Configuration

Set up openstackai for your environment.

---

Environment Variables

OpenAI

export OPENAI_API_KEY=sk-your-api-key

Azure OpenAI

# With API Key
export AZURE_OPENAI_API_KEY=your-api-key
export AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com/
export AZURE_OPENAI_DEPLOYMENT=gpt-4o-mini

# With Azure AD (recommended - no API key needed!)
export AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com/
export AZURE_OPENAI_DEPLOYMENT=gpt-4o-mini
# Uses your az login credentials automatically

Anthropic

export ANTHROPIC_API_KEY=sk-ant-your-api-key

Ollama (Local)

# No configuration needed for default
# Or specify custom endpoint
export OLLAMA_HOST=http://localhost:11434

---

Programmatic Configuration

import openstackai

openstackai.configure(
    api_key="sk-...",
    model="gpt-4o",
    temperature=0.7
)

---

Provider-Specific Configuration

OpenAI

from openstackai.core import OpenAIProvider, LLMConfig

provider = OpenAIProvider(LLMConfig(
    api_key="sk-...",
    model="gpt-4o-mini",
    temperature=0.7,
    max_tokens=4096,
))

Azure OpenAI

from openstackai.core import AzureOpenAIProvider, LLMConfig

# API Key auth
provider = AzureOpenAIProvider(LLMConfig(
    api_key="your-key",
    api_base="https://your-resource.openai.azure.com/",
    model="gpt-4o-mini",
    api_version="2024-02-15-preview",
))

# Azure AD auth (recommended for enterprise)
provider = AzureOpenAIProvider(LLMConfig(
    api_base="https://your-resource.openai.azure.com/",
    model="gpt-4o-mini",
    # No api_key = uses DefaultAzureCredential
))

Anthropic

from openstackai.core import AnthropicProvider, LLMConfig

provider = AnthropicProvider(LLMConfig(
    api_key="sk-ant-...",
    model="claude-3-sonnet-20240229",
    max_tokens=4096,
))

---

Model Selection

openstackai automatically selects the appropriate model based on environment:

1. If AZURE_OPENAI_ENDPOINT is set → Azure OpenAI
2. If OPENAI_API_KEY is set → OpenAI
3. If ANTHROPIC_API_KEY is set → Anthropic
4. If Ollama is running → Ollama

Override with explicit provider:

from openstackai import Agent
from openstackai.core import AzureOpenAIProvider, LLMConfig

agent = Agent(
    name="Bot",
    instructions="...",
    llm=AzureOpenAIProvider(LLMConfig(...))
)

---

YAML Configuration

Define agents in YAML files:

# agents/research_assistant.yaml
name: ResearchAssistant
instructions: |
  You are a research assistant that helps users find information.
  Be thorough and cite your sources.
model: gpt-4o-mini
temperature: 0.7
tools:
  - web_search
  - summarize
memory:
  type: conversation
  max_messages: 50

Load and use:

from openstackai.config import load_agent, AgentBuilder

config = load_agent("agents/research_assistant.yaml")
agent = AgentBuilder.from_config(config).build()

---

Default Settings

Setting	Default	Description
model	gpt-4o-mini	Default model
temperature	0.7	Creativity level
max_tokens	4096	Max output tokens
timeout	60	Request timeout (seconds)
retry_count	3	Retry attempts
retry_delay	1.0	Delay between retries

---

Next Steps

• [[Quick Start]] - Run your first program
• [[Azure AD Auth]] - Enterprise authentication
• [[Agent]] - Create agents