Config Schema Reference

Complete schema for sekuire.yml configuration.

---

Full Schema

sekuire.yml

# Project metadata
project:
  name: string          # Required: Project name
  version: string       # Required: Semantic version (e.g., "1.0.0")
  description: string   # Optional: Project description

# Agent definitions
agents:
  <agent_key>:          # Agent identifier (used with getAgent())
    name: string        # Required: Display name
    system_prompt: string  # Required: Path to system prompt file
    tools: string | array  # Optional: Path to tools.json or inline list
    llm:
      provider: string     # Required: openai, anthropic, google, ollama
      model: string        # Required: Model name (e.g., gpt-4-turbo)
      api_key_env: string  # Required: Env var for API key
      temperature: number  # Optional: 0-2 (default: 0.7)
      max_tokens: number   # Optional: Max response tokens
      base_url: string     # Optional: Custom API base URL
    memory:
      type: string         # Optional: buffer, window (default: buffer)
      max_messages: number # Optional: Max messages to retain

# Deployment config
deployment:
  docker_image: string  # Optional: Container image reference
  runtime: string       # Optional: node, python, rust
  port: number          # Optional: Container port (default: 8000)

# Marketplace metadata (for agent vendors)
marketplace:
  category: string      # Optional: analytics, productivity, etc.
  tags: array           # Optional: Search tags
  pricing:
    model: string       # Optional: subscription, usage, freemium
    price_usd: number   # Optional: Price in USD
    billing: string     # Optional: monthly, yearly, per-request
  compliance: array     # Optional: soc2, hipaa, gdpr

---

project

Field	Type	Required	Description
name	string	Yes	Project name (used in registry)
version	string	Yes	Semantic version
description	string	No	Project description

---

agents.<key>

Each agent is defined under a unique key (e.g., assistant, researcher).

Field	Type	Required	Description
name	string	Yes	Display name
system_prompt	string	Yes	Path to .md file
tools	string | array	No	Tools config

llm

Field	Type	Required	Description
provider	string	Yes	openai, anthropic, google, ollama
model	string	Yes	Model name
api_key_env	string	Yes	Env var name for API key
temperature	number	No	Sampling temperature (0-2)
max_tokens	number	No	Max response tokens
base_url	string	No	Custom API base URL

memory

Field	Type	Required	Description
type	string	No	buffer, summary, window, redis, postgres, file
max_messages	number	No	Max messages to retain (for buffer/summary/window)
config	object	No	Backend-specific config (for redis/postgres/file)

Memory Types

Type	Description
buffer	FIFO queue - oldest messages dropped first (default)
summary	Summarize older messages to save context
window	Sliding window of recent messages
redis	Redis-backed persistent storage
postgres	PostgreSQL-backed persistent storage
file	File-based local storage

Memory Examples

In-Memory (default)

memory:
  type: "buffer"
  max_messages: 10

Summary-based

memory:
  type: "summary"
  max_messages: 20

Redis-backed

memory:
  type: "redis"
  config:
    url: "redis://localhost:6379"  # Override via REDIS_URL

PostgreSQL-backed

memory:
  type: "postgres"
  config:
    url: "postgres://user:pass@localhost/dbname"  # Override via DATABASE_URL

File-based

memory:
  type: "file"
  config:
    path: "./agent_memory.json"

---

deployment

Field	Type	Required	Description
docker_image	string	No	Container image reference
runtime	string	No	node, python, rust
port	number	No	Container port (default: 8000)

---

Supported LLM Providers

Provider	Models	API Key Env
openai	gpt-4, gpt-4-turbo, gpt-4o, gpt-4o-mini	OPENAI_API_KEY
anthropic	claude-3-opus, claude-3-sonnet, claude-3-haiku	ANTHROPIC_API_KEY
google	gemini-pro, gemini-1.5-pro, gemini-2.0-flash	GOOGLE_API_KEY
ollama	llama3, mistral, mixtral, deepseek (local)	N/A
custom	Any OpenAI-compatible API	Custom env var

Custom LLM Providers

You can use any OpenAI-compatible API by setting provider: "custom" and specifying a base_url:

OpenRouter

llm:
  provider: "custom"
  model: "anthropic/claude-3.5-sonnet"
  api_key_env: "OPENROUTER_API_KEY"
  base_url: "https://openrouter.ai/api/v1"

Azure OpenAI

llm:
  provider: "custom"
  model: "gpt-4"
  api_key_env: "AZURE_OPENAI_API_KEY"
  base_url: "https://your-resource.openai.azure.com/openai/deployments/gpt-4"

Together AI

llm:
  provider: "custom"
  model: "meta-llama/Llama-3-70b-chat-hf"
  api_key_env: "TOGETHER_API_KEY"
  base_url: "https://api.together.xyz/v1"

Local vLLM / Text Generation Inference

llm:
  provider: "custom"
  model: "mistral-7b"
  api_key_env: "LOCAL_API_KEY"  # or leave empty
  base_url: "http://localhost:8080/v1"

---

Complete Example

sekuire.yml

project:
  name: "research-assistant"
  version: "1.0.0"
  description: "AI research assistant with web search"

agents:
  researcher:
    name: "Research Agent"
    system_prompt: "./prompts/researcher.md"
    tools:
      - web_search
      - calculator
    llm:
      provider: "anthropic"
      model: "claude-3-opus-20240229"
      api_key_env: "ANTHROPIC_API_KEY"
      temperature: 0.3
      max_tokens: 4000
    memory:
      type: "buffer"
      max_messages: 20

  coder:
    name: "Coding Agent"
    system_prompt: "./prompts/coder.md"
    tools: "./coder-tools.json"
    llm:
      provider: "openai"
      model: "gpt-4-turbo"
      api_key_env: "OPENAI_API_KEY"
      temperature: 0.1

deployment:
  docker_image: "myrepo/research-assistant:v1.0.0"
  runtime: "node"
  port: 8000

tip

Use sekuire init to generate a starter config with sensible defaults.