@timmeck/brain 3.36.15 → 3.36.16

package/README.md

@@ -5,9 +5,9 @@
 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
 [![GitHub stars](https://img.shields.io/github/stars/timmeck/brain-ecosystem?style=social)](https://github.com/timmeck/brain-ecosystem)
 
-**Autonomous Error Memory, Code Intelligence & Self-Improving AI for Claude Code — 134 MCP Tools, 30+ Engines**
+**Autonomous Error Memory, Code Intelligence & Self-Improving AI for Claude Code — 134 MCP Tools, 60+ Engines**
 
-Brain is an MCP server that gives Claude Code a persistent, self-improving memory. It remembers errors, learns solutions, and runs 30+ autonomous engines in a 40-step feedback cycle. It observes itself, detects anomalies, forms and tests hypotheses, distills principles, reasons in chains, feels emotions, evolves strategies genetically, debates itself, gets curious about knowledge gaps, and modifies its own source code. It dreams — consolidating memories during idle periods like biological sleep. 134 MCP tools. 2496 tests.
+Brain is an MCP server that gives Claude Code a persistent, self-improving memory. It remembers errors, learns solutions, and runs 60+ autonomous engines in a 40-step feedback cycle. It observes itself, detects anomalies, forms and tests hypotheses, distills principles, reasons in chains, feels emotions, evolves strategies genetically, debates itself, challenges its own principles (Advocatus Diaboli), gets curious about knowledge gaps, syncs knowledge via Borg collective, loads community plugins, and modifies its own source code. Multi-provider LLM (Anthropic + Ollama). Autonomous web research missions. Live tech radar scanning. 137 MCP tools. 1407 tests.
 
 ## Quick Start
 
@@ -18,6 +18,29 @@ brain setup
 
 That's it. One command configures MCP, hooks, and starts the daemon.
 
+## Architecture
+
+```
+Claude Code  ──MCP stdio──►  Brain Daemon (:7777)
+Cursor/Windsurf ─MCP SSE──►  MCP HTTP Server (:7778)
+Browser ────────HTTP──────►  Mission Control (:7788)
+                              Command Center  (:7790)
+                                    │
+                    ┌───────────────┼───────────────┐
+                    ▼               ▼               ▼
+              Error Memory    Research Engine   60+ Engines
+              Code Intel      Mission Engine    ResearchOrchestrator
+              Synapse Net     LLM Service       40-step feedback loop
+              Prevention      Web Research          │
+              Git Intel       TechRadar         ┌───┴───────────┐
+                    │                           ▼               ▼
+                    ▼                     Self-Modification  Dream Mode
+                  SQLite (~/.brain)       CodeGenerator      Memory
+                                          SelfScanner        Consolidation
+```
+
+Cross-brain communication via IPC named pipes (trading-brain, marketing-brain).
+
 ## Features
 
 ### Error Memory & Code Intelligence
@@ -32,221 +55,188 @@ That's it. One command configures MCP, hooks, and starts the daemon.
 - **Memory System** — Remember preferences, decisions, context, facts, goals, and lessons across sessions
 - **Session Tracking** — Auto-tracks conversation sessions with goals, summaries, and outcomes
 - **Decision History** — Record architecture/design decisions with alternatives and rationale
-- **Semantic Changelog** — Track what changed, why, and how it connects to errors and decisions
-- **Task/Goal Tracking** — Manage tasks with priorities, subtasks, and full context aggregation
+- **Semantic Changelog** — Tracks every file change with semantic meaning, diffs, and context
+- **Task & Goal Tracking** — Persistent tasks with priority, status, and cross-session continuity
 - **Semantic Search** — Local all-MiniLM-L6-v2 embeddings (23MB, no cloud required)
 
