Roadmap

Maticlib is actively developed. Below is the current progress and planned features for the library.

Core Library

[x] Google Gemini integration
[x] Mistral AI integration
[x] OpenAI Responses API integration
[x] Graph-based workflow engine
[x] Parallel node execution
[x] Automated documentation system
[x] Native Tool/Function calling system
[x] API Reference Architecture: Standardized Google-style docstrings with table rendering
[x] Documentation Modularization: Decoupled, component-based Modules & API guide
[ ] Anthropic Claude integration
[ ] Ollama (Local LLM) integration
[ ] Streaming support for all providers

[x] Embeddings support (OpenAI, Google, Mistral)
[x] Context Engineering: Advanced Chunkers & Document Loaders
[x] Text2SQL Engine: Schema Reflection, Transpilation, and Validation Guards
[x] Vector Stores: Native integrations (Chroma, Milvus, Qdrant, Pinecone)
[x] Full Telemetry & Observability: Pipeline Traces and Logging Callbacks
[x] Resilience: Exponential Backoff and Retry Policies
[x] Memory Management: Rolling Window and Conversation Buffers
[x] Formatting Utilities: Converting execution outputs to aligned Markdown tables
[x] RAG & Text2SQL Evaluation: Fuzzy heuristics & pluggable LLM metrics for relevance and accuracy
[ ] Agent Tooling: Built-in tools for Playwright, Selenium, Code Execution, and Bash/CLI scripting.
[ ] MCP Support: Tools to both host MCP servers and consume MCP clients.
[ ] Agent Integration: Higher-level Agent abstractions for complex task decomposition.
[ ] Multi-agent collaboration protocols
[ ] Workflow persistence & checkpointing
[x] Detailed Token Usage Tracking

[ ] FastAPI Integration: One-line utility to expose any MaticGraph or Agent as a streaming REST API.
[ ] MaticUI: A professional-grade interface for testing and production-ready agent interactions (with future support for authentication).