Appearance

Custom Templates ​

Create custom pipeline templates for your specific use cases.

Overview ​

Custom templates let you define reusable pipeline configurations with specific models, prompts, and settings.

Template Structure ​

yaml
name: my-custom-template
description: Template for customer support voice interactions
version: "1.0.0"

# Speech-to-Text configuration
stt:
  model: whisper-1
  language: en
  streaming: false
  word_timestamps: true

# Language Model configuration
llm:
  model: gpt-5.2
  system_prompt: |
    You are a customer support assistant for TechCorp.

    Guidelines:
    - Be helpful and concise
    - Offer to escalate complex issues
    - Never share internal procedures
  max_tokens: 200
  temperature: 0.7
  tools: []

# Text-to-Speech configuration
tts:
  model: eleven_turbo_v2_5
  voice: friendly
  streaming: true
  voice_settings:
    stability: 0.5
    similarity_boost: 0.75

# Post-processing
processing:
  detect_pii: false
  format_output: text

# Compliance settings
compliance:
  audit_log: true
  data_classification: internal

# Cost limits
limits:
  max_audio_seconds: 300
  max_response_tokens: 500

Creating Templates ​

Via API ​

bash
curl -X POST https://api.gateflow.ai/v1/mcp/pipeline-templates \
  -H "Authorization: Bearer gw_prod_admin_key" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "support-voice-bot",
    "description": "Voice bot for customer support",
    "stt": {
      "model": "voxtral-mini-latest",
      "streaming": true
    },
    "llm": {
      "model": "gpt-5-mini",
      "system_prompt": "You are a helpful support assistant.",
      "max_tokens": 150,
      "temperature": 0.7
    },
    "tts": {
      "model": "eleven_turbo_v2_5",
      "voice": "friendly"
    }
  }'

Response:

json
{
  "template_id": "template_xyz789",
  "name": "support-voice-bot",
  "version": "1.0.0",
  "created_at": "2026-02-16T10:00:00Z"
}

Via YAML File ​

bash
# Upload YAML template
curl -X POST https://api.gateflow.ai/v1/mcp/pipeline-templates/upload \
  -H "Authorization: Bearer gw_prod_admin_key" \
  -F "file=@my-template.yaml"

Template Sections ​

STT Configuration ​

yaml
stt:
  model: whisper-1              # STT model
  language: en                  # ISO language code or "auto"
  streaming: true               # Enable streaming transcription
  word_timestamps: false        # Include word-level timing
  prompt: "Technical discussion" # Hint for better accuracy

Available Models:

  • whisper-1 - OpenAI Whisper (best accuracy)
  • voxtral-mini-latest - Mistral Voxtral (low latency)

LLM Configuration ​

yaml
llm:
  model: gpt-5.2                # LLM model
  system_prompt: |              # System instructions
    You are a helpful assistant.
  max_tokens: 200               # Max response tokens
  temperature: 0.7              # Creativity (0-2)
  top_p: 1.0                    # Nucleus sampling
  presence_penalty: 0           # Discourage repetition
  frequency_penalty: 0          # Discourage common tokens
  tools: []                     # Function calling tools

TTS Configuration ​

yaml
tts:
  model: eleven_multilingual_v2 # TTS model
  voice: professional           # Voice ID or standard name
  streaming: true               # Enable streaming audio
  voice_settings:
    stability: 0.5              # Voice consistency
    similarity_boost: 0.75      # Voice matching
    style: 0.0                  # Style exaggeration
    speed: 1.0                  # Speech rate

Available Models:

  • eleven_multilingual_v2 - Best quality
  • eleven_turbo_v2_5 - Low latency
  • eleven_flash_v2_5 - Cost effective
  • tts-1-hd - OpenAI high quality
  • tts-1 - OpenAI standard

Processing Configuration ​

yaml
processing:
  detect_pii: true              # Scan for PII/PHI
  redact_pii: false             # Redact detected PII
  pii_categories:               # Categories to detect
    - email
    - phone
    - ssn
  format_output: json           # Output format
  extract_entities: true        # Named entity extraction