-### 30+ Autonomous Engines
-
-The ResearchOrchestrator runs a 40-step feedback cycle every 5 minutes, coordinating all engines:
-
-#### Core Research (9 Engines)
-
-| Engine | What It Does |
-|--------|-------------|
-| **SelfObserver** | Observes Brain's own performance and generates insights |
-| **AnomalyDetective** | Detects statistical outliers using Z-scores and drift analysis |
-| **CrossDomainEngine** | Finds correlations between events across brains |
-| **AdaptiveStrategy** | Adjusts strategies, reverts if performance drops |
-| **ExperimentEngine** | A/B tests on Brain's own parameters every 5 cycles |
-| **KnowledgeDistiller** | Extracts principles from confirmed hypotheses |
-| **ResearchAgenda** | Prioritizes what to research next |
-| **CounterfactualEngine** | "What if" analysis for hypothetical interventions |
-| **Journal** | Logs all discoveries, experiments, breakthroughs |
-
-#### Intelligence & Awareness
-
-| Engine | What It Does |
-|--------|-------------|
-| **AttentionEngine** | Dynamic focus, context detection, burst detection, engine weight allocation |
-| **TransferEngine** | Cross-brain knowledge transfer, analogies, cross-domain rules |
-| **NarrativeEngine** | Brain explains itself in natural language, finds contradictions, generates digests |
-| **CuriosityEngine** | Knowledge gap detection, UCB1 explore/exploit, blind spot detection |
-| **EmergenceEngine** | Emergent behavior detection, complexity metrics (entropy, phi), phase transitions |
-| **DebateEngine** | Multi-agent debates, advocatus diaboli, consensus synthesis |
-| **ReasoningEngine** | Forward chaining, abductive reasoning, temporal inference, counterfactuals |
-| **EmotionalModel** | 8 emotion dimensions, 6 moods, mood-based behavior recommendations |
-
-#### Meta-Cognition & Optimization
-
-| Engine | What It Does |
-|--------|-------------|
-| **ParameterRegistry** | Central tunable parameter store with 30+ parameters, bounds, snapshots |
-| **MetaCognitionLayer** | Engine performance grading (A-F), frequency adjustment |
-| **AutoExperimentEngine** | Autonomous parameter tuning with snapshot/rollback |
-| **EvolutionEngine** | Genetic algorithm — tournament selection, crossover, mutation, elitism |
-| **GoalEngine** | Goal planning with progress tracking, linear regression forecasting |
-| **MemoryPalace** | Knowledge graph with BFS pathfinding and auto-connection building |
-
-#### Autonomy & Self-Improvement
-
-| Engine | What It Does |
-|--------|-------------|
-| **SelfTestEngine** | Tests if brain truly understands its principles vs memorization |
-| **TeachEngine** | Generates teaching packages for other brains |
-| **DataScout** | External data acquisition from GitHub/npm/HN |
-| **SimulationEngine** | What-if scenarios via CausalGraph + PredictionEngine |
-| **SelfScanner** | Indexes own TypeScript source code with SHA256 change detection |
-| **SelfModificationEngine** | Generates and tests code changes autonomously via Claude API |
-| **BootstrapService** | Cold-start fix: seeds data so engines produce output from cycle 1 |
+### LLM Service
+- **Multi-Provider** — Anthropic Claude + Ollama local models with auto-routing
+- **Smart Caching** — Content-hash cache, avoid duplicate API calls
+- **Rate Limiting** — Per-hour and per-day token budgets with automatic throttling
+- **Usage Tracking** — Calls, tokens, latency, cache hit rate, cost tracking
+- **Tier Routing** — Templates mapped to tiers (critical/standard/bulk), auto-routed to best provider
+
+### Research Missions
+- **5-Phase Pipeline** — Decompose → Gather → Hypothesize → Analyze → Synthesize
