Memind 0.2.0 Release Notes

[me0106]

Highlights

Memind 0.2.0 expands the project from a core memory engine into a broader memory platform for applications and coding agents.

• Added Memory Graph relationships and Memory Thread replay (feat: add graph relationships and Memory Thread capabilities #28, Optimize memory thread replay reads #50).
• Added multimodal raw data ingestion and improved insight/temporal extraction (feat: add multimodal rawdata support #23, feat: refactor and optimize insight tree #24, feat: redesign temporal memory item extraction #26, Improve memory item extraction prompt #31).
• Added admin APIs and UI for buffers, dashboards, items, graphs, and memory management (Add admin buffer, dashboard, and item graph APIs #39, Add Memind UI admin console #43, WIP: add admin memory APIs and UI data layer #87).
• Added Java, Python, TypeScript, Go, and Rust SDKs (feat: add Java client release workflow #55, feat: add official Python client #58, Add official TypeScript client #69, Add official Go API client #71, feat(rust-client): add official Rust SDK #72).
• Added integrations for Claude Code, Codex, OpenClaw, and Hermes (feat: add Claude Code memory integration #32, Add Codex integration plugin #35, Add Memind OpenClaw integration #79, Add Hermes Agent memory provider integration #82, Add coding agent timeline memory support #102, Add agent timeline ingestion for OpenClaw and Hermes #103).
• Added an HTTP MCP server and expanded MCP tools (Add HTTP MCP server support #76, Expand Memind MCP memory tools #111).
• Added durable OpenAPI ingestion, generated server OpenAPI, and a standardized API response contract (fix: add durable Open API ingestion #61, standardize server API response contract #65, feat: add server OpenAPI generation #80).
• Migrated observability to Micrometer with memory-operation metrics and retrieval traces (Add memory observability metrics and retrieval traces #30, Surface retrieval status observability #51, feat: migrate observability to Micrometer #134).

Breaking Changes

• JDBC artifacts and public APIs were restructured (Split JDBC plugin into dialect modules with Jdbi and Hikari-backed access #37). The former memind-plugin-jdbc implementation artifact is now a parent module. Applications must depend on memind-plugin-jdbc-sqlite, memind-plugin-jdbc-mysql, or memind-plugin-jdbc-postgresql; shared integrations can use memind-plugin-jdbc-core. JdbcStore, DriverManagerDataSource, and JdbcExecutor were removed. Plain Java applications must migrate to the relevant dialect plugin factory. Spring Boot applications can continue using memind-plugin-jdbc-starter.
• The server API response contract changed (standardize server API response contract #65). Endpoints now return the standardized ApiResult success/error envelope, public error codes use snake_case, paging uses the new page result contract, and open memory APIs are split into synchronous and asynchronous controllers. Custom clients that parse bare response payloads or call the previous endpoint layout must be updated.
• Retrieval cache APIs were removed (fix: remove retrieval cache #95). RetrievalCache, RetrievalConfig.enableCache(), RetrievalConfig.withoutCache(), and the cache-accepting DefaultMemoryRetriever constructors no longer exist.
• Spring AI model configuration and wiring changed (refactor: route Memind AI through Spring beans #128). Memind no longer constructs provider models from the previous Memind-specific configuration, and the old Memind default chat/embedding model properties were removed. Applications must provide Spring-managed ChatModel/EmbeddingModel beans or configure models through spring.ai.chat-models, spring.ai.embedding-models, and spring.ai.model.*. Slot routing remains under memind.ai.chat.slots.
• The observability integration artifacts changed (feat: migrate observability to Micrometer #134). The old OpenTelemetry tracing plugin/starter and its direct instrumentation path were removed. Replace them with memind-plugin-observability-micrometer or memind-plugin-observability-micrometer-starter, then configure trace export through Micrometer/Spring Boot.
• Project-owned Jackson APIs migrated to Jackson 3 (chore: migrate project code to Jackson 3 #131). Extensions that directly use exposed JSON types must migrate runtime imports from com.fasterxml.jackson.* to tools.jackson.* where applicable and review removed or changed APIs. Jackson annotations remain the compatibility exception.
• The managed framework baseline changed ([Feature]: upgrade Spring Boot to 4.1.0 and Spring AI to 2.0.0 #123). The main reactor now targets Spring Boot 4.1 and Spring AI 2.0 GA. Applications importing the Memind BOM should check dependency convergence and compatibility with custom Spring integrations.

Memory and Retrieval

• Introduced graph relationships and Memory Thread support (feat: add graph relationships and Memory Thread capabilities #28).
• Optimized Memory Thread replay and retrieval status visibility (Optimize memory thread replay reads #50, Surface retrieval status observability #51).
• Improved multimodal ingestion, insight-tree generation, temporal extraction, and extraction prompts (feat: add multimodal rawdata support #23, feat: refactor and optimize insight tree #24, feat: redesign temporal memory item extraction #26, Improve memory item extraction prompt #31).
• Removed the retrieval cache to keep reads consistent with current memory state (fix: remove retrieval cache #95).
• Added configurable AI client routing (Add configurable AI client routing #116).

Platform and Integrations

• Split JDBC persistence into core and database-specific modules for SQLite, MySQL, and PostgreSQL (Split JDBC plugin into dialect modules with Jdbi and Hikari-backed access #37).
• Added Docker Compose deployment support and documentation (feat: docker compose support #45, docs: add Docker Compose quick start to README #60).
• Upgraded to Spring Boot 4.1 and Spring AI 2.0 ([Feature]: upgrade Spring Boot to 4.1.0 and Spring AI to 2.0.0 #123).
• Routed AI provider usage through Spring-managed beans (refactor: route Memind AI through Spring beans #128).
• Migrated the project to Jackson 3 (chore: migrate project code to Jackson 3 #131).
• Added Maven Central release automation (feat: add Maven Central release workflow #44).

Reliability and Security

• Optimized SQL queries and corrected point-in-time temporal overlap behavior (feat: optimize SQL query paths #62, fix: handle point-in-time temporal overlap #63).
• Added recovery for stale Claude Code retries (Recover stale Claude Code retry claims #86).
• Migrated local caching to Caffeine (fix: replace ConversationBufferLocks map with bounded cache #94).
• Made conversation-buffer drains atomic and runtime configuration initialization idempotent (Fix atomic conversation buffer drains for persistent stores #104, fix: make runtime config initialization idempotent #106).
• Sanitized request IDs and hardened URL access against SSRF (fix: sanitize request ids before logging #107, fix: harden HTTP resource fetching against SSRF #109).
• Corrected item deletion and vector-cleanup ordering (fix: delete memory items before vector cleanup #110).

Upgrade Checklist

1. Select the JDBC dialect module required by your database, or retain memind-plugin-jdbc-starter for Spring Boot usage.
2. Migrate server clients to the standardized response envelope and current synchronous/asynchronous endpoints.
3. Remove use of the deleted retrieval cache APIs.
4. Move AI provider/default model configuration to Spring AI properties and verify the required model beans.
5. Replace the old tracing dependency with the Micrometer observability plugin or starter.
6. Check Jackson 3 imports/APIs and Spring Boot 4.1/Spring AI 2.0 dependency compatibility.

Full changelog: 0.1.0...0.2.0