@hasna/skills 0.1.46 → 0.1.48

package/skills/transcript/SKILL.md

@@ -1,100 +1,80 @@
 ---
 name: transcript
-description: Transcribe audio and video files using ElevenLabs Scribe, OpenAI Whisper, or Google Gemini. Supports automatic chunking for large files, speaker diarization, timestamps, and multiple output formats (text, SRT, VTT, JSON).
+description: Transcribe audio, video, YouTube, Vimeo, and generic media URLs with iapp-transcriber or the hosted Skills runtime. Supports OpenAI GPT-4o transcription, OpenAI diarization, ElevenLabs Scribe v2, DeepGram, chunking, source metadata, subtitles, and JSON outputs.
 ---
 
-# Audio Transcription Skill
+# Transcript
 
-This skill provides high-quality speech-to-text transcription using multiple AI providers. It automatically handles large files through compression and chunking.
+Create transcripts from local audio/video files or media URLs. Use this skill when the user asks to transcribe, caption, diarize, summarize, or package spoken audio/video content.
 
-This CLI is API-backed. Set `SKILLS_API_KEY` when routing through the hosted skills/connectors runtime; provider-specific keys are managed by that runtime.
+## Choose The Runtime
 
-## Supported Providers
+- Use the hosted Skills runtime when the user explicitly runs `skills run transcript`, needs remote execution, or has only `SKILLS_API_KEY` configured.
+- Use local `iapp-transcriber` when you are on this machine and need direct access to local files, YouTube/Vimeo/generic `yt-dlp` sources, transcript DB records, MCP tools, comments, exports, or OpenLoops follow-up workflows.
+- The local command is `transcriber` when installed, or `bun run src/cli/index.ts` from `/home/hasna/Workspace/hasnaxyz/internalapp/iapp-transcriber`.
 
-### ElevenLabs Scribe
-- **Accuracy**: 96.7% for English (industry-leading)
-- **Max file size**: 3GB / 10 hours
-- **Features**: Speaker diarization (up to 32 speakers), word-level timestamps
-- **Cost**: $0.40/hour
-- **Best for**: Multi-speaker recordings, highest accuracy needs
+## Hosted Usage
 
-### OpenAI Whisper
-- **Accuracy**: Excellent
-- **Max file size**: 25MB (automatic chunking for larger files)
-- **Features**: Segment timestamps, language detection
-- **Cost**: $0.006/min ($0.003/min with GPT-4o Mini)
-- **Best for**: Standard transcription, good balance of cost and quality
+```bash
+skills run transcript --source ./meeting.mp3 --title "Design review" --provider openai
+skills run transcribe --source https://www.youtube.com/watch?v=... --provider openai --diarize
+```
 
-### Google Gemini
-- **Accuracy**: Very good
-- **Max file size**: 2GB
-- **Features**: Multimodal analysis, summarization capabilities
-- **Cost**: ~$0.09-0.23/hour (generous free tier available)
-- **Best for**: Cost-sensitive projects, multimodal needs
+Poll hosted runs with `skills runs status <run-id>` and download outputs with `skills exports download <run-id>`.
 
-## Usage
+## Local Usage
 
-### Basic Transcription
 ```bash
-bun run src/index.ts transcribe \
-  --provider openai \
-  --input ./recording.mp3
+transcriber transcribe ./meeting.mp3 --provider openai --json
+transcriber transcribe https://www.youtube.com/watch?v=... --provider openai --model gpt-4o-transcribe --json
+transcriber transcribe ./meeting.mp3 --provider openai --diarize --json
+transcriber export <transcript-id> --format srt --output captions.srt
 ```
 
-### With Speaker Diarization
-```bash
-bun run src/index.ts transcribe \
-  --provider elevenlabs \
-  --input ./meeting.mp3 \
-  --diarize \
-  --timestamps \
-  --format srt
-```
+Local provider defaults:
 
-### Export to Subtitles
-```bash
-bun run src/index.ts transcribe \
-  --provider gemini \
-  --input ./video.mp4 \
-  --format vtt \
-  --output ./captions.vtt
-```
+- `openai`: default, uses `gpt-4o-transcribe`; `--diarize` uses `gpt-4o-transcribe-diarize`.
+- `elevenlabs`: uses `scribe_v2`, supports diarization and keyterms.
+- `deepgram`: uses Nova-3, supports diarization.
 