+- **Web Research** — Brave Search + Jina Reader + Playwright + Firecrawl fallback chain
+- **Autonomous** — Brain decides what to research and executes independently
+- **Source Tracking** — Every finding traced back to its original source
+
+### TechRadar
+- **Daily Scanning** — Tracks trending repos, tech news, library updates
+- **Repo Watching** — Monitor specific repos for changes (3 default repos configured)
+- **LLM Relevance Scoring** — AI judges how relevant each finding is to your stack
+- **Digest Generation** — Daily summaries of what's new and relevant
+
+### 60+ Autonomous Engines
+
+The ResearchOrchestrator runs a 40-step feedback cycle every 5 minutes:
+
+#### Core Research Engines
+
+| Engine | Purpose |
+|--------|---------|
+| SelfObserver | Monitors Brain's own performance metrics and behavior |
+| AnomalyDetective | Detects statistical anomalies in error patterns and metrics |
+| DataScout | Discovers external data sources and imports relevant data |
+| SignalScanner | Scans GitHub repos, HN, crypto markets for signals |
+| TechRadar | Daily tech landscape scanning with relevance scoring |
+| HypothesisEngine | Generates and statistically tests hypotheses about patterns |
+| ExperimentEngine | Proposes, runs, and measures controlled experiments |
+| AutoExperimentEngine | Autonomously discovers and runs parameter experiments |
+| SimulationEngine | What-if scenarios and counterfactual reasoning |
+| KnowledgeDistiller | Extracts principles and anti-patterns from experience |
+
+#### Intelligence Engines
+
+| Engine | Purpose |
+|--------|---------|
+| AttentionEngine | Dynamic focus allocation across topics and engines |
+| CausalGraph | Discovers cause-effect relationships between events |
+| CrossDomainEngine | Finds correlations across brain domains |
+| PatternExtractor | Mines recurring code and error patterns |
+| TransferEngine | Transfers knowledge between domains via analogies |
+| NarrativeEngine | Generates natural language explanations of findings |
+| CuriosityEngine | Detects knowledge gaps and generates exploration questions |
+| ResearchAgendaEngine | Prioritizes what to investigate next |
+| CounterfactualEngine | "What if X hadn't happened?" reasoning |
+
+#### Meta-Cognition Engines
+
+| Engine | Purpose |
+|--------|---------|
+| MetaCognitionLayer | Evaluates engine effectiveness, produces report cards |
+| DebateEngine | Multi-perspective reasoning with synthesis |
+| EmergenceEngine | Detects emergent properties from engine interactions |
+| ConceptAbstraction | Forms hierarchical concept taxonomies |
+| MemoryPalace | Builds associative knowledge graph for navigation |
+| ReasoningEngine | Deductive, abductive, and temporal inference chains |
+| EmotionalModel | Frustration, curiosity, satisfaction — influences priorities |
+| SelfTestEngine | Tests understanding of its own principles |
+| TeachEngine | Packages knowledge for transfer to other brains |
+
+#### Autonomy Engines
+
+| Engine | Purpose |
+|--------|---------|
+| SelfModificationEngine | Generates code improvements, tests before applying |
+| CodeGenerator | Produces new code from learned patterns |
+| CodeMiner | Extracts reusable patterns from codebases |
+| GoalEngine | Sets, tracks, and forecasts autonomous goals |
+| EvolutionEngine | Genetic algorithm for strategy optimization |
+| AdaptiveStrategyEngine | Real-time parameter adaptation based on outcomes |
+| DreamEngine | Offline memory consolidation during idle |
+| ResearchOrchestrator | Orchestrates the entire 40-step feedback cycle |
 
