clip.m4a\",\"engine\":\"local\"}"); RC=$?
+check "local missing: exit 1" "$RC" "1"
+check "local missing: needsSetup" "$(printf '%s' "$OUT" | "$JQ" -r .needsSetup)" "true"
+check "local missing: missing parts" "$(printf '%s' "$OUT" | "$JQ" -c .missing)" '["whisper-cli","whisper-model"]'
+
+# 3. auto engine, no local engine, no OpenAI key → needsSetup (not a bare error).
+ln -sf "$(command -v curl)" "$T/bin/curl"
+OUT=$(run "$T/bin" "{\"audioFile\":\"$T/clip.m4a\"}"); RC=$?
+check "auto no engine: exit 1" "$RC" "1"
+check "auto no engine: needsSetup" "$(printf '%s' "$OUT" | "$JQ" -r .needsSetup)" "true"
+check "auto no engine: message" "$(printf '%s' "$OUT" | "$JQ" -r '.error | test("no OpenAI API key")')" "true"
+
+# 4. engine=openai explicitly without a key → plain error, no needsSetup (installing does not help).
+OUT=$(run "$T/bin" "{\"audioFile\":\"$T/clip.m4a\",\"engine\":\"openai\"}")
+check "openai no key: not needsSetup" "$(printf '%s' "$OUT" | "$JQ" -r '.needsSetup // false')" "false"
+
+# 5. Everything present (whisper-cli in $CREWLY_HOME/bin, model in ~/.cache) → transcript.
+mkdir -p "$T/home/.crewly/bin" "$T/home/.cache/whisper-models"
+: > "$T/home/.cache/whisper-models/ggml-large-v3-turbo-q5_0.bin"
+cat > "$T/home/.crewly/bin/whisper-cli" <<'EOF'
+#!/bin/bash
+while [ $# -gt 0 ]; do case "$1" in -of) prefix="$2"; shift 2 ;; *) shift ;; esac; done
+printf '{"result":{"language":"zh"},"transcription":[{"offsets":{"from":0,"to":1500},"text":" 你好，今天开会"},{"offsets":{"from":1500,"to":1600},"text":" [BLANK_AUDIO]"}]}' > "${prefix}.json"
+EOF
+chmod 755 "$T/home/.crewly/bin/whisper-cli"
+OUT=$(run "$T/bin" "{\"audioFile\":\"$T/clip.m4a\",\"outputFile\":\"$T/out.md\"}"); RC=$?
+check "local ok: exit 0" "$RC" "0"
+check "local ok: success" "$(printf '%s' "$OUT" | "$JQ" -r .success)" "true"
+check "local ok: engine" "$(printf '%s' "$OUT" | "$JQ" -r .engine)" "whisper.cpp"
+check "local ok: text" "$(printf '%s' "$OUT" | "$JQ" -r .text)" "你好，今天开会"
+check "local ok: blank tag dropped" "$(printf '%s' "$OUT" | "$JQ" -r .segmentCount)" "1"
+check "local ok: markdown written" "$(grep -c '你好，今天开会' "$T/out.md")" "2"
+
+# 6. The shipped setup block is valid JSON and names the skill's own install script.
+check "setup: whisper linux script" "$("$JQ" -r '.setup.steps[] | select(.id=="whisper-cli") | .install.linux.script' "$HERE/skill.json")" "install-whisper-cpp.sh"
+[ -f "$HERE/install-whisper-cpp.sh" ] && OK=yes || OK=no
+check "setup: install script exists" "$OK" "yes"
+bash -n "$HERE/install-whisper-cpp.sh" && OK=yes || OK=no
+check "setup: install script parses" "$OK" "yes"
+
+echo "=== Results: $PASS passed, $FAIL failed ==="
+[ "$FAIL" -eq 0 ]

package/config/skills/agent/transcribe-audio/install-whisper-cpp.sh

@@ -0,0 +1,126 @@
+#!/bin/bash
+# =============================================================================
+# install-whisper-cpp.sh — put a working `whisper-cli` in $CREWLY_BIN_DIR on Linux.
+#
+# Run by the Crewly skill-setup runner (the `whisper-cli` step of this skill's
+# setup block) when whisper-cli is not already installed. Never prompts.
+#
+# 1. Prebuilt: the official whisper.cpp release ships Ubuntu binaries
+#    (whisper-bin-ubuntu-{x64,arm64}.tar.gz). The pinned tarball is downloaded,
+#    its sha256 checked, extracted to $CREWLY_HOME/opt/whisper.cpp-<tag>/, and a
+#    wrapper $CREWLY_BIN_DIR/whisper-cli is written that sets LD_LIBRARY_PATH to
+#    the bundled shared libraries. A smoke test (`whisper-cli --help`) decides
+#    whether the binary runs here (it needs a recent enough glibc).
