v1.1.0

v1.1.0

The first minor release after v1.0.5, gathering the fixes and hardening that surfaced from real upgrades — migration/drift repair, MCP read-path resilience, and a rounded-out fp permission surface. A quiet, clean release line.

Features

• MCP server-instructions block for tool discovery and routing, so agents pick the right tool
• MCP path lookups now steer to footprinter_folder, with an empty-result hint
• footprinter_folder surfaces unlisted_file_count on the VIEWER payload
• MCP read path hardened against live schema migrations (no more transient "no such column" / 0-record status)
• footprinter_read hoists identity fields to the top of the response
• Search summary reports the total match count, not just the number returned (LIMIT+1 truncation detection)
• fp permission set bulk-records policy updates from a CSV
• fp permission check extended to entity-id scopes (email/chat/visit + numeric file/folder ids)
• fp doctor link-folders backfills parent_folder_id for existing local folders
• fp doctor architecture WARN now gives concrete pipx reinstall remediation
• .claude/ and .context/ indexed as listed by default
• MCP setup filters client paths to installed clients only
• Vectorization status filter is now config-driven; hardcoded security limits moved to config
• New verify-install QA tier with a --with-pytest mode

Fixes

• Visibility-enum migration no longer crashes on upgrade — CHECK constraints rebuild before values update
• Added projects.project_name → name migration for existing databases
• Closed the clients.updated_at schema-drift gap and added an fp doctor drift guard
• Bare fp ingest no longer re-vectorizes the whole corpus or loses progress on interrupt
• Conditional vectorized_at clearing + a repair command
• MCP VIEWER leak fixes: hide unlisted projects from navigation; gate unlisted/removed folders on exact-path lookup
• parent_folder_id populated for local folders at ingest
• fp doctor accuracy: separate DB-open errors from FTS health-check failures; mcp_config check validates command+args and fails closed on lookup failure
• Handler return values now propagate to the process exit code
• fp permission check not-found tip uses scope strings; DB closed on all exit paths

Maintenance

~57 internal commits: extraction of search/status/policy logic into service layers, schema.py split into DDL/FTS modules, connection-factory centralization, config-limit unification, CI lint enforcement + Node 24 runner bump, QA-harness hardening, and documentation/reference drift cleanup. No user-facing behavior change.