 ### Self-Improvement Loop
-- **Hypothesis Engine** — Forms hypotheses from observed patterns, tests them statistically
-- **Auto-Experiments** — Tests its own parameters (Z-Threshold, Decay Rate, Research Interval)
-- **Self-Metrics** — Tracks own cycle data: anomaly_count, insight_count, cycle_duration_ms
-- **Frustration Detection** — If a suggestion is ignored 3x, generates alternative approaches
-- **Improvement Requests** — Writes "Tell Claude:" suggestions to `~/.brain/improvement-requests.md`
-- **Knowledge Distillation** — Confirmed hypotheses become reusable principles
 
-### Dream Mode
-- **Memory Replay** — Spreading Activation strengthens active memory pathways
-- **Synapse Pruning** — Weak connections (weight < 0.15) are deleted
-- **Memory Compression** — Similar memories are clustered and merged
-- **Importance Decay** — Old memories gradually lose importance, eventually archived
-- **Triggers** — Every 20 cycles, after 5 minutes idle, or manually
-
-### Mission Control Dashboard (`:7788`)
-
-<p align="center"><img src="../../docs/assets/mission-control.png" alt="Mission Control Dashboard" width="700"></p>
+Brain continuously improves itself through a closed-loop cycle:
 
-- **7 Tabs** — Overview, Consciousness, Thoughts, CodeGen, Self-Mod, Engines, Intelligence
-- **Consciousness Entity** — Living animated visualization of the brain's emotional/consciousness state: pulsing orb colored by mood (flow=cyan, excited=gold, anxious=red, reflective=violet, bored=gray, determined=white), organic tentacles, 8-dimension ring, ambient particles, floating thoughts, inner eye tracking attention focus
-- **Thought Stream** — Real-time feed of every engine's activity with significance filtering
-- **CodeGen** — Generate, review, approve/reject code with syntax highlighting
-- **Self-Mod** — Pending modifications with side-by-side diff viewer
-- **Engine Status** — Live status cards for all 30+ engines
-- **Intelligence** — Attention topics, transfer analogies, cross-domain rules
-- **Trigger Cycle** — Button to manually trigger a feedback cycle
-
-### Prediction Engine
-- **Holt-Winters** — Double Exponential Smoothing for metric forecasting
-- **EWMA Fallback** — When insufficient data for Holt-Winters
-- **Auto-Calibration** — Confidence calibrated against actual outcomes
-- **Self-Resolving** — Predictions are checked against reality and scored
+1. **Observe** — SelfObserver records performance metrics (error rates, resolution times, cache hits)
+2. **Hypothesize** — HypothesisEngine generates testable theories about what could work better
+3. **Experiment** — AutoExperimentEngine runs controlled A/B tests on parameters
+4. **Measure** — MetaCognitionLayer evaluates which experiments improved outcomes
+5. **Adapt** — AdaptiveStrategy applies winning parameters, reverts failures
+6. **Evolve** — EvolutionEngine genetically breeds the best strategy combinations
 
-### AutoResponder
-- **Anomaly Response** — Automatically adjusts parameters when anomalies are detected
-- **Configurable Rules** — Define which anomalies trigger which actions
-- **Cooldown System** — Prevents response storms
-- **Journal Logging** — Every auto-response is logged for review
-
-### Code Generation & Mining
-- **CodeGenerator** — Claude API integration, uses brain knowledge (principles, anti-patterns, strategies, patterns) as context
-- **Human-in-the-Loop** — Generated code is reviewed: approve or reject with notes
-- **CodeMiner** — Mines GitHub repos (README, package.json, directory trees) for pattern learning
-- **PatternExtractor** — Extracts dependency, tech stack, structure, and README patterns from mined repos
-- **CodeGen Tab** — Integrated in Mission Control at http://localhost:7788 for reviewing and managing code generations
-- **Signal Scanner** — Tracks GitHub trending repos, Hacker News mentions, crypto signals
-
-### Dashboards
-
-| Dashboard | Port | What It Shows |
-|-----------|------|--------------|
-| **Mission Control** | 7788 | Unified 7-tab dashboard: Overview, Consciousness (Entity visualization), Thoughts, CodeGen, Self-Mod, Engines, Intelligence |
-
-## Architecture
+Frustration Detection: EmotionalModel tracks repeated failures. High frustration triggers more aggressive exploration via CuriosityEngine.
 
