interest-memory v0.1.0

Initial open-source release of interest-memory — a lightweight, self-hostable memory service for AI agents.

Highlights

• Prefix-window parallel extraction → three-stage correction (subjectivity exemption, relation verdict, evidence locators) → per-interest-point wiki write → related-page reconciliation
• Strict audit trail: change_log, evidence located to URL/turns/query, contradiction closure
• Progressive disclosure: lean recall injection vs full memory_search on demand
• Resource footprint (measured): ~19 MB binary, idle 20–40 MB RSS, peak <75 MB; initial disk ~20 MB (~38 MB after a week of use, transcripts dominate)
• Self-host friendly: single binary + config, OpenAI-compatible base_url (local Ollama/vLLM capable), guided installer + systemd
• Bridges: Hermes, opencode, pi, openclaw, claudecode, codex, reasonix + shared MCP server
• CI: go test -race + vet + fmt, node:test for all bridges

Docs

Bilingual README (zh/en) with architecture diagram, REST API reference, resource benchmark; CHANGELOG / CONTRIBUTING / SECURITY.