feat(mem0-plugin): improve auto-triggering — pre-fetch, dedup, skill enforcement v0.2.3 (#5237)

2026-05-23 17:20:54 +05:30
parent 16a7702d09
commit 99beb007ab
54 changed files with 1817 additions and 1162 deletions
@@ -0,0 +1,50 @@
+---
+name: context-loader
+description: Pre-load relevant memories for current task
+---
+
+# Context Loader
+
+Pre-fetches relevant memories to prime context before working on a task.
+
+## When to use
+
+- Session start (auto-triggered by `on_session_start.sh` for onboarded projects)
+- User starts work on a specific feature or file set
+- Complex multi-step task begins
+- User says "what do we know about X" or "context for X"
+
+## Steps
+
+1. **Extract topics** from current message/task. Identify: file paths, module names, feature areas, error patterns.
+
+2. **Run 2-4 parallel `search_memories` calls** with different angles:
+
+   | Query angle | Filter | Purpose |
+   |---|---|---|
+   | Feature/module name | `{"metadata": {"type": "decision"}}` | Architecture decisions |
+   | File paths mentioned | `{"metadata": {"type": "convention"}}` | Coding patterns |
+   | Error keywords (if any) | `{"metadata": {"type": "anti_pattern"}}` | Known pitfalls |
+   | Broad project context | no metadata filter | Catch-all |
+
+   All calls must include `user_id` and `app_id` filters.
+
+3. **Deduplicate** results by memory ID across all search responses.
+
+4. **Output compact context block** (max 10 memories):
+
+```
+context-loader: loaded <N> memories for "<task summary>"
+  - [decision] <content> [mem0:<short_id>]
+  - [convention] <content> [mem0:<short_id>]
+  - [anti_pattern] <content> [mem0:<short_id>]
+```
+
+5. If **zero results**: output nothing. Don't announce empty context.
+
+## Constraints
+
+- **Read-only** — never modify or delete memories
+- **Max 10 memories** returned (most relevant only)
+- **Silent on empty** — only surfaces findings if relevant context exists
+- Skip memories already visible in current session context