Skip to content

Memind 0.2.0

Latest
Latest

Choose a tag to compare

View all tags
@me0106 me0106 released this 13 Aug 08:36
0.2.0
7536460
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

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 (#28, #50).
  • Added multimodal raw data ingestion and improved insight/temporal extraction (#23, #24, #26, #31).
  • Added admin APIs and UI for buffers, dashboards, items, graphs, and memory management (#39, #43, #87).
  • Added Java, Python, TypeScript, Go, and Rust SDKs (#55, #58, #69, #71, #72).
  • Added integrations for Claude Code, Codex, OpenClaw, and Hermes (#32, #35, #79, #82, #102, #103).
  • Added an HTTP MCP server and expanded MCP tools (#76, #111).
  • Added durable OpenAPI ingestion, generated server OpenAPI, and a standardized API response contract (#61, #65, #80).
  • Migrated observability to Micrometer with memory-operation metrics and retrieval traces (#30, #51, #134).

Breaking Changes

  • JDBC artifacts and public APIs were restructured (#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 (#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 (#95). RetrievalCache, RetrievalConfig.enableCache(), RetrievalConfig.withoutCache(), and the cache-accepting DefaultMemoryRetriever constructors no longer exist.
  • Spring AI model configuration and wiring changed (#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 (#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 (#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 (#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 (#28).
  • Optimized Memory Thread replay and retrieval status visibility (#50, #51).
  • Improved multimodal ingestion, insight-tree generation, temporal extraction, and extraction prompts (#23, #24, #26, #31).
  • Removed the retrieval cache to keep reads consistent with current memory state (#95).
  • Added configurable AI client routing (#116).

Platform and Integrations

  • Split JDBC persistence into core and database-specific modules for SQLite, MySQL, and PostgreSQL (#37).
  • Added Docker Compose deployment support and documentation (#45, #60).
  • Upgraded to Spring Boot 4.1 and Spring AI 2.0 (#123).
  • Routed AI provider usage through Spring-managed beans (#128).
  • Migrated the project to Jackson 3 (#131).
  • Added Maven Central release automation (#44).

Reliability and Security

  • Optimized SQL queries and corrected point-in-time temporal overlap behavior (#62, #63).
  • Added recovery for stale Claude Code retries (#86).
  • Migrated local caching to Caffeine (#94).
  • Made conversation-buffer drains atomic and runtime configuration initialization idempotent (#104, #106).
  • Sanitized request IDs and hardened URL access against SSRF (#107, #109).
  • Corrected item deletion and vector-cleanup ordering (#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

Assets 2
Loading