pyproject.toml

[project]
name = "ai-gateway"
version = "0.1.0"
description = "Comprehensive environment for the AI Gateway labs (Jupyter notebooks, Bicep, APIM policies)."
readme = "README.md"
requires-python = ">=3.12"
# Notes:
# - azure-functions-worker is intentionally omitted (managed by the Azure Functions platform).
# - Per-lab pyproject.toml files exist alongside lab notebooks for labs with extra/pinned deps.
dependencies = [
    "requests",
    "jsonref",
    "python-dotenv",
    "pandas",
    "matplotlib",
    "redis",
    "msal",
    "openai[realtime]",
    "azure-cli>=2.66",
    "azure-identity",
    "opentelemetry-api",
    "opentelemetry-sdk",
    "azure-monitor-opentelemetry",
    "azure-monitor-ingestion",
    "azure-ai-projects>=2.2.0",
    "azure-ai-agents",
    "azure-ai-inference[prompts]",
    "azure-search-documents",
    "azure-functions",
    "azure-mgmt-apimanagement",
    # MCP labs (root environment)
    "mcp==1.21.2",
    "autogen-core",
    "autogen-ext[openai,azure,mcp]",
    "autogen-agentchat",
    "semantic-kernel[mcp]==1.33.0",
]

[tool.uv]
package = false
# Azure CLI's transitive dependencies (azure-batch, azure-keyvault-administration)
# are pre-release; allow them globally so the root environment resolves cleanly.
prerelease = "allow"

[tool.uv.workspace]
# Each lab is an independent project with its own pin set, so labs are NOT
# workspace members (their versions conflict). They are installed from their
# own pyproject.toml via `uv sync` inside the lab directory.
members = []
exclude = [
    "labs/_deprecated/**",
    "labs/**",
    "shared/mcp-servers/**",
    "tools/mock-server",
    ".github/skills/**",
]