+# 2. Source: when there is no prebuilt for this CPU, or it does not run, the
+#    same tag is built with cmake into $CREWLY_HOME/opt/ and whisper-cli copied
+#    to $CREWLY_BIN_DIR. Needs cmake, a C++ compiler, curl and tar; when they
+#    are missing and root/passwordless sudo is available ($CREWLY_SUDO), they
+#    are installed with apt-get, otherwise the script says what to install.
+#
+# Environment (set by the runner):
+#   CREWLY_HOME      Crewly home (default ~/.crewly)
+#   CREWLY_BIN_DIR   where whisper-cli goes (default $CREWLY_HOME/bin)
+#   CREWLY_SUDO      '' = root, 'sudo -n' = passwordless sudo, 'unavailable'
+#   WHISPER_CPP_SKIP_PREBUILT=1   force the source build (testing)
+#   WHISPER_CPP_DOWNLOAD_BASE     override the download host (testing)
+# =============================================================================
+set -euo pipefail
+
+# ── Pinned release (sha256 from the GitHub release asset digests) ─────────────
+WHISPER_TAG="v1.9.2"
+SHA256_UBUNTU_X64="46811a3ecf584307480a220b9ef5ff81b7b22dc41577cbc274ce3afc61f753b1"
+SHA256_UBUNTU_ARM64="7e26fa6a36d9174d5c0bf033ccbc026c3b5e569e2ee787058241346ef5392719"
+DOWNLOAD_BASE="${WHISPER_CPP_DOWNLOAD_BASE:-https://github.com/ggml-org/whisper.cpp}"
+BUILD_JOBS="$(nproc 2>/dev/null || echo 2)"
+
+CREWLY_HOME="${CREWLY_HOME:-${HOME}/.crewly}"
+BIN_DIR="${CREWLY_BIN_DIR:-${CREWLY_HOME}/bin}"
+OPT_DIR="${CREWLY_HOME}/opt"
+SUDO_MODE="${CREWLY_SUDO-unavailable}"
+
+log() { printf '[install-whisper-cpp] %s\n' "$*" >&2; }
+die() { log "ERROR: $*"; exit 1; }
+
+[ "$(uname -s)" = "Linux" ] || die "this installer is for Linux (on macOS the setup uses: brew install whisper-cpp)"
+mkdir -p "$BIN_DIR" "$OPT_DIR"
+
+WORK="$(mktemp -d "${TMPDIR:-/tmp}/whisper-cpp-XXXXXX")"
+trap 'rm -rf "$WORK"' EXIT
+
+sha256_of() {
+  if command -v sha256sum >/dev/null 2>&1; then sha256sum "$1" | cut -d' ' -f1
+  else shasum -a 256 "$1" | cut -d' ' -f1; fi
+}
+
+# Write the wrapper that makes the bundled .so files findable.
+write_wrapper() {
+  local real="$1" libdir="$2"
+  cat > "${BIN_DIR}/whisper-cli" <<EOF
+#!/bin/sh
+# Crewly wrapper for whisper.cpp ${WHISPER_TAG} (written by install-whisper-cpp.sh)
+LD_LIBRARY_PATH="${libdir}\${LD_LIBRARY_PATH:+:\$LD_LIBRARY_PATH}" exec "${real}" "\$@"
+EOF
+  chmod 755 "${BIN_DIR}/whisper-cli"
+}
+
+smoke_test() { "${BIN_DIR}/whisper-cli" --help >/dev/null 2>&1; }
+
+try_prebuilt() {
+  local arch asset sha
+  arch="$(uname -m)"
+  case "$arch" in
+    x86_64|amd64) asset="whisper-bin-ubuntu-x64.tar.gz"; sha="$SHA256_UBUNTU_X64" ;;
+    aarch64|arm64) asset="whisper-bin-ubuntu-arm64.tar.gz"; sha="$SHA256_UBUNTU_ARM64" ;;
+    *) log "no prebuilt binary for ${arch}"; return 1 ;;
+  esac
+  local url="${DOWNLOAD_BASE}/releases/download/${WHISPER_TAG}/${asset}"
+  log "downloading ${url}"
+  curl -fsSL --retry 3 -o "${WORK}/${asset}" "$url" || { log "download failed"; return 1; }
+  local got; got="$(sha256_of "${WORK}/${asset}")"
+  [ "$got" = "$sha" ] || { log "checksum mismatch for ${asset}: expected ${sha}, got ${got}"; return 1; }
+  local dest="${OPT_DIR}/whisper.cpp-${WHISPER_TAG}"
+  rm -rf "$dest"; mkdir -p "$dest"
+  tar -xzf "${WORK}/${asset}" -C "$dest" --strip-components=1
+  [ -x "${dest}/whisper-cli" ] || { log "whisper-cli not found in ${asset}"; return 1; }
