📦 Wyoming Event Format

Wyoming is a streaming peer-to-peer protocol using newline-delimited JSON (JSONL) plus optional binary data.

It is designed to be:

• 🧩 Simple to implement
• 🔄 Streaming-friendly (audio chunking)
• 🔐 Language-agnostic and open

---

🔠 Message Structure

Each Wyoming message is made of:

{ "type": "...", "data_length": ..., "payload_length": ... }\n
<data_length bytes>       (optional, UTF-8 JSON)
<payload_length bytes>    (optional, binary payload)

• Header (required): JSON object, UTF-8, newline terminated
• Data section (optional): UTF-8 JSON
• Payload (optional): binary blob (e.g., PCM audio)

---

🧪 Minimal Example (Audio Chunk)

{ "type": "audio-chunk", "data_length": 0, "payload_length": 3200 }

Followed by 3200 bytes of raw PCM audio.

---

🔄 With Additional Metadata

{ "type": "transcribe", "data_length": 45, "payload_length": 0 }

Followed by:

{ "name": "whisper", "language": "en" }

No binary payload in this case.

---

🧰 Python Example: Creating an Event

from wyoming.event import Event

event = Event(
    type="synthesize",
    data={"text": "Hello world"},
    payload=b""
)

raw_bytes = event.to_bytes()

---

🧰 Python Example: Reading an Event from a Stream

from wyoming.event import read_event

async def handle(reader):
    event = await read_event(reader)
    print(event.type, event.data)

---

🧪 CLI Test (Netcat)

To verify a Wyoming-compatible server is running:

echo '{"type":"describe"}' | nc localhost 10200

Expect response like:

{ "type": "info", ... }

---

🌐 Format Summary

Component	Description
type	Event type string (e.g. transcribe)
data_length	Length of optional JSON block in bytes
payload_length	Length of optional binary in bytes

Wyoming works over: - TCP (tcp://host:port) - Unix sockets (unix:///path) - Standard I/O (stdio://)