# Experimental Features and Product Boundary
This document classifies the current v0.1.x repository surface. It is the operational reference for
what is production-facing, optional, experimental, or deferred.
## Production-Facing MVP
The v0.1.x product boundary is an embedded local-first persistent memory engine:
| Embedded Rust SDK and CLI | Production-facing |
| Memory `put/get/delete/list/search` | Production-facing |
| WAL-backed recovery | Production-facing |
| Namespaces and scalar metadata filters | Production-facing |
| Derived namespace and metadata indexes | Production-facing |
| HNSW vector retrieval | Production-facing |
| BM25 lexical retrieval | Production-facing |
| Hybrid Retrieval v1 with deterministic RRF | Production-facing |
| Basic phrase filtering | Production-facing |
| Manual rebuild and structural audit flows | Production-facing |
| JSONL export/import | Production-facing |
| Source-installed Python SDK | Production-facing |
## Optional Wrapper
| Local `vanta-server` binary | Optional wrapper around the embedded core |
The server exists for local development and network exposure around the same embedded engine. It is
not the primary product identity for this release.
## Experimental or Not MVP
These surfaces may exist in the repository, but they are not stable product claims for v0.1.x:
| IQL/LISP/DQL parser, evaluator, and executor paths | **Archived (2024-06-10)** - Runtime LISP evaluation abandoned due to borrow checker issues and GIL blocking. See `archive/experimental-quarantine-2024-06/` |
| MCP API | Experimental integration surface |
| LLM/Ollama integration | External optional integration, not core dependency |
| Governance and maintenance semantics | **Archived (2024-06-10)** - Runtime governance framework incompatible with compile-time governance via IQL AST Pass. Useful utilities extracted to `src/utils/` |
| Graph traversal beyond stored local edges | Experimental, not a graph database claim |
| Docker/Ollama examples | Experimental development examples |
## Extracted Utilities (Production-Ready)
The following utilities were extracted from experimental governance and are now part of the core API:
| DuplicatePreventionFilter | Bloom filter for duplicate prevention in multi-writer scenarios | `vantadb::utils::DuplicatePreventionFilter` |
| OriginCollisionTracker | Collision tracking and friction metrics for multi-agent coordination | `vantadb::utils::OriginCollisionTracker` |
| compute_confidence_friction | Functional friction metric computation for conflict analysis | `vantadb::utils::compute_confidence_friction` |
These utilities are stateless, well-tested, and suitable for production use in multi-writer and multi-agent scenarios.
Visible in-tree examples and docs must carry the same boundary:
- `docs/experimental/IQL.md`
- `examples/docker/docker-compose.ollama.yml`
- `examples/docker/Dockerfile`
- `examples/docker/start.sh`
- `examples/python/langchain_rag.py`
## Deferred
The following are explicitly outside the v0.1.x MVP:
| Agent metacognition and automatic memory consolidation | Deferred |
| Plugins and marketplace | Deferred |
| RBAC, true multi-tenancy, quotas, and enterprise audit | Deferred |
| HA, replication, clustering, and cloud managed service | Deferred |
| Production PyPI publication and signed installers | Deferred |
| Advanced ranking, snippets, highlighting, Unicode folding, stopwords, stemming | Deferred |
| SQL, general OLTP, warehouse, and time-series workloads | Deferred |