agentic-flow 1.7.2 → 1.7.4

package/dist/reasoningbank/index.js

@@ -1,38 +1,38 @@
 /**
- * ReasoningBank Plugin for Agentic-Flow
+ * ReasoningBank - Closed-loop memory system for AI agents
+ * Based on arXiv:2509.25140 (Google DeepMind)
  *
- * Main entry point and public API
- *
- * Paper: https://arxiv.org/html/2509.25140v1
- *
- * This is the Node.js backend using SQLite for persistent storage.
- * For browser environments, use './wasm-adapter.js' instead.
- * For automatic backend selection, use './backend-selector.js'.
+ * @since v1.7.0 - Integrated AgentDB for optimal performance
  */
-// Core algorithms
+// New hybrid backend (recommended for new code)
+export { HybridReasoningBank } from './HybridBackend.js';
+export { AdvancedMemorySystem } from './AdvancedMemory.js';
+// Re-export AgentDB controllers for advanced usage
+export { ReflexionMemory } from 'agentdb/controllers/ReflexionMemory';
+export { SkillLibrary } from 'agentdb/controllers/SkillLibrary';
+export { CausalMemoryGraph } from 'agentdb/controllers/CausalMemoryGraph';
+export { CausalRecall } from 'agentdb/controllers/CausalRecall';
+export { NightlyLearner } from 'agentdb/controllers/NightlyLearner';
+export { EmbeddingService } from 'agentdb/controllers/EmbeddingService';
+// Original ReasoningBank implementations (backwards compatibility)
 export { retrieveMemories, formatMemoriesForPrompt } from './core/retrieve.js';
 export { judgeTrajectory } from './core/judge.js';
 export { distillMemories } from './core/distill.js';
 export { consolidate, shouldConsolidate } from './core/consolidate.js';
 export { mattsParallel, mattsSequential } from './core/matts.js';
-// Utilities
 export { computeEmbedding, clearEmbeddingCache } from './utils/embeddings.js';
 export { mmrSelection, cosineSimilarity } from './utils/mmr.js';
 export { scrubPII, containsPII, scrubMemory } from './utils/pii-scrubber.js';
 export { loadConfig } from './utils/config.js';
-// Import for internal use
-import { loadConfig } from './utils/config.js';
+// Re-export database utilities
 import * as db from './db/queries.js';
+export { db };
+// Original functions (backwards compatibility)
+import { loadConfig } from './utils/config.js';
 import { retrieveMemories } from './core/retrieve.js';
 import { judgeTrajectory } from './core/judge.js';
 import { distillMemories } from './core/distill.js';
-import { shouldConsolidate, consolidate } from './core/consolidate.js';
-// Database
-export { db };
-/**
- * Initialize ReasoningBank
- * Run migrations and check configuration
- */
+import { shouldConsolidate as shouldCons, consolidate as cons } from './core/consolidate.js';
 export async function initialize() {
     const config = loadConfig();
     console.log('[ReasoningBank] Initializing...');
@@ -40,7 +40,6 @@ export async function initialize() {
     console.log(`[ReasoningBank] Database: ${process.env.CLAUDE_FLOW_DB_PATH || '.swarm/memory.db'}`);
     console.log(`[ReasoningBank] Embeddings: ${config.embeddings.provider}`);
     console.log(`[ReasoningBank] Retrieval k: ${config.retrieve.k}`);
-    // Run migrations to create database and tables
     try {
         await db.runMigrations();
         console.log(`[ReasoningBank] Database migrated successfully`);
@@ -49,7 +48,6 @@ export async function initialize() {
         console.error('[ReasoningBank] Migration error:', error);
         throw new Error('ReasoningBank initialization failed: could not run migrations');
     }
-    // Check database connection
     try {
         const dbConn = db.getDb();
         const tables = dbConn.prepare("SELECT name FROM sqlite_master WHERE type='table' AND name LIKE 'pattern%'").all();
@@ -61,48 +59,29 @@ export async function initialize() {
     }
     console.log('[ReasoningBank] Initialization complete');
 }
-/**
- * Full task execution with ReasoningBank
- * Combines retrieve → execute → judge → distill → consolidate
- */
 export async function runTask(options) {
     console.log(`[ReasoningBank] Running task: ${options.taskId}`);
-    // 1. Retrieve memories
     const memories = await retrieveMemories(options.query, {
         domain: options.domain,
         agent: options.agentId
     });
     console.log(`[ReasoningBank] Retrieved ${memories.length} memories`);
-    // 2. Execute task with memories
     const trajectory = await options.executeFn(memories);
-    // 3. Judge trajectory
     const verdict = await judgeTrajectory(trajectory, options.query);
     console.log(`[ReasoningBank] Verdict: ${verdict.label} (${verdict.confidence})`);
-    // 4. Distill new memories
     const newMemories = await distillMemories(trajectory, verdict, options.query, {
         taskId: options.taskId,
         agentId: options.agentId,
         domain: options.domain
     });
     console.log(`[ReasoningBank] Distilled ${newMemories.length} new memories`);
-    // 5. Consolidate if needed
     let consolidated = false;
-    if (shouldConsolidate()) {
+    if (shouldCons()) {
         console.log('[ReasoningBank] Running consolidation...');
-        await consolidate();
+        await cons();
         consolidated = true;
     }
-    return {
-        verdict,
-        usedMemories: memories,
-        newMemories,
-        consolidated
-    };
+    return { verdict, usedMemories: memories, newMemories, consolidated };
 }
-// AgentDB Integration
-// Ultra-fast vector database drop-in replacement
-// 150x-12,500x faster than legacy implementation
-export { createAgentDBAdapter, createDefaultAgentDBAdapter, migrateToAgentDB, validateMigration } from './agentdb-adapter.js';
-// Version info
-export const VERSION = '1.0.0';
+export const VERSION = '1.7.1';
 export const PAPER_URL = 'https://arxiv.org/html/2509.25140v1';

