This directory contains example workflows demonstrating KDeps v2 features.

Basic LLM chatbot with validation and error handling.

• YAML configuration
• HTTP API with CORS
• Preflight validation
• LLM integration with Ollama

Run:

cd chatbot && kdeps run workflow.yaml --dev

Advanced HTTP client features with authentication, caching, and retries.

• Multiple authentication methods (Bearer, API Key, Basic, OAuth2)
• Intelligent retry logic with exponential backoff
• Response caching with TTL
• Proxy and TLS configuration
• Custom headers and timeout management

Run:

cd http-advanced && kdeps run workflow.yaml --dev

Shell command execution with timeout and output capture.

• System command execution
• Output parsing and formatting
• Timeout handling
• Cross-platform shell support

Run:

cd shell-exec && kdeps run workflow.yaml --dev

Advanced SQL operations with batch processing and multiple result formats.

• Named database connections
• Batch operations (paramsBatch)
• CSV and JSON result formatting
• Transaction support
• Connection pooling

Features:

• PostgreSQL analytics queries (CSV output)
• MySQL inventory management
• Batch user status updates
• Multi-database transactions

Run:

cd sql-advanced && kdeps run workflow.yaml --dev

File upload handling with multipart form-data support.

• Single and multiple file uploads
• File metadata access (count, names, types)
• File content, path, and MIME type access
• Integration with get() and info() functions

Run:

cd file-upload && kdeps run workflow.yaml --dev

Vision model integration with image analysis.

• Image file upload via multipart form-data
• Vision model integration (moondream, llava)
• Image analysis with natural language queries
• JSON structured responses

Prerequisites: Install vision model: ollama pull moondream:1.8b

Run:

cd vision && kdeps run workflow.yaml --dev

AI phone IVR using the native telephony: resource (TwiML for Twilio-compatible providers).

• Keypad menu with match results (result.status, result.interpretation)
• Static opening-hours line, speech-question gather
• LLM answers the spoken question as text for provider-side TTS

Prerequisites: ollama pull llama3.2:1b (for the answer route)

Run:

cd telephony-bot && kdeps run workflow.yaml --dev

LLM tool calling / function calling with automatic tool execution.

• Tool/function definitions in ChatConfig
• Tool parameter schemas (OpenAI Functions format)
• Automatic tool execution when LLM calls tools
• Tool results fed back to LLM for final response
• Multi-turn conversations with tools
• Multiple tools per request

Features:

• ✅ Tool definitions work
• ✅ Automatic tool execution
• ✅ Tool resources (Python, SQL, HTTP, Exec, APIResponse)
• ✅ Tool arguments automatically stored in memory

Run:

cd tools && kdeps run workflow.yaml --dev

Session management with persistent storage.

• SQLite-backed persistent session storage
• Session TTL configuration
• Session data storage and retrieval
• Preflight checks using session data

Run:

cd session-auth && kdeps run workflow.yaml --dev

Demonstrates automatic environment variable scoping per component, auto-scaffolded .env templates, and kdeps component update.

• TRANSLATOR_OPENAI_API_KEY overrides OPENAI_API_KEY inside translator only
• On first run, kdeps auto-creates .env template with all env() vars listed blank
• On first run, kdeps auto-creates README.md from component metadata
• kdeps component update merges new vars into an existing .env
• Summarizer falls back to extractive summary when no API key is set

Run:

export OPENAI_API_KEY=sk-...          # global fallback
export TRANSLATOR_OPENAI_API_KEY=sk-... # translator only (optional)
cd auto-env && kdeps run workflow.yaml

A sub-workflow designed to be called exclusively via run.component. Declares sources: [component] so no HTTP server or listener is started.

• sources: [component] - no listener started, driven by parent
• Reusable text-transformation sub-module
• Access via run.component from any parent workflow
• Demonstrates component.description field

Invoke from a parent:

# This workflow has no standalone entry-point.
# Call it from a parent workflow resource:
#   run:
#     component:
#       name: component-input-source
#       with:
#         text: "hello world"
#         style: title
kdeps validate examples/component-input-source

Demonstrates automatic dependency installation and per-run cleanup via setup and teardown lifecycle hooks in a custom local component.

• setup.pythonPackages — installed once via uv pip
• setup.osPackages — installed once via apt-get / apk / brew
• setup.commands — run once after packages are ready
• teardown.commands — run after every invocation

Run:

cd component-setup-teardown && kdeps run workflow.yaml

Single-shot document summarizer using the file input source. Reads content from --file, stdin, or KDEPS_FILE_PATH and returns a structured AI analysis.

• sources: [file] — runs once and exits
• Input via --file, stdin pipe, or KDEPS_FILE_PATH
• input('fileContent') / input('filePath') in resources
• Structured JSON output with title, summary, key points

Run:

cd file-processor
echo "The Eiffel Tower was built in 1889 for the World's Fair." | kdeps run workflow.yaml
# or
kdeps run workflow.yaml --file /path/to/document.txt

Shows the built-in input component — pre-installed, no kdeps component install needed. Collects named input slots and returns structured JSON.

• Uses the built-in input component with query, text, and key slots
• 14 named slots: query, prompt, text, data, key, value, a–h
• Passes collected inputs to an LLM for Q&A
• Access component output with output('actionId')

Run:

cd input-component && kdeps run workflow.yaml

1. Choose an example directory
2. Update database connections if needed
3. Run: kdeps run workflow.yaml --dev
4. Test endpoints with curl or your browser

1. Start with Chatbot - Learn basic YAML, HTTP API, and LLM integration
2. Advanced SQL - Master database operations, batch processing, and result formatting
3. Build Your Own - Combine features from examples

Most examples require databases. See each example's README for setup instructions.

For quick testing, you can use SQLite (no setup required) by changing connection strings to:

connection: "sqlite:///data.db"

Use --dev flag for hot reload during development:

kdeps run workflow.yaml --dev

This automatically restarts the server when you modify workflow files or resources.