Compliance Configuration ​

yaml
compliance:
  audit_log: true               # Log all invocations
  data_classification: internal # Classification level
  retention_days: 90            # Log retention
  hipaa: false                  # HIPAA compliance
  matter_id: null               # Legal matter tracking

Using Custom Templates ​

In MCP Tools ​

python
from gateflow_mcp import MCPClient

client = MCPClient(agent_id="agent_abc123", api_key="gf-agent-...")

result = client.call_tool(
    name="voice/pipeline",
    arguments={
        "audio": audio_b64,
        "template": "support-voice-bot"  # Your custom template
    }
)

With Overrides ​

python
result = client.call_tool(
    name="voice/pipeline",
    arguments={
        "audio": audio_b64,
        "template": "support-voice-bot",
        "overrides": {
            "llm": {
                "max_tokens": 300  # Override for this call
            }
        }
    }
)

Template Versioning ​

Create New Version ​

bash
curl -X POST https://api.gateflow.ai/v1/mcp/pipeline-templates/support-voice-bot/versions \
  -H "Authorization: Bearer gw_prod_admin_key" \
  -H "Content-Type: application/json" \
  -d '{
    "version": "1.1.0",
    "changes": "Updated system prompt",
    "llm": {
      "system_prompt": "You are an improved support assistant..."
    }
  }'

List Versions ​

bash
curl https://api.gateflow.ai/v1/mcp/pipeline-templates/support-voice-bot/versions \
  -H "Authorization: Bearer gw_prod_admin_key"

Use Specific Version ​

python
result = client.call_tool(
    name="voice/pipeline",
    arguments={
        "audio": audio_b64,
        "template": "support-voice-bot@1.0.0"  # Pin to version
    }
)

Template Management ​

List Templates ​

bash
curl https://api.gateflow.ai/v1/mcp/pipeline-templates \
  -H "Authorization: Bearer gw_prod_admin_key"

Get Template ​

bash
curl https://api.gateflow.ai/v1/mcp/pipeline-templates/support-voice-bot \
  -H "Authorization: Bearer gw_prod_admin_key"

Update Template ​

bash
curl -X PATCH https://api.gateflow.ai/v1/mcp/pipeline-templates/support-voice-bot \
  -H "Authorization: Bearer gw_prod_admin_key" \
  -H "Content-Type: application/json" \
  -d '{
    "llm": {
      "max_tokens": 250
    }
  }'

Delete Template ​

bash
curl -X DELETE https://api.gateflow.ai/v1/mcp/pipeline-templates/support-voice-bot \
  -H "Authorization: Bearer gw_prod_admin_key"

Assigning to Agents ​

bash
curl -X PATCH https://api.gateflow.ai/v1/mcp/agents/agent_abc123 \
  -H "Authorization: Bearer gw_prod_admin_key" \
  -H "Content-Type: application/json" \
  -d '{
    "permissions": {
      "pipelines": ["support-voice-bot", "voice-agent-fast"]
    }
  }'

Example Templates ​

Meeting Transcription ​

yaml
name: meeting-transcription
stt:
  model: whisper-1
  word_timestamps: true
llm:
  model: gpt-5.2
  system_prompt: |
    Summarize this meeting transcript:
    - Key decisions made
    - Action items with owners
    - Follow-up required
  max_tokens: 500
processing:
  format_output: markdown

Language Learning ​

yaml
name: language-tutor
stt:
  model: whisper-1
  language: auto
llm:
  model: gpt-5.2
  system_prompt: |
    You are a language tutor.
    Correct grammar mistakes gently.
    Respond in the same language as the user.
  temperature: 0.8
tts:
  model: eleven_multilingual_v2
  voice: friendly

Best Practices ​

  1. Version your templates - Track changes over time
  2. Test thoroughly - Validate before production use
  3. Document purpose - Clear descriptions help maintenance
  4. Set appropriate limits - Prevent cost overruns
  5. Use meaningful names - Descriptive template names

Next Steps ​

Built with reliability in mind.

Copyright 2026 GateFlow