+  write_wrapper "${dest}/whisper-cli" "$dest"
+  smoke_test || { log "the prebuilt whisper-cli does not run on this system (glibc too old?)"; return 1; }
+  log "installed prebuilt whisper.cpp ${WHISPER_TAG} → ${BIN_DIR}/whisper-cli"
+}
+
+ensure_build_tools() {
+  local missing=()
+  command -v cmake >/dev/null 2>&1 || missing+=(cmake)
+  { command -v c++ >/dev/null 2>&1 || command -v g++ >/dev/null 2>&1; } || missing+=(build-essential)
+  command -v curl >/dev/null 2>&1 || missing+=(curl)
+  [ ${#missing[@]} -eq 0 ] && return 0
+  if [ "$SUDO_MODE" = "unavailable" ] || ! command -v apt-get >/dev/null 2>&1; then
+    die "building whisper.cpp needs: ${missing[*]}. Install them (e.g. sudo apt-get install -y ${missing[*]}) and run the setup again."
+  fi
+  log "installing build tools: ${missing[*]}"
+  # shellcheck disable=SC2086 # SUDO_MODE is '' or 'sudo -n'
+  $SUDO_MODE env DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends "${missing[@]}" </dev/null \
+    || { $SUDO_MODE apt-get update </dev/null && $SUDO_MODE env DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends "${missing[@]}" </dev/null; } \
+    || die "apt-get could not install ${missing[*]}"
+}
+
+build_from_source() {
+  ensure_build_tools
+  local url="${DOWNLOAD_BASE}/archive/refs/tags/${WHISPER_TAG}.tar.gz"
+  log "building whisper.cpp ${WHISPER_TAG} from source (${url})"
+  curl -fsSL --retry 3 -o "${WORK}/src.tar.gz" "$url" || die "could not download ${url}"
+  mkdir -p "${WORK}/src"
+  tar -xzf "${WORK}/src.tar.gz" -C "${WORK}/src" --strip-components=1
+  cmake -S "${WORK}/src" -B "${WORK}/build" -DCMAKE_BUILD_TYPE=Release -DBUILD_SHARED_LIBS=OFF -DWHISPER_BUILD_TESTS=OFF >&2
+  cmake --build "${WORK}/build" --config Release --target whisper-cli -j "$BUILD_JOBS" >&2
+  local dest="${OPT_DIR}/whisper.cpp-${WHISPER_TAG}-src"
+  rm -rf "$dest"; mkdir -p "$dest"
+  cp "${WORK}/build/bin/whisper-cli" "${dest}/whisper-cli"
+  write_wrapper "${dest}/whisper-cli" "$dest"
+  smoke_test || die "the whisper-cli built from source does not run"
+  log "built whisper.cpp ${WHISPER_TAG} → ${BIN_DIR}/whisper-cli"
+}
+
+if [ "${WHISPER_CPP_SKIP_PREBUILT:-0}" != "1" ] && try_prebuilt; then
+  exit 0
+fi
+build_from_source

package/config/skills/agent/transcribe-audio/install-whisper-cpp.test.sh

@@ -0,0 +1,69 @@
+#!/bin/bash
+# Tests for install-whisper-cpp.sh — run with: bash install-whisper-cpp.test.sh </dev/null
+# Fakes uname (Linux x86_64), curl (serves a crafted release tarball) and
+# sha256sum on an isolated PATH; nothing is downloaded or built. Exit 0 on pass.
+set -u
+HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+SCRIPT="$HERE/install-whisper-cpp.sh"
+PASS=0; FAIL=0
+check() {
+  local name="$1"; local got="$2"; local want="$3"
+  if [ "$got" = "$want" ]; then PASS=$((PASS+1)); else FAIL=$((FAIL+1)); echo "FAIL: $name"; echo "  want: $want"; echo "  got:  $got"; fi
+}
+BASH_BIN="${TEST_BASH:-$(command -v bash)}"
+T="$(mktemp -d)"; trap 'rm -rf "$T"' EXIT
+mkdir -p "$T/bin" "$T/pkg/whisper-bin-ubuntu-x64"
+for tool in mktemp cat rm tar mkdir chmod cp cut dirname grep sed head; do
+  src="$(command -v "$tool" 2>/dev/null)" && ln -sf "$src" "$T/bin/$tool"
+done
+EXPECTED_SHA="$(grep '^SHA256_UBUNTU_X64=' "$SCRIPT" | cut -d'"' -f2)"
+
+# The release tarball: whisper-cli that answers --help.
