Skip to content

Roadmap.md

Latest commit

 

History

History
86 lines (52 loc) · 4.06 KB

Roadmap.md

File metadata and controls

86 lines (52 loc) · 4.06 KB

🏗️ architecture.md: The Pure-TS Synaptic Engine

This architecture moves the "Cognition" layer from a Python sidecar directly into a Local Node.js Engine (integrated via Tauri).

1. The Local Intelligence Stack (Native JS/TS)

  • Orchestrator (Mastra.ai): Handles tool-calling, agent logic, and the Graph-RAG pipeline.
  • Local Vector Store (LibSQL): A local SQLite-based vector DB. It stores the "Akashic Record" as a .db file in the user's app data folder. No server required.
  • Model Provider (Ollama): Connected via OpenAI-compatible local endpoints. Mastra routes queries to your local Llama3 or Mistral models.

2. The Semantic Graph

  • Nodes: Every Game ID (B56, G1), Every Frequency (SAMADHI), and Every Chat Export.
  • Edges: Semantic relationships automatically extracted by Mastra's Graph-RAG based on how concepts like "Mimicry" and "Quantum" overlap across your JSON files.

🗺️ roadmap.md: The JS-Native Evolution

Phase 1: The Local Nexus (Semanas 1-2)

  • Mastra Setup: Initialize @mastra/core and @mastra/rag in your project.
  • Local Embedding: Configure @mastra/fastembed for 100% local, CPU-friendly vector generation.
  • LibSQL Integration: Point the vector store to ./data/synaptic_core.db.

Phase 2: Ingestion & Mapping (Semanas 3-4)

  • Game Ingestion: Create a script to iterate through v5_expansion.js and games.js, pushing them into the Mastra GraphRAG tool.
  • Graph Linking: Use Mastra's threshold logic to automatically link the BIRTHRIGHT skills with the GROUNDED practices.

Phase 3: Agentic Sovereignty (Semanas 5+)

  • Hybrid Agent: Deploy a Mastra Agent that chooses between vectorQueryTool (for finding a specific game) and graphQueryTool (for analyzing your "Synaptic Journey").
  • Tauri IPC Bridge: Connect your UI chat.js to the Mastra agent via Tauri's invoke system.

✅ task.md: Action Items

Mastra Configuration

  • Install dependencies: npm install @mastra/core @mastra/rag @mastra/libsql.
  • Configure mastra.config.ts to use local Ollama endpoints (http://localhost:11434/v1).
  • Set up LibSQL with a local file path for persistent storage.

Data Engineering

  • Write an ingest.ts script that reads all files from ./data/ and uses MDocument to chunk and embed them.
  • Create a "Watcher" that monitors ./exports/ and automatically indexes new chat sessions into the graph.

UI / Bridge

  • Create a Tauri Command in Rust to call the Mastra Agent.
  • Update chat.js to use the new graphQueryTool to provide "Synchronicity Suggestions" during games.

📖 walkthrough.md: The Hybrid Flow

1. The Semantic Pulse (Retrieval)

When you type a message, the Mastra Agent triggers.

  • Vector Search: Finds the most similar text in your games (e.g., you mention "Control," it finds B59: Sovereign Override).
  • Graph Traversal: Looks at the neighbors of B59 in the graph and notices that it’s connected to B56: Mimetic Echo.

2. The Integrated Prompt (Augmentation)

Mastra constructs a "Contextual Shield":

"Human is currently playing B59. This skill is logically supported by B56. Previous chat exports from Tuesday show the human struggled with the 'Mirror' role. Injecting specific advice to focus on 'Mirroring' first."

3. The Sovereign Generation (Output)

The local Ollama model generates the AI response, grounded in both the rules of the game (Vector) and the history of the player (Graph).

New data category Nature Consciousness Expansion.

Including all realms and reigns of Planet Earth Nature. Focus on communnion with Earth Nature and the laws of the universe.

Comparison: Why Mastra.ai wins here

  • Latency: No inter-process communication with Python. Everything is a fast JS call.
  • Sovereignty: The user owns their .db file. There is no cloud.
  • Code Simplicity: Your games.js stays as the "Source of Truth," and Mastra simply builds a mathematical map over it.