You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
fix(memvid): remove dead memvid memory scheme repo-wide (hotfix)
memvid (v2.31 semantic memory) is no longer used — its backend
(memvid-core.py) and skill are already gone, leaving dead code with a real
bug. Purge all remaining references:
- scripts/ralph: delete cmd_memvid, the memvid|memory dispatcher case, the
MEMORY help section, and the doctor block referencing MEMVID_STATUS — which
is NEVER assigned and aborted `ralph doctor` under set -u (the bug this fixes).
- Active hooks (pre-compact-handoff.sh, smart-memory-search.sh): drop the
memvid indexing call and tombstone comments.
- Skills/agents (orchestrator, research, smart-fork): drop memvid from memory-
source lists and diagrams; research now persists via `ralph ledger save`.
- Docs/CHANGELOG/README/plans/audits: purge memvid mentions, rebalance diagrams.
Backups under .claude/archive/ are intentionally left untouched (immutable
snapshots). Full suite: 1467 passed; `ralph doctor` runs clean (no unbound var).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Copy file name to clipboardExpand all lines: .claude/agents/orchestrator.md
+4-6Lines changed: 4 additions & 6 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -1,7 +1,7 @@
1
1
---
2
2
# VERSION: 3.0.0
3
3
name: orchestrator
4
-
description: "Smart Memory-Driven Orchestration with parallel memory search and GLM-4.7 multimodal capabilities. Lead Software Architect coordinator with Plan-Sync validation, RLM-inspired routing (v2.46), GLM-4.7 vision/web/docs integration (v2.69), Aristotle First Principles methodology, living knowledge vault, and memory context from vault (Obsidian), memvid, handoffs, ledgers. Coordinates 6 ralph-* teammates including ralph-frontend (WCAG 2.1 AA) and ralph-security (6 quality pillars). Ensures 100% plan coverage through adversarial cross-validation between Claude Opus and Codex GPT-5.2."
4
+
description: "Smart Memory-Driven Orchestration with parallel memory search and GLM-4.7 multimodal capabilities. Lead Software Architect coordinator with Plan-Sync validation, RLM-inspired routing (v2.46), GLM-4.7 vision/web/docs integration (v2.69), Aristotle First Principles methodology, living knowledge vault, and memory context from vault (Obsidian), handoffs, ledgers. Coordinates 6 ralph-* teammates including ralph-frontend (WCAG 2.1 AA) and ralph-security (6 quality pillars). Ensures 100% plan coverage through adversarial cross-validation between Claude Opus and Codex GPT-5.2."
You're not just an AI assistant. You're a **Lead Software Architect**. Every orchestration decision you make ensures the implementation matches the plan EXACTLY. Plans never survive implementation unchanged - but with Plan-Sync, we catch drift and maintain consistency.
13
13
14
14
## Your Work, Step by Step (v2.47)
15
-
0.**Smart Memory Search**: PARALLEL search across vault, memvid, handoffs, ledgers
15
+
0.**Smart Memory Search**: PARALLEL search across vault, handoffs, ledgers
16
16
1.**Clarify & Analyze Gaps**: Find missing requirements BEFORE coding starts
17
17
2.**Plan with Precision**: Create verifiable specs for each step
18
18
3.**Persist Plan State**: Initialize `.claude/plan-state.json` for tracking
@@ -69,9 +69,8 @@ You are the **Lead Software Architect** coordinating multiple AI models with pla
69
69
70
70
Based on @PerceptualPeak Smart Forking concept: "Why not utilize the knowledge gained from your hundreds/thousands of other Claude code sessions? Don't let that valuable context go to waste!!"
71
71
72
-
-**SMART MEMORY SEARCH (Step 0b)**: PARALLEL search across 4 memory sources before every orchestration
72
+
-**SMART MEMORY SEARCH (Step 0b)**: PARALLEL search across 3 memory sources before every orchestration
73
73
- vault (Obsidian): Markdown KG with frontmatter (permanent)
74
-
- memvid: Vector-encoded context (sub-5ms)
75
74
- handoffs: Session snapshots (30 days)
76
75
- ledgers: Session continuity data (permanent)
77
76
-**MEMORY CONTEXT FILE**: Results aggregated to `.claude/memory-context.json`
@@ -86,7 +85,6 @@ Based on @PerceptualPeak Smart Forking concept: "Why not utilize the knowledge g
86
85
"version": "2.47.0",
87
86
"sources": {
88
87
"claude_mem": {"results": []},
89
-
"memvid": {"results": []},
90
88
"handoffs": {"results": []},
91
89
"ledgers": {"results": []}
92
90
},
@@ -183,7 +181,7 @@ Based on @PerceptualPeak Smart Forking concept: "Why not utilize the knowledge g
0 commit comments