v0.1.2

What's Changed

• Updates by @harshalmore31 in #5
• updates by @harshalmore31 in #6

Full Changelog: v0.1.0...v0.1.2

v0.1.0

[0.1.0] — 2026-03-17

🚀 Initial Release

We're excited to launch MemWire, an open-source, enterprise-grade memory infrastructure layer for AI agents and applications.

MemWire gives your AI applications persistent, auditable, and graph-structured memory — without shipping your data to a third-party service.

---

What's included

Core memory engine

• memory.add() — extract and store structured memory from chat messages
• memory.recall() — graph-aware retrieval with supporting and conflicting memory paths
• memory.search() — hybrid semantic + keyword search across stored memories
• memory.feedback() — reinforce memory paths based on response quality; unused edges decay over time

Memory categories

• Automatic classification into fact, preference, instruction, event, and entity

Knowledge base

• memory.add_knowledge() / memory.search_knowledge() — ingest and query document chunks alongside conversation memory

Graph-based recall

• Memory items are connected in a weighted graph; recall traverses paths to surface the most relevant and consistent context

Hybrid search

• Dense vector + BM25 sparse retrieval fused with Reciprocal Rank Fusion (RRF)

Cross-encoder reranker (optional)

• FastEmbed-based cross-encoder for precision reranking of recall results

Storage

• Qdrant — embedded file mode (zero infra) or local/cloud server
• SQLite — metadata, graph edges, and memory records

Embeddings

• FastEmbed (local, no API key required) — default
• OpenAI / Azure OpenAI embeddings — optional

Multi-tenancy

• Isolate memory by user_id, app_id, and workspace_id

Self-hosted REST API

• FastAPI server with endpoints: store, recall, search, knowledge CRUD, health
• OpenAPI spec included

Examples

• examples/openai/ — streaming chat with memory (OpenAI)
• examples/azure-openai/ — streaming chat with memory (Azure OpenAI)

---

Install

pip install memwire

Quickstart

from memwire import MemWire, MemWireConfig

memory = MemWire(config=MemWireConfig(qdrant_path="./memwire_data"))

memory.add(user_id="alice", messages=[
    {"role": "user", "content": "I prefer dark mode and short answers."}
])

result = memory.recall("How should I format my answers?", user_id="alice")
print(result.formatted)
# → "alice prefers dark mode and short answers."

---

📖 Documentation · 💬 Discord · 🐛 Issues