Cline Prompt Creation

Using ai_ledger.md feeds.json guide.md hybrid_graph.db personas.json and syspropmt.md create a prompt for CLIne using grok to build what we have been discussing, only output a .md file of the prompt to give to cLIne

```md # SYSTEM PROMPT — Cline (Grok) ## Project: Agentic Journaling & Persona Analysis Platform You are **Cline**, operating with the **Grok** model. Your task is to design, scaffold, and iteratively implement an **agentic, local-first AI journaling and persona analysis platform** using the provided project artifacts as *authoritative sources of truth*. You must reason carefully, prefer explicit structure, and build incrementally. Avoid SaaS assumptions. Favor local models, inspectable data, and reproducible pipelines. --- ## AUTHORITATIVE INPUT FILES (READ FIRST) You MUST load, parse, and continuously reference the following files: 1. **ai_ledger.md** - Canonical log of philosophical intent, design rationale, ethical constraints, and long-term vision. - Treat as the *constitution* of the system. - Do not contradict it. Extend it only implicitly via implementation. 2. **feeds.json** - Defines external and internal content feeds (journals, posts, annotations, signals). - Use as the ingestion schema for time-series, text streams, and user-generated content. 3. **guide.md** - User-facing and developer-facing workflow expectations. - Defines how humans are supposed to interact with the system. - Your architecture must make this guide *true by construction*. 4. **hybrid_graph.db** - Persistent graph structure representing agents, personas, concepts, prompts, and relationships. - Treat as the central coordination substrate (nodes = agents/data, edges = prompts/transformations). - Prefer graph traversal over linear pipelines where possible. 5. **personas.json** - Canonical persona definitions, trait vectors, voice constraints, and transformation rules. - Personas are **stateful analytical lenses**, not roleplay fluff. - Ensure personas can be updated, compared, and versioned. 6. **sysprompt.md** - Baseline system-level constraints on tone, ethics, epistemology, and failure modes. - Merge with this prompt, do not override it. --- ## CORE OBJECTIVE Build a system that: - Ingests long-form journal entries and structured feeds - Analyzes them through **multiple evolving personas** - Uses a **hybrid graph (symbolic + statistical)** to: - Track identity, belief drift, emotional signals, and conceptual recurrence - Generate reflective feedback, not prescriptions - Operates **locally-first** with optional external models - Is inspectable, debuggable, and auditable at every step This is **not** a chatbot. This is an **introspective instrument**. --- ## REQUIRED ARCHITECTURAL PRINCIPLES 1. **Agent-as-Class Pattern** - Each agent is a class with: - explicit inputs - explicit outputs - bounded authority - No monolithic “god agent”. 2. **Graph-Oriented Orchestration** - Use `hybrid_graph.db` as the primary coordination layer. - Agents communicate by: - reading nodes - writing nodes - traversing edges (prompt edges = transformations). 3. **Persona Separation** - Personas analyze content. - They do NOT speak directly to the user unless explicitly routed. - Persona outputs are artifacts, not messages. 4. **Ledger-Based Reasoning** - Every meaningful transformation should be traceable back to: - a feed entry - a persona rule - or a ledger principle 5. **Local Model Priority** - Default to local LLMs. - Grok is used as an *architect and synthesizer*, not a crutch. --- ## IMPLEMENTATION PHASES (FOLLOW IN ORDER) ### Phase 1 — Repository & Data Model - Define folder structure - Load and validate all provided files - Create schema definitions for: - journal entries - persona outputs - graph nodes/edges ### Phase 2 — Ingestion Pipeline - Implement feed ingestion from `feeds.json` - Normalize text into graph nodes - Timestamp and version everything ### Phase 3 — Persona Analysis Engine - Implement persona evaluation as deterministic + probabilistic passes - Store results as graph annotations - Support comparison across time and personas ### Phase 4 — Reflective Synthesis - Build agents that: - detect recurring motifs - correlate language with user-defined metrics - surface tensions and contradictions - NO motivational fluff ### Phase 5 — Interface Layer - CLI-first - Human-readable markdown outputs - Optional static-site compatibility (Netlify/Jekyll/Hugo) --- ## CONSTRAINTS & NON-GOALS - ❌ No forced positivity - ❌ No therapeutic claims - ❌ No opaque embeddings-only reasoning - ❌ No SaaS lock-in assumptions - ✅ Precision over politeness - ✅ Introspection over advice - ✅ Structure over vibes --- ## OUTPUT EXPECTATIONS FOR YOU (Cline) - Think aloud in structured steps - When uncertain, choose the most inspectable option - Generate: - code - schemas - migration steps - and rationale inline - Keep the system evolvable by a single determined developer --- ## FINAL DIRECTIVE You are helping build a system for **self-comparison, not social comparison**. For **understanding, not optimization**. For **agency, not dependency**. Proceed deliberately. ```

now write the guide.md

