# ORA System Review Task List

- [x] Initial System Review <!-- id: 10 -->
    - [x] Explore codebase structure <!-- id: 0 -->
    - [x] Identify key configuration and entry points <!-- id: 1 -->
    - [x] Analyze Services <!-- id: 2 -->
        - [x] Review `orchestrator.py` <!-- id: 5 -->
        - [x] Review `memory.py` <!-- id: 6 -->
        - [x] Review `guardrails.py` <!-- id: 7 -->
        - [x] Review `intent.py` & `emotion.py` <!-- id: 8 -->
        - [x] Review spiritual practices (prayer/discernment/lectio) <!-- id: 9 -->
        - [x] Review infrastructure (main.py, config.py) <!-- id: 11 -->
        - [x] Review peripheral services (llm.py, telemetry.py) <!-- id: 12 -->
    - [x] Synthesize Review Report <!-- id: 13 -->
        - [x] Document Architecture/Data Flow <!-- id: 14 -->
        - [x] Identify Missing Persistent Layers <!-- id: 15 -->
    - [x] Define "Not to build" list <!-- id: 16 -->
    - [x] Create One-week MVP plan <!-- id: 17 -->

- [x] Implement MVP Core Services <!-- id: 18 -->
    - [x] Implement Audit Persistence (File logging) <!-- id: 19 -->
    - [x] Implement Journal Service (SQLite) <!-- id: 20 -->
    - [x] Implement User Profiles (SQLite) <!-- id: 21 -->
    - [x] Connect Bible RAG (Vector Search) <!-- id: 22 -->

- [x] Phase 2: Advanced Reasoning & Study <!-- id: 30 -->
    - [x] Research RLM Blog Post <!-- id: 31 -->
    - [x] Implement RLM (for Deep Study) <!-- id: 31_sub -->
        - [x] Design Python REPL Sandbox <!-- id: 32 -->
        - [x] Implement `RLMService` wrapper <!-- id: 33 -->
    - [x] Integrate TRM (for Logic Solvers) <!-- id: 34 -->
        - [x] Define Theological Logic dataset schema <!-- id: 35 -->
    - [x] Phase 2 Architecture Upgrades <!-- id: 44 -->
        - [x] Migrate Storage to LanceDB <!-- id: 45 -->
        - [x] Ingest `bible-nlp/biblenlp-corpus` <!-- id: 51 -->
    - [x] Evaluate Pre-trained Bible Models <!-- id: 53_eval -->
        - [x] Test `Christian-Bible-Expert-12B` <!-- id: 54 -->
        - [x] Create `Gabriel-8x7B` via Ollama <!-- id: 55 -->
    - [x] Setup Unsloth Fine-tuning Pipeline <!-- id: 47 -->
    - [x] Start "Reasoning Trace" Dataset Collection <!-- id: 48 -->

- [x] Phase 3: Frontend Mastery & Integration <!-- id: 60 -->
    - [x] Migrate `soapbox-demo` to root as `frontend`
    - [x] Connect ChatWidget to ORA Backend (Port 6000)
    - [x] Implement Agent Handoff Visualizations
    - [x] Create Handover Prompt for Claude

- [/] Phase 4: Model Optimization & Fine-tuning <!-- id: 70 -->
    - [/] Benchmark `gabriel-8x7b` baseline capabilities
    - [x] Curate theological dataset for specialist fine-tuning
    - [x] Perform LoRA fine-tuning on `gabriel-8x7b` using Unsloth
    - [ ] Evaluate fine-tuned model vs baseline
    - [ ] Deploy optimized reasoning model to production

- [x] General Research Phase <!-- id: 39 -->
    - [x] Research Local Vector Databases (Chroma/LanceDB) <!-- id: 40 -->
    - [x] Research Secure REPL Sandboxes (for RLM) <!-- id: 42 -->
    - [x] Research Fine-tuning Logic for Small Models <!-- id: 43 -->
    - [x] Evaluate new Bible Datasets (Helsinki, BibleNLP, etc.) <!-- id: 50 -->
    - [x] Evaluate Pre-trained Bible Models (Gabriel, Christian-Bible-Expert) <!-- id: 53 -->