+printf '#!/bin/sh\n[ "$1" = "--help" ] && echo "usage: whisper-cli" && exit 0\necho "ran $*"\n' > "$T/pkg/whisper-bin-ubuntu-x64/whisper-cli"
+chmod 755 "$T/pkg/whisper-bin-ubuntu-x64/whisper-cli"
+(cd "$T/pkg" && tar -czf "$T/release.tar.gz" whisper-bin-ubuntu-x64)
+
+printf '#!/bin/sh\ncase "$1" in -s) echo Linux ;; -m) echo x86_64 ;; *) echo Linux ;; esac\n' > "$T/bin/uname"
+printf '#!/bin/sh\necho 4\n' > "$T/bin/nproc"
+# curl -fsSL --retry 3 -o <out> <url>: log the URL, copy the tarball.
+cat > "$T/bin/curl" <<EOT
+#!/bin/sh
+while [ \$# -gt 0 ]; do case "\$1" in -o) out="\$2"; shift 2 ;; -*) shift ;; *) url="\$1"; shift ;; esac; done
+echo "\$url" >> "$T/curl.log"
+cp "$T/release.tar.gz" "\$out"
+EOT
+cat > "$T/bin/sha256sum" <<'EOT'
+#!/bin/sh
+echo "${FAKE_SHA}  $1"
+EOT
+chmod 755 "$T/bin/"*
+
+run() { env -i HOME="$T/home" PATH="$T/bin" CREWLY_HOME="$T/home/.crewly" CREWLY_SUDO=unavailable "$@" "$BASH_BIN" "$SCRIPT" </dev/null 2>"$T/err.log"; }
+
+# 1. Prebuilt: checksum matches → extracted, wrapper written, smoke test passes.
+run FAKE_SHA="$EXPECTED_SHA"; RC=$?
+check "prebuilt: exit 0" "$RC" "0"
+check "prebuilt: pinned release URL" "$(tail -n 1 "$T/curl.log")" "https://github.com/ggml-org/whisper.cpp/releases/download/v1.9.2/whisper-bin-ubuntu-x64.tar.gz"
+check "prebuilt: wrapper is executable" "$([ -x "$T/home/.crewly/bin/whisper-cli" ] && echo yes)" "yes"
+check "prebuilt: wrapper runs the extracted binary" "$(PATH="$T/bin" "$T/home/.crewly/bin/whisper-cli" -m x)" "ran -m x"
+check "prebuilt: sets LD_LIBRARY_PATH to the bundle" "$(grep -c "LD_LIBRARY_PATH=\"$T/home/.crewly/opt/whisper.cpp-v1.9.2" "$T/home/.crewly/bin/whisper-cli")" "1"
+
+# 2. Checksum mismatch → prebuilt rejected, source build needs tools; no sudo → clear failure.
+rm -rf "$T/home"; : > "$T/curl.log"
+run FAKE_SHA=0000; RC=$?
+check "mismatch: exit 1" "$RC" "1"
+grep -q "checksum mismatch for whisper-bin-ubuntu-x64.tar.gz" "$T/err.log" && OK=yes || OK=no
+check "mismatch: says so" "$OK" "yes"
+grep -q "building whisper.cpp needs: cmake build-essential. Install them (e.g. sudo apt-get install -y cmake build-essential)" "$T/err.log" && OK=yes || OK=no
+check "no build tools + no sudo: tells the owner what to install" "$OK" "yes"
+check "mismatch: no whisper-cli left behind" "$([ -e "$T/home/.crewly/bin/whisper-cli" ] && echo present || echo absent)" "absent"
+
+# 3. Not Linux → refuses (macOS uses brew).
+printf '#!/bin/sh\necho Darwin\n' > "$T/bin/uname"
+run FAKE_SHA="$EXPECTED_SHA"; RC=$?
+check "darwin: exit 1" "$RC" "1"
+grep -q "this installer is for Linux" "$T/err.log" && OK=yes || OK=no
+check "darwin: says brew is used there" "$OK" "yes"
+
+echo "=== Results: $PASS passed, $FAIL failed ==="
+[ "$FAIL" -eq 0 ]