Vask

Voice-to-Action AI Framework — enterprise-ready automation that connects voice input, LLM reasoning, and tool execution into streamlined business workflows.

Built by Techforcement to help growing businesses automate operations and scale efficiently.

What Vask Does

Vask turns natural language — spoken or typed — into automated business actions. Instead of switching between tools, writing scripts, or waiting on manual processes, teams describe what they need and Vask handles the rest.

For business operations:

• Automate CRM updates, customer follow-ups, and sales logging via voice
• Trigger multi-step workflows across Slack, Google Workspace, and databases
• Monitor services and respond to incidents with AI-guided remediation
• Process inbound webhooks from Stripe, GitHub, Twilio into automated actions

For technical teams:

• Multi-provider LLM support (Gemini, Claude, OpenAI, OpenRouter)
• Plugin system for Slack, Google Workspace, SQL databases
• YAML-driven workflow engine with conditional branching and human approval gates
• REST API, CLI, and MCP server deployment options
• RBAC authentication, rate limiting, and audit logging

Quick Start

pip install -e .
vask --help

Core Commands

vask ask "Summarize today's sales pipeline"       # Text query to LLM
vask record                                        # Voice → transcribe → LLM
vask transcribe                                    # Voice → text only
vask serve                                         # Start REST API server
vask mcp                                           # Start MCP server
vask plugins                                       # List available plugins

Configuration

Config lives at ~/.config/vask/config.toml:

[defaults]
input = "mic"
transcription = "openai-whisper"
llm = "gemini"
output = "terminal"

[providers.gemini]
type = "llm"
api_key_env = "GEMINI_API_KEY"
model = "gemini-2.0-flash"

[providers.claude]
type = "llm"
api_key_env = "ANTHROPIC_API_KEY"
model = "claude-sonnet-4-20250514"

[server]
host = "0.0.0.0"
port = 8420
cors_origins = ["*"]

Set API keys as environment variables:

export GEMINI_API_KEY="..."
export OPENAI_API_KEY="..."
export ANTHROPIC_API_KEY="..."

Workflow Examples

Vask workflows are YAML-defined automation pipelines. See the examples/ directory:

Workflow	Use Case
crm_call_log.yaml	Log sales calls → extract CRM data → notify Slack
incident_response.yaml	Check service health → AI analysis → approve → notify
client_onboarding.yaml	Welcome email → provision workspace → schedule kickoff
lead_qualification.yaml	Score inbound leads → route to sales → update CRM
weekly_report.yaml	Pull metrics → generate summary → distribute to stakeholders

API Server

vask serve --port 8420

Key endpoints:

• POST /v1/ask — Query LLM with text
• POST /v1/transcribe — Transcribe audio (base64)
• POST /v1/sessions — Manage conversation sessions
• POST /v1/workflows/{id}/run — Execute automation workflows
• POST /v1/tools/execute — Run tools directly
• POST /v1/webhooks/{id}/receive — Inbound webhook receiver
• GET /health — Health check with provider status

MCP Server

Integrate Vask as tools in Claude or other MCP-compatible clients:

{
  "mcpServers": {
    "vask": {
      "command": "vask",
      "args": ["mcp"]
    }
  }
}

Plugins

Plugin	Tools
Slack	Send messages, search, react, set status, list channels
Google Workspace	Gmail, Calendar, Sheets operations
SQL	Execute queries against SQLite, PostgreSQL, MySQL

Enable plugins via config or at runtime through the API.

Deployment

# Docker
docker compose up -d

# Direct
vask serve --host 0.0.0.0 --port 8420

Architecture

Voice/Text Input → Transcription (Whisper) → LLM Reasoning → Tool Execution → Output
                                                    ↕
                                          Workflow Engine
                                          Plugin System
                                          Webhook Handler

Modular provider pattern — swap any component (input, transcription, LLM, output, tools) without changing the pipeline.

License

MIT