Pipelines

Execute full voice pipelines combining STT, LLM, and TTS in a single request.

Endpoint

POST /v1/audio/pipelines

Authentication

Authorization: Bearer gw_prod_...

Request

Headers

Header	Required	Description
Authorization	Yes	Bearer token with API key
Content-Type	Yes	multipart/form-data or application/json

Body Parameters

Parameter	Type	Required	Description
audio	file/string	Yes	Audio file or base64-encoded audio
template	string	Yes	Pipeline template ID
context	string	No	System prompt for LLM
stream	boolean	No	Enable streaming response
overrides	object	No	Override template settings

Pipeline Templates

Template	Use Case	Components
voice-agent-fast	Real-time assistant	Voxtral → GPT-5-mini → ElevenLabs Turbo
voice-agent-premium	High-quality assistant	Whisper → GPT-5.2 → ElevenLabs v2
ambient-scribe	Medical transcription	Whisper → GPT-5.2 + PII detection
legal-dictation	Legal documents	Whisper → GPT-5.2 + formatting

Examples

Basic Pipeline

curl -X POST https://api.gateflow.ai/v1/audio/pipelines \
  -H "Authorization: Bearer gw_prod_..." \
  -F "audio=@question.mp3" \
  -F "template=voice-agent-fast" \
  -F "context=You are a helpful assistant."

Python

import requests
import base64

# With file upload
with open("question.mp3", "rb") as f:
    response = requests.post(
        "https://api.gateflow.ai/v1/audio/pipelines",
        headers={"Authorization": "Bearer gw_prod_..."},
        files={"audio": f},
        data={
            "template": "voice-agent-fast",
            "context": "You are a helpful assistant."
        }
    )

result = response.json()
print(f"Transcription: {result['transcription']}")
print(f"Response: {result['response']}")

# Save audio response
audio_bytes = base64.b64decode(result["audio"])
with open("response.mp3", "wb") as f:
    f.write(audio_bytes)

With Base64 Audio

import base64

with open("question.mp3", "rb") as f:
    audio_b64 = base64.b64encode(f.read()).decode()

response = requests.post(
    "https://api.gateflow.ai/v1/audio/pipelines",
    headers={
        "Authorization": "Bearer gw_prod_...",
        "Content-Type": "application/json"
    },
    json={
        "audio": audio_b64,
        "template": "voice-agent-fast",
        "context": "You are a helpful assistant."
    }
)

Streaming Pipeline

import httpx
import json
import base64

async def stream_pipeline(audio_file):
    with open(audio_file, "rb") as f:
        audio_b64 = base64.b64encode(f.read()).decode()

    async with httpx.AsyncClient() as client:
        async with client.stream(
            "POST",
            "https://api.gateflow.ai/v1/audio/pipelines",
            headers={
                "Authorization": "Bearer gw_prod_...",
                "Content-Type": "application/json"
            },
            json={
                "audio": audio_b64,
                "template": "voice-agent-fast",
                "stream": True
            }
        ) as response:
            async for line in response.aiter_lines():
                event = json.loads(line)

                if event["type"] == "transcription.partial":
                    print(f"Hearing: {event['text']}")
                elif event["type"] == "transcription.final":
                    print(f"You said: {event['text']}")
                elif event["type"] == "llm.token":
                    print(event["token"], end="", flush=True)
                elif event["type"] == "audio.chunk":
                    audio = base64.b64decode(event["data"])
                    play_audio(audio)
                elif event["type"] == "done":
                    print("\n--- Complete ---")

Template Overrides

response = requests.post(
    "https://api.gateflow.ai/v1/audio/pipelines",
    headers={
        "Authorization": "Bearer gw_prod_...",
        "Content-Type": "application/json"
    },
    json={
        "audio": audio_b64,
        "template": "voice-agent-fast",
        "overrides": {
            "llm": {
                "model": "claude-sonnet-4-5-20250929",
                "max_tokens": 300
            },
            "tts": {
                "voice": "professional",
                "speed": 1.1
            }
        }
    }
)

Medical Scribe

response = requests.post(
    "https://api.gateflow.ai/v1/audio/pipelines",
    headers={"Authorization": "Bearer gw_prod_..."},
    files={"audio": open("patient_visit.mp3", "rb")},
    data={"template": "ambient-scribe"}
)

result = response.json()
print(f"Transcription: {result['transcription']}")
print(f"Clinical Note: {result['structured_note']}")
print(f"PII Detected: {result['pii_detected']}")

Response

Standard Response

{
  "id": "pipeline_abc123",
  "transcription": "What's the weather like today?",
  "response": "I don't have access to real-time weather data, but I can help you find a weather service or app that can provide current conditions for your location.",
  "audio": "base64_encoded_audio_data...",
  "audio_format": "mp3",
  "usage": {
    "stt_seconds": 3.5,
    "llm_tokens": {"prompt": 45, "completion": 52},
    "tts_characters": 156
  },
  "latency": {
    "stt_ms": 450,
    "llm_ms": 380,
    "tts_ms": 320,
    "total_ms": 1150
  },
  "providers": {
    "stt": "mistral",
    "llm": "openai",
    "tts": "elevenlabs"
  }
}

Ambient Scribe Response

{
  "id": "pipeline_abc123",
  "transcription": "Patient reports headache for three days, took ibuprofen with minimal relief...",
  "structured_note": {
    "chief_complaint": "Headache",
    "duration": "3 days",
    "symptoms": ["headache", "fatigue"],
    "medications_mentioned": ["ibuprofen"],
    "allergies_mentioned": [],
    "vital_signs": null
  },
  "pii_detected": [
    {"type": "patient_name", "value": "[REDACTED]", "confidence": 0.95}
  ],
  "audit_id": "audit_xyz789",
  "compliance": {
    "hipaa": true,
    "audit_logged": true
  }
}

Streaming Events

Event Type	Description	Data
transcription.partial	Partial transcript	{text}
transcription.final	Final transcript	{text}
llm.token	LLM token	{token}
llm.done	LLM complete	{text}
audio.chunk	TTS audio chunk	{data} (base64)
done	Pipeline complete	{usage, latency}
error	Error occurred	{message, recoverable}

Custom Pipelines

Create Custom Template

curl -X POST https://api.gateflow.ai/v1/management/pipeline-templates \
  -H "Authorization: Bearer gw_prod_admin_key" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "my-custom-pipeline",
    "stt": {
      "model": "whisper-1",
      "language": "en"
    },
    "llm": {
      "model": "gpt-5.2",
      "system_prompt": "You are a customer service agent.",
      "max_tokens": 200,
      "temperature": 0.7
    },
    "tts": {
      "model": "eleven_turbo_v2_5",
      "voice": "friendly"
    }
  }'

List Templates

curl https://api.gateflow.ai/v1/management/pipeline-templates \
  -H "Authorization: Bearer gw_prod_..."

Errors

Code	Description
400	Invalid audio format or parameters
401	Invalid API key
404	Template not found
413	Audio file too large
429	Rate limit exceeded
500	Pipeline execution error

See Also

• Pipeline Templates - Template configuration
• Streaming Speech - Streaming guide
• Transcriptions - STT API
• Speech Synthesis - TTS API