-```
-Claude Code / Cursor / Browser
-         |
-    MCP / HTTP / REST
-         |
-    +----+----+
-    | BrainCore|
-    +----+----+
-         |
-  +------+------+------+------+------+
-  |      |      |      |      |      |
-Error  Code  Synapse Memory  Git  Embedding
-Memory Intel Network System Intel Engine
-  |      |      |      |      |      |
-  +------+------+------+------+------+
-         |
-    ResearchOrchestrator (40 steps)
-         |
-  +------+------+------+------+------+------+------+
-  |      |      |      |      |      |      |      |
-Dream  Pred.  AutoR  Hypo.  CodeGen Signal Attention
-Engine Engine espond Engine (Claude) Scanner Engine
-  |      |      |      |      |      |      |      |
-  +------+------+------+------+------+------+------+
-         |
-  30+ Engines in Feedback Loops
-  (Research, Intelligence, MetaCog, Evolution,
-   Reasoning, Emotions, Self-Modification, ...)
-         |
-      SQLite (better-sqlite3, WAL mode)
-```
-
-## MCP Tools (134 tools)
-
-**Error & Code**: brain_report_error, brain_query_error, brain_report_solution, brain_report_attempt, brain_find_reusable_code, brain_register_code, brain_check_code_similarity
-
-**Memory & Sessions**: brain_remember, brain_recall, brain_session_start, brain_session_end, brain_session_history
+### Dream Mode
 
-**Decisions, Changes, Tasks, Docs**: brain_record_decision, brain_query_decisions, brain_record_change, brain_query_changes, brain_add_task, brain_update_task, brain_list_tasks, brain_task_context, brain_index_project, brain_query_docs, brain_project_context
+During idle periods (no active conversations), DreamEngine performs memory consolidation:
 
-**Research & Learning**: brain_explore, brain_connections, brain_insights, brain_rate_insight, brain_suggest, brain_explain_learning, brain_override_rule, brain_get_suggestions
+- **Replay** — Re-processes important experiences to strengthen synaptic connections
+- **Pruning** — Removes low-value memories and weak synapse connections
+- **Compression** — Merges similar patterns into generalized principles
+- **Decay** — Time-based weight reduction on unused knowledge
+- **Triggers** — Starts automatically after configurable idle period, or manually via `dream.start`
 
-**9 Research Engines** (5 tools each): self_observer, anomaly_detective, cross_domain, adaptive_strategy, experiment, knowledge_distiller, research_agenda, counterfactual, journal — each with _status, _list, _get, _summary, _config
+### Prediction Engine
 
-**Dream Mode**: brain_dream_status, brain_dream_consolidate, brain_dream_history
+Forecasts future metrics using statistical models:
 
-**Consciousness**: brain_consciousness_status, brain_consciousness_thoughts
+- **Holt-Winters** — Triple exponential smoothing for seasonal patterns
+- **EWMA** — Exponential weighted moving average for trend detection
+- **Auto-Calibration** — Tracks prediction accuracy and adjusts model parameters
+- **Domain-Aware** — Separate models per domain (errors, performance, learning)
 
-**Prediction**: brain_predict, brain_prediction_accuracy, brain_predictions_list
+### AutoResponder
 
-**AutoResponder**: brain_responder_status, brain_responder_history, brain_responder_rules
+Automatic anomaly response system:
 
-**CodeGen**: brain_generate_code, brain_codegen_status, brain_codegen_review, brain_codeminer_status, brain_codeminer_patterns
+- **Rule-Based** — Configurable rules: "if error_rate > threshold, trigger learning cycle"
+- **Cooldown** — Prevents response storms with per-rule cooldown periods
+- **Action Types** — Learning cycles, notifications, parameter adjustments, dream triggers
+- **History** — Full audit trail of what was detected and what action was taken
 
-**Attention**: brain_focus_status, brain_focus_set, brain_focus_history
+### Code Generation & Mining
 
-**Transfer**: brain_transfer_status, brain_transfer_analogies, brain_transfer_rules
+- **CodeGenerator** — Generates code improvements via Claude API with full diff preview
+- **CodeMiner** — Analyzes codebases to extract reusable patterns and modules
+- **PatternExtractor** — Identifies recurring code patterns across projects
+- **SignalScanner** — Monitors GitHub trending, Hacker News, crypto markets for relevant signals
+- **Self-Improvement Proposals** — Engines can propose improvements to their own source code
 
-**Narrative**: brain_explain, brain_ask, brain_weekly_digest, brain_contradictions
+### Self-Modification
+- **SelfScanner** — Indexes own TypeScript source code with SHA256 change detection
+- **SelfModificationEngine** — Generates improvements via Claude API, tests before applying
+- **Experiment Ledger** — Tracks hypothesis, risk level, metrics before/after for every modification
+- **Safety** — All modifications require explicit approval, automatic rollback on test failure
 
