Skip to main content

Changelog

All notable changes to openstackai will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[0.4.0] - 2026-02-16​

Release Summary​

Major release achieving feature parity with industry-leading AI agent SDKs including OpenAI Agents SDK, Microsoft Semantic Kernel, Google ADK, Strands Agents, and Anthropic SDK.

Added​

Phase 1 Features (Core)​

  • Runner Pattern - Structured agent execution inspired by OpenAI Agents SDK (29 tests)
  • Agent Config YAML - No-code agent configuration from YAML/JSON files (24 tests)
  • Agents as Plugins - Plugin architecture inspired by MS Semantic Kernel (22 tests)
  • OpenAPI Tools - Auto-generate tools from OpenAPI specs (40 tests)
  • Token Counting - Token counting utilities inspired by Anthropic SDK (40 tests)

Phase 2 Features (Advanced)​

  • Tool Auto-Discovery - Automatic tool loading from directories (12 tests)
  • Context Caching - Cache context for performance optimization (7 tests)
  • Session Rewind - Checkpoint and rollback session states (6 tests)
  • Multimodal Vision - Image/vision input support (12 tests)
  • Vector DB Connectors - ChromaDB, Pinecone, Weaviate, Qdrant, Azure AI Search (7 tests)

Phase 3 Features (Enterprise)​

  • A2A Protocol - Agent-to-Agent communication protocol (12 tests)
  • Development UI - Built-in debugging and testing UI (9 tests)
  • Voice Streaming - Real-time voice input/output (10 tests)

Phase 4 Features (Architecture)​

  • Kernel Registry Pattern - Service registry inspired by MS Semantic Kernel (35 tests)

Changed​

  • Azure AD Authentication - AzureOpenAIProvider now auto-detects DefaultAzureCredential
  • Professional Repository Structure - Improved documentation, GitHub templates, workflows

Test Coverage​

  • 671 tests passing (636 core + 35 kernel)

[0.1.0] - 2026-02-12​

🎉 Initial Release - "The Pandas of AI"​

This is the first public release of openstackai, the revolutionary AI library that makes complex AI tasks as simple as pandas operations.

Added​

One-Liner Functions​

  • ask() - Ask any question, get intelligent answers
  • research() - Deep research on any topic
  • summarize() - Summarize text, files, or URLs
  • extract() - Extract structured data from text
  • generate() - Generate content of various types
  • translate() - Translate between languages
  • chat() - Interactive chat sessions with memory
  • agent() - Create custom AI agents

Modules​

  • rag - RAG operations in 2 lines

    • rag.index() - Index documents
    • rag.ask() - Query documents
    • rag.from_url() - RAG from URL
    • rag.from_text() - RAG from text

  • fetch - Real-time data fetching

    • fetch.weather() - Get weather data
    • fetch.news() - Get news articles
    • fetch.stock() - Get stock data
    • fetch.crypto() - Get cryptocurrency data
    • fetch.facts() - Get facts on topics

  • analyze - Data and text analysis

    • analyze.data() - Analyze DataFrames/dicts
    • analyze.text() - Analyze text content
    • analyze.sentiment() - Sentiment analysis
    • analyze.compare() - Compare items

  • code - Code operations

    • code.write() - Generate code
    • code.review() - Review code quality
    • code.debug() - Debug errors
    • code.explain() - Explain code
    • code.refactor() - Refactor code
    • code.convert() - Convert between languages

Core Features​

  • Zero-configuration setup (auto-detects API keys)
  • Multi-provider support (OpenAI, Anthropic, Azure)
  • Full type hints with .pyi stub files
  • Lazy imports for fast startup
  • Sensible defaults for all operations

Prebuilt Agent Personas​

  • coder - Expert programmer
  • researcher - Academic researcher
  • writer - Content writer
  • analyst - Data analyst
  • teacher - Patient educator
  • advisor - Strategic advisor
  • critic - Constructive critic
  • creative - Creative thinker
  • editor - Content editor
  • python_expert - Python specialist

Infrastructure​

  • Configuration management with configure()
  • Environment variable support
  • Azure OpenAI integration
  • Comprehensive error handling

Technical Details​

  • Python Version: 3.9+
  • Dependencies: openai, anthropic (optional), pydantic
  • License: MIT

[Unreleased]​

Planned Features​

  • Async/await support for all functions
  • Streaming responses
  • Custom skill plugins
  • Agent memory persistence
  • Multi-agent orchestration
  • Local LLM support (Ollama, llama.cpp)
  • Web UI for agent management
  • MCP (Model Context Protocol) integration

Migration Guide​

From LangChain​

# Before (LangChain)
from langchain.llms import OpenAI
from langchain.chains import LLMChain
from langchain.prompts import PromptTemplate

llm = OpenAI(api_key=os.environ.get("OPENAI_API_KEY"))
template = PromptTemplate(input_variables=["q"], template="{q}")
chain = LLMChain(llm=llm, prompt=template)
result = chain.run("What is AI?")

# After (openstackai)
from openstackai import ask
result = ask("What is AI?")

From LlamaIndex​

# Before (LlamaIndex)
from llama_index import VectorStoreIndex, SimpleDirectoryReader

documents = SimpleDirectoryReader("./docs").load_data()
index = VectorStoreIndex.from_documents(documents)
query_engine = index.as_query_engine()
response = query_engine.query("What is the summary?")

# After (openstackai)
from openstackai import rag
response = rag.ask("./docs", "What is the summary?")

Version History​

VersionDateHighlights
0.1.02026-02-12Initial release with full easy API

For feature requests or bug reports, please visit our GitHub Issues.