v1.1.0 · Expand recall on demand · Review failures and revise Experiences · Unify tag management · Improve Agent handoffs and diagnostics

PowerContext 1.1.0 adds an optional recall sufficiency gate, evidence-based recurring failure records, tags across all built-in Artifact families, and native MiniMax embeddings, alongside expanded Pi and OpenClaw workflows and better Source provenance.

Highlights

• Optionally expand thin recall with at most two additional searches while preserving Scopes, families, result limits, and byte budgets. The gate is disabled by default. Recognized standalone execution instructions no longer count as lexical evidence; Topic query embedding during preparation falls back to full-text search after 250 ms.
• Manage Topic Memory through generic Artifact create, replace, revision, lineage, tag, and publication operations, and customize Topic Memory processing Prompts by Scope.
• Organize all seven built-in Artifact families with tags, including Profile and saved Prompt Artifacts. Queries that omit families now cover every built-in family.
• Record exact evidence of recurring failures and verified avoidance. Three consecutive recurrences against one Experience revision can trigger a revision proposal for review when the repair surface is Experience content; approval and execution remain explicit.
• Use the Pi Full core integration profile for Memory changes, statistics, Topic Memory queries, structured Handoffs, Task Outcomes, candidate generation and inspection, and External Skill discovery and import. Durable writes require confirmation; candidate generation does not approve or install a Skill.
• Use OpenClaw /pc commands and work coordination workflows, and follow aligned Agent guidance and layered Skill routing across integrations.
• Inspect recalled-context token reduction in Claude Code, OpenCode, and Codex, observe DSH recall and capture status through /pc, and trace Memory writes.
• Read Profiles in the personal Dashboard, export the exact rendered Handoff Markdown, and read Experiences through dedicated CLI commands.
• Strengthen resource access decisions with shared authorization filters and a transaction snapshot boundary, and reduce database transactions when reading Scope statistics.
• Batch SQLite vector completeness checks while detecting stale Memory heads and missing or mismatched vectors.
• Use native MiniMax embeddings with separate document and query requests, batching, response validation, and timeout handling.
• Preserve registered Source types and identifiers in Artifact provenance, process persisted remote Source evidence for Topic Memory, distinguish missing Memory from pending ownership, support Markdown Handoff Reports over MCP, and extend WorkBuddy prompt Hooks to 30 seconds.
• Improve Windows integration setup, command encoding, service inspection, and file URI handling, including native Codex MCP authentication verification. Windows remains experimental.
• Upgrade the Server, clients, and Agent integrations together. SourceTypeReference.source_type is now an open string. Back up existing databases and complete startup schema updates with one Server instance before starting other instances. Older deployments with unfinished Artifact processing migration must complete that migration first.

uv tool install --force "powercontext[cli,server]==1.1.0"