Skip to content

ROADMAP.md

Latest commit

 

History

History
157 lines (107 loc) · 8.69 KB

ROADMAP.md

File metadata and controls

157 lines (107 loc) · 8.69 KB
Raw

Moss Roadmap

This is a living document. We update it as priorities shift based on community feedback and production learnings. If something here excites you, open an issue or PR — we move fast on contributions.

Shipped

SDKs & Runtimes

  • Python SDK (moss) — async-first, type-safe
  • TypeScript SDK (@moss-dev/moss) — full feature parity with Python
  • Elixir SDK (moss) — Hex package for Phoenix / LiveView apps
  • Go SDK — bindings-backed manage + local query via libmoss (sdks/go/)
  • C bindings — example usage, metadata filtering, session management
  • Bun runtime support — native Bun example application
  • WebAssembly runtime — client-side semantic search in the browser, no server required

Search & Indexing

  • Built-in embedding models (moss-minilm)
  • Custom embedding support (bring your own OpenAI, Cohere, etc.)
  • Hybrid search — combine semantic search with BM25 keyword matching
  • Metadata filtering ($eq, $and, $in, $near)
  • Document management (add, upsert, get, delete)
  • Text classification — zero-shot classification via Moss similarity (python-classification)
  • Benchmarks directory — reproducible latency/throughput scripts comparing Moss vs Pinecone, Qdrant, and Chroma on standardized datasets

Framework Integrations

  • LangChain integration — MossRetriever (BaseRetriever) + get_moss_tool() factory; canonical Python integration pattern
  • DSPy integration — notebook-based DSPy + Moss retrieval
  • Haystack integration — MossDocumentStore and MossRetriever drop-in components for RAG pipelines
  • CrewAI integration — Moss as a retrieval tool for CrewAI agents; travel-planning demo
  • AutoGen integration — multi-agent e-commerce support with Moss sub-10ms context retrieval
  • LlamaIndex integration — full-stack PDF processing with Liteparse + Moss vector search
  • Mastra integration — Moss wrapped as a createTool() primitive for Mastra agents (TypeScript)
  • AWS Strands Agents integration — strands-agents-moss package for Strands Agents workflows
  • Daytona sandbox integration — log ingestion Q&A agent; code execution in isolated Daytona sandboxes
  • Cognee + Daytona integration — multi-agent shared memory graph (Cognee + Moss) across isolated Daytona sandboxes
  • LangGraph integration — retrieval node for stateful multi-agent workflows
  • Langflow integration — drag-and-drop MossRetrieverComponent and MossSearchComponent
  • n8n community node — create index, manage documents, and query from n8n workflows

Voice AI

  • Pipecat voice agent integration — pipecat-moss package + quickstart bot
  • Ollama + Moss + Pipecat reference architecture — fully local LLM voice agent; single docker compose up
  • Hume AI + Ollama + Pipecat — local voice agent with Hume AI (Octave) expressive TTS
  • LiveKit voice agent integration — LiveKit agent + React frontend deployed to Vercel
  • ElevenLabs integration — elevenlabs-moss package; knowledge-base-backed Conversational AI bot
  • Agora Conversational AI integration — agora-moss MCP server; search_knowledge_base tool for Agora voice agents
  • VAPI integration — vapi-moss webhook adapter with HMAC verification; LLM-directed retrieval via Custom Tool
  • TEN Framework integration — MossSessionManager + voice-assistant app with session-scoped grounding

Developer Tools & Packages

  • Moss CLI — moss index, moss query, moss documents — manage indexes and run queries without writing code
  • MCP server — expose Moss as a Model Context Protocol server so any MCP-compatible AI tool (Claude, Cursor, Windsurf) can do semantic search
  • Vercel AI SDK integration — @moss-tools/vercel-sdk tool wrappers: search, create index, manage documents
  • VitePress search plugin — cloud search on first keystroke, on-device after index download; live demo on Vercel
  • Markdown documentation indexer — moss-md-indexer parses and chunks Markdown docs for upload to Moss
  • Zo computer skill — Moss semantic search skill for the Zo computer platform

