How neagi/reefiki works
REEFIKI is a local, multi-project wiki memory for AI agents. Its principal workflow distills session material: capture into an inbox, filter for reusable knowledge, save it as typed Markdown wiki pages, then link and index those pages so a later agent can retrieve context. Supporting capabilities include onboarding and project setup, review and promotion governance, health checks, and a read-only local operations dashboard. Some subsystem links are based on README descriptions because their implementation was not sampled; those are marked unverified where appropriate.
Main entry points: review_queues.py Review queues · index_search.py Search index · snapshot.py Workspace snapshot · selection_gate.py Knowledge filter
Components 20
Memory workflow
- Project inbox
- Knowledge filter
scripts/reefiki_core/selection_gate.py - Memory save
scripts/reefiki_core/save.py - Markdown wiki
- Markdown parsing
scripts/reefiki_core/markdown.py - Wiki writing
scripts/reefiki_core/wiki_write.py
Recall and indexing
- Search index
scripts/reefiki_core/index_search.py - Context packing
scripts/reefiki_core/memory_pack.py - Memory lookup
scripts/reefiki_core/memory_lookup.py
Governance and safety
- Review queues
scripts/reefiki_core/review_queues.py - Memory promotion
scripts/reefiki_core/promotion.py - Policy safety
scripts/reefiki_memory/__init__.py - Health checks
scripts/reefiki_core/health.py
Workspace operations
- Project onboarding
scripts/reefiki_core/onboarding.py - Workspace setup
scripts/reefiki_core/init_workspace.py - REEFIKI CLI
scripts/reefiki.py
Operations dashboard
- Dashboard server
scripts/reefiki_core/ops_dashboard/server.py - Workspace snapshot
scripts/reefiki_core/ops_dashboard/snapshot.py
Other components
- AI agent
- Local operator
Connections 22
- AI agent → Project inbox: captures material
README.md - Project inbox → Knowledge filter: screens for reuse
README.md - Knowledge filter → Memory save: selects knowledge
README.md - Memory save → Wiki writing: writes pages
- Wiki writing → Markdown wiki: creates pages
README.md - Markdown wiki → Search index: indexes pages
- Search index → Memory lookup: searches memory
- Memory lookup → Context packing: packs context
- Context packing → AI agent: returns context
- AI agent → Review queues: reviews candidates
- Review queues → Memory promotion: queues knowledge
- Policy safety → Knowledge filter: checks inputs
- Health checks → Review queues: checks queues
scripts/reefiki_core/health.py - Project onboarding → Workspace setup: guides setup
- AI agent → REEFIKI CLI: runs commands
scripts/reefiki.py - REEFIKI CLI → Workspace setup: dispatches setup
scripts/reefiki.py - Local operator → Dashboard server: requests status
scripts/reefiki_core/ops_dashboard/server.py - Dashboard server → Workspace snapshot: serves snapshot
scripts/reefiki_core/ops_dashboard/server.py - Workspace snapshot → Health checks: reads health
scripts/reefiki_core/ops_dashboard/snapshot.py - Workspace snapshot → Review queues: reads queues
scripts/reefiki_core/ops_dashboard/snapshot.py - Search index → Markdown parsing: parses pages
scripts/reefiki_core/index_search.py - Markdown parsing → Markdown wiki: parses page format
scripts/reefiki_core/markdown.py