-**Curiosity**: brain_curiosity_status, brain_curiosity_gaps, brain_curiosity_questions, brain_curiosity_explore
+### Notifications
+- **Discord, Telegram, Email** — Multi-channel alert routing
+- **Notification Bridge** — IPC-based cross-brain notification relay
+- **Configurable** — All providers optional, graceful fallback
+- **Event Routing** — Different events route to different channels
 
-**Emergence**: brain_emergence_status, brain_emergence_detect, brain_emergence_complexity_metrics, brain_emergence_journal
+### Dashboards
 
-**Debate**: brain_debate_start, brain_debate_synthesize, brain_debate_perspective, brain_debate_history
+| Dashboard | Port | What It Shows |
+|-----------|------|--------------|
+| **Mission Control** | 7788 | 7-tab: Overview, Consciousness Entity, Thoughts, CodeGen, Self-Mod, Engines, Intelligence |
+| **Command Center** | 7790 | 8-page: Ecosystem, Learning, Trading, Marketing, Cross-Brain & Borg, Debates & Challenges, Activity & Missions, Infrastructure |
 
-**MetaCognition**: brain_metacognition_status, brain_engine_report, brain_auto_experiment_status, brain_parameter_registry
+**Mission Control** — The Consciousness Entity visualization shows Brain's current emotional state, active thought streams, engine activity heatmap, and real-time thought generation. CodeGen tab shows pending code proposals. Self-Mod tab shows modification history with diffs.
 
-**SelfAware**: brain_selftest_run, brain_selftest_results, brain_teach_status, brain_teach_create, brain_datascout_status, brain_datascout_scan, brain_simulation_run, brain_simulation_results, brain_simulation_scenarios, brain_palace_status, brain_palace_map, brain_palace_path, brain_palace_build, brain_goal_status, brain_goal_create, brain_goal_progress
+**Command Center** — Live overview of the entire ecosystem: all 3 brains, 60+ engines, error log, self-modification feed, research missions, knowledge growth chart, engine dependency flow, quick actions, animated Borg network, peer graph, debate history, Advocatus Diaboli challenges with resilience bars, watchdog daemon monitoring, LLM usage tracking.
 
-**Evolution**: brain_evolution_status, brain_evolution_history, brain_evolution_best, brain_evolution_run
+## MCP Tools (137 tools)
 
-**Reasoning**: brain_reasoning_status, brain_reason, brain_explain_why, brain_what_if
+**Error & Code**: brain_report_error, brain_query_error, brain_report_solution, brain_report_attempt, brain_find_reusable_code, brain_register_code, brain_check_code_similarity
 
-**Emotions**: brain_emotional_status, brain_mood_history, brain_mood_influences, brain_mood_advice
+**Memory & Sessions**: brain_remember, brain_recall, brain_session_start, brain_session_end, brain_session_history
 
-**Self-Modification**: brain_selfmod_status, brain_selfmod_pending, brain_selfmod_approve, brain_selfmod_history
+**Research Engines** (5 tools each): self_observer, anomaly_detective, cross_domain, adaptive_strategy, experiment, knowledge_distiller, research_agenda, counterfactual, journal
 
-**Ecosystem**: brain_status, brain_notifications, brain_ecosystem_status, brain_query_peer, brain_error_trading_context, brain_unified_status
+**Dream, Consciousness, Prediction, AutoResponder, Attention, Transfer, Narrative, Curiosity, Emergence, Debate, Challenge (Advocatus Diaboli), MetaCognition, Evolution, Reasoning, Emotions, Self-Modification, Ecosystem, Borg, Plugins** — full tool suites for each
 
