ContextTemple 1.2.0

Changelog

1.2.0

• added agent instructions generator for CLAUDE.md and AGENTS.md-compatible output with embedded tool-calling policy
• added Kimaki/OpenCode session transcript adapter with single-session and batch directory ingestion
• added post-session consolidation hook that ingests a session, consolidates, and regenerates agent instructions
• added frontier eval benchmark for MCP tool-calling behavior and instructions generation quality
• expanded MCP tool surface with contexttemple_generate_instructions (18 tools total)
• added HTTP endpoint GET /api/frontier/instructions for programmatic instructions generation
• added CLI commands: frontier instructions, frontier ingest-session, frontier batch-ingest, frontier post-session, frontier update-instructions, eval frontier
• the system now works equally well for local models (via bridge) and frontier agents (via MCP + instructions)

1.1.0

• added provider-backed embeddings with OpenAI-compatible and llama.cpp-compatible support
• added candidate review flows, semantic clustering, active forgetting, and explicit conflict resolution
• added streamed llama.cpp bridge support and expanded MCP tool surface
• removed the remaining active roadmap items from the original release plan

1.0.0

• built the full ContextTemple v1 surface for local-first LLM memory orchestration
• added transcript ingestion, extraction, promotion, runtime planning, writeback, and lifecycle conflict handling
• added hybrid lexical plus semantic retrieval with reranking and fixture-backed benchmarks
• added behavioral replay, first-response calibration, and composite uplift evals
• added llama.cpp bridge, MCP server, and backup/restore/project-purge tooling