Roadmap

Forward-looking plan from v0.4.2 (July 2026). Dates are estimates.

---

Shipped in v0.4.2

• Package rename: privysha → asha (PyPI, imports, CLI)
• ANCHOR mission-aware agent governance (anchor(), guards, sandbox, human approval)
• Framework adapters: CrewAI, LangChain, AutoGen, LlamaIndex, MCP, LangGraph, generic
• CI/publish workflows updated for asha package layout
• Integration test job with optional framework deps in CI

Shipped in v0.4.1

• Frozen root API (process, sanitize, optimize, Agent)
• PromptProcessor as sole orchestrator (security → compile → optimize)
• Typed ProcessResult / SanitizeResult / OptimizeResult
• Unified mode= safety semantics
• runtime/resolve.py - no compat on hot path
• Architecture boundary tests in CI
• Dead legacy code removed (Pipeline, ModelRouter, debug/ package, etc.)
• wrap_llm / auto_patch in asha.integrations

---

v0.5.x (target: H2 2026)

• API freeze - deprecation warnings only, no breaking removals
• Container sandbox backends (docker, bwrap)
• Prompt caching for repeated patterns
• Expanded international PII formats
• PyPI classifier → Beta

---

v1.0.0 (target: after community feedback)

• Stable public API guarantee
• Migration guide finalized
• Production readiness checklist (security review, SLA docs)
• Documented support policy

Not committed for 1.0: managed cloud service, multi-tenant SaaS.

---

Preview features (may change anytime)

Feature	Module
ANCHOR framework adapters (non-CrewAI)	runtime/anchor/adapters/
AshaFit local advisor	runtime/local_advisor/
Hybrid ML PII	core/hybrid_pii.py + pii_pipeline/
auto_patch()	integrations/auto_patch.py
Framework middleware	integrations/fastapi, langchain, etc.

---

Long-term ideas

• Cost-aware multi-model routing
• Prometheus metrics export expansion
• Prompt caching engine
• Native function-calling in asha.Agent tool loop

See developer-preview.md for current limitations.