Data Connectors

  • MongoDB connector — sync MongoDB collections → Moss index
  • MySQL / MariaDB connector — sync tables → Moss index via PyMySQL
  • SQLite connector — sync SQLite tables → Moss index
  • Supabase connector — sync Supabase tables → Moss index via PostgREST

Apps & Deployment

  • Next.js example app — Next.js 16 browser-based semantic search UI (@moss-dev/moss-web)
  • Docker deployment examples (ECS / Kubernetes patterns) — Python + JS SDK in containers

In Progress

  • Firecrawl cookbook — crawl a URL with Firecrawl and index the content directly into Moss; turnkey web knowledge base for agents
  • Unstructured cookbook — ingest PDF, DOCX, and HTML files via Unstructured and load into Moss; doc-parsing connector example
  • Google ADK integration — Moss as a retrieval tool for Google's Agent Development Kit
  • Smolagents integration — lightweight retrieval tool for Hugging Face's agent framework

Next Up — Community Contributions Welcome

These are well-scoped and ready for contributors. Each one has (or will have) a corresponding GitHub issue with detailed instructions.

New SDK Bindings

  • Swift bindings — for iOS/macOS apps with on-device retrieval (good first issue)
  • Rust bindings — for performance-critical pipelines (good first issue)
  • Kotlin bindings — for Android apps and Spring Boot backend services (good first issue)

Voice AI Ecosystem

  • Daily.co integration — real-time audio pipeline with semantic context injection
  • Twilio integration — retrieval for phone-based AI agents (IVR, call center bots)

Developer Tools

  • VS Code extension — semantic search over your codebase directly from the editor sidebar

Search Quality

  • Reranking support — plug in cross-encoder rerankers as a post-retrieval step
  • Multi-vector retrieval — support ColBERT-style late interaction models

Data Ingestion

  • Chunking strategies — built-in text splitters (sentence, paragraph, recursive, semantic)

Future

These are bigger bets we're exploring. They're directional, not committed — community input will shape what gets built.

Local-First AI Stack

  • vLLM-based local inference + local search — a fully local pipeline: your model, your embeddings, your search, your hardware. No API calls. This is a natural fit for the privacy-first voice AI use case and can meaningfully cut latency for on-premise deployments.

Evaluation & Quality

  • LLM-as-a-judge evaluation framework — automated retrieval quality scoring using LLM judges. We want to lay the foundation and let the community decide the direction — what metrics matter, which judges to support, how to benchmark fairly.
  • Retrieval quality dashboard — visualize query performance, relevance scores, and failure modes over time

Browser & Edge

  • Edge runtime support — run Moss in Cloudflare Workers, Deno Deploy, and Vercel Edge Functions

Advanced Retrieval

  • Query expansion — LLM-powered query rewriting to improve recall on short or ambiguous queries
  • Sparse-dense fusion (SPLADE) — learned sparse retrieval to complement BM25 hybrid, improving precision on rare terms
  • Contextual retrieval — pre-chunking contextualization to make every chunk self-contained and more retrievable

More Data Connectors

Connect knowledge sources to Moss without writing custom ETL.

  • GitHub connector — index code, issues, PRs, and docs from repositories
  • Notion connector — sync and index Notion workspace pages
  • Confluence connector — enterprise knowledge base indexing
  • S3/GCS sync — auto-index documents from cloud storage buckets on upload

How to Contribute

  1. Pick something from "Next Up" — these are ready for PRs
  2. Check the issues — look for good first issue and help wanted labels
  3. Propose something new — open an issue describing what you want to build. We're open to ideas that aren't on this list.
  4. Read the Contributing Guide — fork, branch from main, PR

If you're unsure where to start, drop a message in Discord and we'll point you in the right direction.