 ## CLI Commands
 
 ```
 brain setup              One-command setup: MCP + hooks + daemon
-brain start              Start the Brain daemon
-brain stop               Stop the daemon
-brain status             Show stats
+brain start / stop       Daemon management (with watchdog)
+brain status             Stats: errors, solutions, engines, synapses
 brain doctor             Health check: daemon, DB, MCP, hooks
 brain query <text>       Search for errors and solutions
-brain modules            List registered code modules
-brain insights           Show research insights
-brain network            Explore the synapse network
 brain learn              Trigger a learning cycle
-brain explain <id>       Full error report
-brain rules              List active learned rules
-brain synapses           Show strongest synapse connections
-brain config             View and manage configuration
-brain export             Export Brain data as JSON
-brain import <dir>       Import a project directory
-brain dashboard          Generate interactive HTML dashboard
 brain peers              Show peer brains in the ecosystem
+brain dashboard          Generate interactive HTML dashboard
+brain missions           Research mission management (create, list, report)
+brain watchdog           Watchdog daemon status and control
+brain service            Windows service management (install, uninstall, status)
+brain borg               Borg collective sync (status, enable, disable, sync, history)
+brain plugins            Community plugins (list, routes, tools)
+brain export             Export Brain data as JSON
 ```
 
 ## Configuration
@@ -256,20 +246,20 @@ brain peers              Show peer brains in the ecosystem
 | `BRAIN_DATA_DIR` | `~/.brain` | Data directory |
 | `BRAIN_LOG_LEVEL` | `info` | Log level |
 | `BRAIN_API_PORT` | `7777` | REST API port |
-| `BRAIN_API_KEY` | — | API authentication key |
 | `BRAIN_MCP_HTTP_PORT` | `7778` | MCP HTTP/SSE port |
-| `BRAIN_EMBEDDINGS_ENABLED` | `true` | Enable local embeddings |
-| `ANTHROPIC_API_KEY` | — | Enables CodeGenerator + CodeGen Dashboard |
+| `ANTHROPIC_API_KEY` | — | Enables LLM features, CodeGen, Self-Mod |
+| `BRAVE_SEARCH_API_KEY` | — | Enables web research missions |
 | `GITHUB_TOKEN` | — | Enables CodeMiner + Signal Scanner |
+| `OLLAMA_HOST` | `http://localhost:11434` | Ollama local model endpoint |
 
 ## Brain Ecosystem
 
-| Brain | Version | Purpose | Ports |
-|-------|---------|---------|-------|
-| **Brain** | v3.34.0 | Error memory, code intelligence, full autonomy & self-modification | **7777** / 7778 / 7788 |
-| [Trading Brain](../trading-brain) | v2.29.0 | Adaptive trading intelligence with signal learning & backtesting | 7779 / 7780 |
-| [Marketing Brain](../marketing-brain) | v1.30.0 | Content strategy, engagement & cross-platform optimization | 7781 / 7782 / 7783 |
-| [Brain Core](../brain-core) | v2.34.0 | Shared infrastructure — 30+ engines, synapses, IPC, MCP, dream, consciousness, codegen | — |
+| Brain | Purpose | Ports |
+|-------|---------|-------|
+| **Brain** (this) | Error memory, code intelligence, full autonomy & self-modification | **7777** / 7778 / 7788 / 7790 |
+| [Trading Brain](../trading-brain) | Adaptive trading intelligence with signal learning & paper trading | 7779 / 7780 |
+| [Marketing Brain](../marketing-brain) | Content strategy, social engagement & cross-platform optimization | 7781 / 7782 / 7783 |
+| [Brain Core](../brain-core) | Shared infrastructure — 60+ engines | — |
 
 ## Support
 

package/dist/brain.d.ts

@@ -22,7 +22,10 @@ export declare class BrainCore {
     private emergenceEngine;
     private debateEngine;
     private peerNetwork;
+    private pluginRegistry;
+    private borgSync;
     private cleanupTimer;
+    private retentionTimer;
     private config;
     private configPath?;
     private restarting;
@@ -30,6 +33,7 @@ export declare class BrainCore {
     private restartWindowStart;
     start(configPath?: string): void;
     private logCrash;
+    private runRetentionCleanup;
     private cleanup;
     restart(): void;
     stop(): void;