ai-cast-cli

Publish AI-generated podcasts to Sui blockchain + Walrus decentralized storage.

Zero-install CLI for AI agents — run with npx, get JSON output, no interaction needed.

For Agents

# No install needed. Just run:
npx ai-cast-cli --json <command>

Every command outputs structured JSON when --json is set. Parse status field for success/failure.

First-time setup (run once)

# Step 1: Check environment (TTS model, ffmpeg, etc.)
npx ai-cast-cli install --check

# Step 2: Install missing dependencies (TTS model ~500MB)
npx ai-cast-cli install

# Step 3: Connect Sui wallet
npx ai-cast-cli init --package-id 0x10c32bf076865c211bec10e170e2640d08e3515a957754cfdeac890b5a7f2214

# Step 4: Create creator profile
npx ai-cast-cli --json profile create --name "AI Weekly" --bio "AI-powered podcast" --category tech
# → {"status":"ok","objectId":"0x...","digest":"..."}

Generate and publish a podcast (daily workflow)

# Step 1: Fetch article(s)
npx ai-cast-cli --json fetch -u https://example.com/article -o /tmp/articles.json
# → {"status":"ok","output":"/tmp/articles.json","count":1,"articles":[{"title":"...","chars":4500}]}

# Step 2: Generate script (requires KIMI_API_KEY env var)
KIMI_API_KEY=sk-xxx npx ai-cast-cli --json script -i /tmp/articles.json -s deep_dive -o /tmp/script.txt
# → {"status":"ok","output":"/tmp/script.txt","chars":2800,"style":"deep_dive"}

# Step 3: Text-to-speech (requires MLX TTS model installed via `install` command)
npx ai-cast-cli --json speak -i /tmp/script.txt -v serena -o /tmp/episode.wav
# → {"status":"ok","output":"/tmp/episode.wav","voice":"serena"}

# Step 4: Publish to Sui/Walrus
npx ai-cast-cli --json publish -a /tmp/episode.wav -t "Article Title" -s /tmp/script.txt --tags ai,tech --source-url https://example.com/article
# → {"status":"ok","podcastId":"0x...","audioBlobId":"...","digest":"..."}

Batch publish (multiple URLs)

# Create a file with one URL per line
echo "https://article1.com
https://article2.com
https://article3.com" > urls.txt

# Publish each URL as a separate podcast
npx ai-cast-cli --json batch -f urls.txt --style deep_dive --voice serena --tags ai

# Or aggregate all into one podcast
npx ai-cast-cli --json batch -f urls.txt --aggregate --style news --tags weekly

Query data

# List my podcasts
npx ai-cast-cli --json list
# → {"status":"ok","podcasts":[{"objectId":"0x...","title":"...","audioBlobId":"..."}]}

# Check balance and earnings
npx ai-cast-cli --json balance
# → {"status":"ok","suiBalance":0.89,"tipsReceived":0.5,"subscriptionIncome":0.4}

Environment Variables

Skip interactive init by setting env vars (ideal for agents and CI/CD):

| Variable | Required | Description | |---|---|---| | AI_CAST_PACKAGE_ID | Yes | Smart contract package ID on Sui | | AI_CAST_KEYSTORE | Yes | Path to Sui keystore file | | KIMI_API_KEY | For script | LLM API key for podcast script generation | | AI_CAST_NETWORK | No | testnet (default) / mainnet / devnet | | AI_CAST_ADDRESS | No | Active wallet address (auto-detected from keystore) |

# Full agent usage with env vars — no init needed
AI_CAST_PACKAGE_ID=0x10c32bf076865c211bec10e170e2640d08e3515a957754cfdeac890b5a7f2214 \
AI_CAST_KEYSTORE=~/.sui/sui_config/sui.keystore \
KIMI_API_KEY=sk-xxx \
npx ai-cast-cli --json fetch -u https://example.com && \
npx ai-cast-cli --json script -i articles.json && \
npx ai-cast-cli --json speak -i script.txt && \
npx ai-cast-cli --json publish -a /tmp/podcast_*.wav -t "Auto Episode"

All Commands

| Command | Description | Key Options | |---|---|---| | install | Check/install runtime environment | --check (check only) | | fetch | Fetch article content | -u <url> (multiple allowed) | | script | Generate podcast script via LLM | -i <input>, -s <style> | | speak | Text-to-speech synthesis | -i <input>, -v <voice> | | publish | Upload to Walrus + register on Sui | -a <audio>, -t <title>, --tags | | batch | Batch generate + publish | -f <file>, --aggregate | | init | Configure wallet connection | --package-id <id> | | profile create | Create on-chain creator profile | --name, --category | | list | List published podcasts | | | balance | Show SUI balance + earnings | |

Options for publish

| Option | Description | |---|---| | -a, --audio <path> | Audio file (required). WAV auto-converts to Opus | | -t, --title <title> | Podcast title (required) | | -s, --script <path> | Transcript file | | -d, --description | Description text | | --style <style> | deep_dive / news / story / interview | | --tags <tags> | Comma-separated tags: ai,web3,defi | | --tier <tier> | free (default) / premium (SEAL encrypted) | | --source-url <url> | Original article URL | | --retention <n> | Walrus storage epochs (default: 5) |

Voices for speak

| Voice | Gender | |---|---| | serena (default) | Female | | vivian | Female | | sohee | Female | | ryan | Male | | aiden | Male | | eric | Male | | dylan | Male |

Styles for script

| Style | Description | |---|---| | deep_dive (default) | In-depth analysis, logical structure | | news | Fast-paced, information-dense | | story | Narrative, emotionally engaging | | interview | Conversational, casual |

Categories for profile create

tech / finance / news / culture / education / entertainment

Architecture

Article URL → fetch → LLM script → TTS audio → ffmpeg (WAV→Opus) → Walrus upload → Sui on-chain

• Sui: Creator profiles, podcast metadata, subscriptions, tips, SEAL policies
• Walrus: Audio files, transcripts (decentralized blob storage)
• SEAL: Encryption for premium/paywalled content

JSON Output Schema

All commands with --json return:

// Success
{"status": "ok", "key": "value", ...}

// Error
{"error": "error message", "details": "additional info"}

License

MIT