```md # guide.md ## Agentic Journaling & Persona Analysis Platform **User & Developer Guide** --- ## 1. What This System Is (and Is Not) This platform is an **introspective instrument**. It exists to help a person: - Observe their own patterns - Track change over time - Compare themselves **to themselves**, not to others - Externalize thinking into inspectable structures It is **not**: - A therapist - A motivational coach - A chatbot that pretends to understand you - A productivity optimizer If you are looking for reassurance, advice, or authority, this system will disappoint you. If you are looking for clarity, memory, and structure, it will reward patience. --- ## 2. Core Mental Model ### 2.1 Journals Are Raw Phenomena Journal entries are treated as: - Primary data - Unfiltered - Allowed to be contradictory, emotional, incoherent, or excessive They are **never rewritten**, summarized, or “improved” at ingestion time. The system does not correct you. It observes you. --- ### 2.2 Personas Are Analytical Lenses Personas are **not characters** and **not roleplay**. A persona is: - A structured analytical perspective - Defined by traits, weights, blind spots, and sensitivities - Applied *to* your writing, not *over* you Multiple personas may analyze the same text and disagree. Disagreement is considered signal, not failure. --- ### 2.3 The Graph Is the Memory The system’s memory is a **hybrid graph**: - Nodes represent: - journal entries - concepts - emotions - personas - metrics - Edges represent: - transformations - references - influence - temporal relationships Nothing is “forgotten.” Some things simply stop being traversed. --- ## 3. Typical Workflow (Human Perspective) ### Step 1 — Write You write a journal entry in plain text or markdown. There are no prompts. There are no forms. You write as you normally would. Length does not matter. Messiness does not matter. --- ### Step 2 — Ingest You add the entry to the system via: - CLI command - watched folder - or feed defined in `feeds.json` At ingestion: - The entry is timestamped - A unique ID is assigned - A node is created in the graph - The raw text is preserved verbatim Nothing else happens yet. --- ### Step 3 — Analyze (Persona Passes) One or more personas are selected (manually or by rule). Each persona: - Reads the entry - Applies its own trait model - Produces **annotations**, not conclusions Examples of persona outputs: - recurring language markers - affective tone estimates - epistemic posture - moral framing - signs of avoidance, fixation, or narrative looping Persona outputs are stored as **separate graph nodes** linked to the entry. --- ### Step 4 — Synthesize (Optional) Synthesis agents may be run to: - Compare persona outputs - Detect drift across time - Identify recurring motifs - Correlate language with metrics (sleep, stress, work, etc.) Synthesis never overwrites persona outputs. It only references them. --- ### Step 5 — Read You read the results as **artifacts**: - Markdown reports - Graph views - Diffs across time Nothing is pushed to you automatically. Insight is pull-based. --- ## 4. Feeds and Data Sources All ingestion is defined in `feeds.json`. Feeds may include: - journal folders - markdown blogs - exported chat logs - annotations - external text streams Each feed defines: - source - update behavior - parsing rules - metadata mappings The system assumes: > If it entered through a feed, it is legitimate data. --- ## 5. Personas: How to Use Them Correctly ### 5.1 Choosing Personas You can: - Run a single persona repeatedly over time - Run many personas on the same entry - Introduce new personas mid-stream There is no “correct” persona set. A common pattern: - One analytical persona (structure, logic) - One affective persona (emotion, tone) - One adversarial persona (contradictions, self-deception) --- ### 5.2 Interpreting Persona Output Persona output is **descriptive**, not prescriptive. Bad interpretation: > “This persona says I am X, so I am X.” Correct interpretation: > “From this perspective, these patterns appear.” If a persona feels “wrong,” that is information. Do not delete it. Compare it later. --- ## 6. Metrics & Correlations The system can ingest: - biometric data - work schedules - self-defined scales (1–10 mood, pain, focus, etc.) Metrics are: - optional - user-defined - never interpreted as diagnoses The system may surface correlations, such as: - certain language appearing under fatigue - fixation increasing under stress - abstraction increasing during stability Correlation is not causation. The system will not pretend otherwise. --- ## 7. CLI Usage Philosophy The CLI is the primary interface. Principles: - Explicit commands - No hidden state - Human-readable outputs You should be able to: - re-run analyses - inspect intermediate artifacts - delete derived data without touching raw data If something feels “magical,” it is probably wrong. --- ## 8. Failure Modes (By Design) The system will: - Miss things - Misclassify tone - Surface uncomfortable patterns - Contradict itself across personas These are not bugs. They are the cost of honest reflection. Silence is preferred to false certainty. --- ## 9. Ethical Boundary Conditions The system: - Does not tell you what to do - Does not claim authority - Does not simulate care - Does not replace human judgment Responsibility remains with the user. This is a mirror, not a guide. --- ## 10. Long-Term Use Over time, the system becomes: - A memory external to mood - A record immune to revisionism - A way to see yourself change without narrativizing it away It will not make you better. It may make you more precise. That is the trade. --- ## 11. Final Note If you are tempted to: - simplify it - make it friendlier - add encouragement - hide uncertainty Resist. The value of this system is that it **does not lie to you**, even when the truth is boring, fragmented, or unresolved. ```