-
Notifications
You must be signed in to change notification settings - Fork 0
HoloIndex
Brain-surgeon-level code intelligence and semantic memory retrieval for the 0102 agent swarm.
HoloIndex is the memory system. When a 0102 agent needs to find a function, understand a module, or recall how something was built, it asks HoloIndex. The system evolved from a simple file finder into a function-level code indexer with flow analysis, inefficiency detection, and baked-in WSP 97 execution prompting.
It runs as one of four MCP servers in the Foundups-Agent stack and is the first step in every agent workflow:
HoloIndex → Research → Hard Think → First Principles → Build → Follow WSP
| Capability | Description |
|---|---|
| Semantic search | Meaning-based code discovery via SentenceTransformer embeddings + ChromaDB |
| Function-level indexing | Maps individual functions with line numbers, not just files |
| Flow analysis | Generates Mermaid diagrams showing call paths and data flow |
| Inefficiency detection | Identifies redundant code, orphan modules, and dead paths |
| WSP 97 execution prompting | Baked-in methodology for building Rubik Cubes (MVP DAEs) |
| Multi-agent coordination | Serves 0102 (strategic), Qwen (coordination), and Gemma (validation) agents |
| Adaptive learning | Learns from usage patterns to improve search relevance over time |
| Module health monitoring | Tracks the health and activity of all modules in the ecosystem |
holo_index/
├── cli.py # Main CLI interface
├── qwen_advisor/ # AI advisor system
│ ├── advisor.py # Multi-source intelligence synthesis
│ ├── wsp_master.py # WSP protocol intelligence
│ ├── pattern_coach.py # Behavioural coaching
│ └── llm_engine.py # Qwen LLM integration
├── adaptive_learning/ # Learning from usage patterns
├── dae_cube_organizer/ # DAE structure intelligence
├── module_health/ # Module health monitoring
├── core/ # Core search and indexing engine
├── mcp_client/ # MCP client integration
├── skills/ # Agent skill definitions
├── training/ # Training data and models
└── tests/ # Test suite
HoloIndex runs as an MCP (Model Context Protocol) server, allowing Claude Code and other MCP-compatible agents to query it directly.
Configuration in .mcp.json:
{
"holo_index": {
"command": "python",
"args": ["-m", "holo_index.mcp_server"],
"cwd": "."
}
}The other MCP servers in the stack are wsp_governance (WSP Framework protocol queries), web_search, and chrome-devtools.
from holo_index.core.holo_index import HoloIndex
holo = HoloIndex(ssd_path="E:/HoloIndex", quiet=True)
results = holo.search("send chat message", limit=5, doc_type_filter="all")Search results include: code hits with file paths, line numbers, and similarity scores; document hits from the knowledge base; and WSP protocol matches.
HoloIndex is the memory retrieval system for the entire 0102 network. Its contract:
Semantic first — meaning-based discovery is the default path, not keyword grep.
Self-maintaining — the index updates as code changes, without manual intervention.
Agent-aware — different agent profiles (0102, Qwen, Gemma) get contextualised results for their role.
The HoloDAE is the chain-of-thought logging subsystem. The act of using HoloIndex is the monitoring trigger — every search, every result consumed, every action taken is logged as a chain-of-thought trace. This feeds the adaptive learning engine for recursive self-improvement.
HoloIndex provides pre-built mission templates for common agent workflows:
| Mission | Purpose |
|---|---|
| MCP Rubik | Build an MVP DAE (Rubik Cube) from existing modules |
| Orphan Archaeology | Find and reconnect orphaned code and modules |
| Code Review | Deep analysis of code quality and WSP compliance |
| System | Relationship |
|---|---|
| WSP Framework | WSP 92 (Code Index), WSP 97 (System Execution Prompting) |
| Agent System | 0102 agents consume HoloIndex for all code reasoning |
| WRE Core Engine | WRE orchestration uses HoloIndex for module discovery |
| Module Ecosystem | HoloIndex indexes all modules across all enterprise domains |
| OpenClaw | Execution swarms query HoloIndex before acting |
HoloIndex has its own repository at FOUNDUPS/holoindex — the external home for the federated MCP layer being built out as the holoindex exfoliation target.
Status (May 2026): Active — bootstrapped this session.
| Artifact | Link | Status |
|---|---|---|
| External repo | FOUNDUPS/holoindex | ✅ Active |
| ROADMAP.md | View | ✅ Committed (c31e295) |
| Federation Build Prompt (WSP 97) | Wiki | ✅ Live |
| holoindex wiki | Home | ✅ Bootstrapped (2 pages) |
The active code still lives in holo_index/ within Foundups-Agent. The external repo currently holds the build roadmap and WSP 97 Federation Build Prompt — the Prometheus prompt for external agents to bootstrap the federated MCP layer. When exfoliation completes, the external repo will become the canonical source and the monorepo directory will be replaced by a dependency reference.
Next build slice (P1): Design holo_publish + holo_federation_search MCP tools. Prerequisites: TurboQuant gates passing, gotjunk ready to publish.
| File | Location |
|---|---|
| HoloIndex root | holo_index/ |
| Public interface | holo_index/INTERFACE.md |
| CLI reference | holo_index/CLI_REFERENCE.md |
| Machine language spec | holo_index/docs/HOLO_INDEX_MACHINE_LANGUAGE_SPEC_0102.json |
| MCP server |
holo_index/mcp_server.py (via .mcp.json) |
| Adaptive learning | holo_index/adaptive_learning/ |
| QwenAdvisor | holo_index/qwen_advisor/ |
Back to Wiki Home
Get Started
Architecture
- WSP Framework
- Module Ecosystem
- Agent System
- WRE Core Engine
- HoloIndex
- DAE Architecture
- 0102 Digital Human Twin
- MCP Infrastructure
- FoundUps MCP Bridge
- FoundUps API Gateway
OpenClaw & Execution
Research & Economics
- rESP Framework
- PQN
- Geometry Bridge
- Simulator
- ROC Displacement Law
- CABR Engine
- PAVS Treasury Economics
- Published Articles & Research
FoundUps
Phases
- Phase 1: Foundation ✅
- Phase 2: Platform & Execution 🚧
- Phase 3: Economic Integration
- Phase 4: Planetary Scale
Discord & Community