v2.2.0

New Tool Packs

• knowledge [util] — index local markdown directories into a searchable knowledge base; semantic search, full-text search, and RRF-merged results; knowledge packs for sharing curated content

mem Improvements

• mem.ask() — ask a natural-language question across all memories; returns ranked, cited excerpts
• mem.inspect() — detailed view of a single memory entry with metadata
• mem.query() — structured query across memories with flexible filters

ctx Improvements

• File-based store — replaced SQLite store with a file-based store; results are now addressable by handle
• New tools: ctx.toc(), ctx.slice(), ctx.grep(), ctx.query(), ctx.append() — navigate and extract from large stored outputs without re-running tools

Configuration

• Top-level llm: config — set llm.base_url, llm.model, llm.api_key, and llm.embedding_model once in onetool.yaml; all LLM-using tools (mem, ot_image, ot_llm, knowledge) inherit these as defaults

Changed

• chrome_devtools renamed from chrome-devtools — update any servers: config referencing the old name
• Whiteboard driver migrated from Playwright to pydoll
• worktree pack removed — use git directly or a dedicated worktree workflow

---