-### View Provider Info
-```bash
-bun run src/index.ts providers
-```
+Local requirements:
 
-## Output Formats
+- A configured provider credential for the selected local provider.
+- `yt-dlp` for remote media URLs. Set `YTDLP_PATH` if needed.
+- `ffmpeg`/`ffprobe`; the local app bundles npm ffmpeg/ffprobe and also respects `FFMPEG_PATH` and `FFPROBE_PATH`.
 
-| Format | Extension | Description |
-|--------|-----------|-------------|
-| text | .txt | Plain text transcript |
-| srt | .srt | SubRip subtitle format |
-| vtt | .vtt | WebVTT subtitle format |
-| json | .json | Full structured data with metadata |
+## Workflow
 
-## Large File Handling
+1. Inspect source metadata first for URLs:
 
-The skill automatically handles files larger than provider limits:
+   ```bash
+   transcriber info <url> --json
+   ```
 
-- **Compression**: For OpenAI, files are first compressed using Opus codec
-- **Chunking**: Files are split into 10-minute segments with overlap
-- **Merging**: Results are intelligently merged to avoid duplicates
+2. Download audio when the user asks to keep media:
 
-## Configuration
+   ```bash
+   transcriber download <url> --format mp3 --json
+   ```
 
-```bash
-export SKILLS_API_KEY=your_skill_api_key
-```
+3. Transcribe with JSON for automation:
 
-## Dependencies
+   ```bash
+   transcriber transcribe <path-or-url> --provider openai --json
+   ```
 
-For chunking support (OpenAI with large files):
-- `ffmpeg` - Audio processing
-- `ffprobe` - Duration detection
+4. Export or post-process:
 
-Install on macOS:
-```bash
-brew install ffmpeg
-```
+   ```bash
+   transcriber get <id> --json
+   transcriber export <id> --format txt --output transcript.txt
+   transcriber summarize <id>
+   ```
+
+5. For repeat work, create OpenLoops command loops around JSON-producing commands, for example `transcriber feed check --json --dry-run`.
+
+## Safety
+
+- Only fetch URLs the user is authorized to process.
+- The local app rejects private/local URL hosts by default; set `TRANSCRIBER_ALLOW_PRIVATE_URLS=1` only for trusted internal sources.
+- Prefer `--json` for scripts and OpenLoops so failures include a structured transcript record and nonzero exit code.

package/skills/transcript/package.json

@@ -2,7 +2,7 @@
   "name": "transcript",
   "version": "1.0.0",
   "private": true,
-  "description": "Audio/video transcription skill supporting ElevenLabs, OpenAI Whisper, and Google Gemini with automatic chunking for large files",
+  "description": "Audio/video transcription skill backed by iapp-transcriber, OpenAI GPT-4o transcription, ElevenLabs Scribe v2, and hosted Skills runtime",
   "type": "module",
   "skills": {
     "runtime": "hosted",

package/skills/apidocs/.claude/settings.json

@@ -1,5 +0,0 @@
-{
-  "env": {
-    "CLAUDE_CODE_TASK_LIST_ID": "service-apidocs-dev"
-  }
-}

package/skills/hook/bunfig.toml

@@ -1,5 +0,0 @@
-[install]
-peer = false
-
-[run]
-bun = true

package/skills/implementation/bunfig.toml

@@ -1,5 +0,0 @@
-[install]
-peer = false
-
-[run]
-bun = true