package/dist/utils/cli.js

@@ -49,16 +49,6 @@ export function parseArgs() {
         options.mode = 'reasoningbank';
         return options;
     }
-    // Check for agentdb command
-    if (args[0] === 'agentdb') {
-        options.mode = 'agentdb';
-        return options;
-    }
-    // Check for skills command
-    if (args[0] === 'skills') {
-        options.mode = 'skills';
-        return options;
-    }
     for (let i = 0; i < args.length; i++) {
         const arg = args[i];
         switch (arg) {
@@ -165,7 +155,6 @@ USAGE:
   npx agentic-flow [COMMAND] [OPTIONS]
 
 COMMANDS:
-  skills <command>        Claude Code Skills management (init, create, list)
   reasoningbank <cmd>     Memory system that learns from experience (demo, test, init)
   claude-code [options]   Spawn Claude Code with proxy + Agent Booster (57x faster edits)
   mcp <command> [server]  Manage MCP servers (start, stop, status, list)
@@ -182,12 +171,6 @@ REASONINGBANK COMMANDS:
   npx agentic-flow reasoningbank benchmark    Run performance benchmarks
   npx agentic-flow reasoningbank status       Show memory statistics
 
-SKILLS COMMANDS:
-  npx agentic-flow skills init [location]   Initialize skills directories (personal/project/both)
-  npx agentic-flow skills create            Create example agentic-flow skills
-  npx agentic-flow skills list              List all installed skills
-  npx agentic-flow skills help              Show skills help
-
 MCP COMMANDS:
   npx agentic-flow mcp start [server]    Start MCP server(s)
   npx agentic-flow mcp stop [server]     Stop MCP server(s)
@@ -233,11 +216,6 @@ OPTIONS:
   --help, -h                  Show this help message
 
 EXAMPLES:
-  # Skills (Claude Code integration!)
-  npx agentic-flow skills init              # Initialize skills directories
-  npx agentic-flow skills create            # Create agentdb-quickstart skill
-  npx agentic-flow skills list              # List all installed skills
-
   # ReasoningBank (Learn from agent experience!)
   npx agentic-flow reasoningbank demo       # See 0% → 100% success transformation
   npx agentic-flow reasoningbank test       # Run 27 validation tests

package/docs/AGENTDB_TESTING.md

@@ -0,0 +1,411 @@
+# AgentDB CLI - Local Testing Guide
+
+## Quick Start
+
+### Option 1: Run the Full Test Suite (Recommended)
+```bash
+# From the project root: /workspaces/agentic-flow/agentic-flow
+./scripts/test-agentdb.sh
+```
+
+This will run 13 comprehensive tests covering all CLI features!
+
+### Option 2: Manual Testing
+
+#### 1. Ensure you're in the correct directory
+```bash
+cd /workspaces/agentic-flow/agentic-flow
+pwd  # Should show: /workspaces/agentic-flow/agentic-flow
+```
+
+#### 2. Build the Project (if not already built)
+```bash
+npm run build
+```
+
+#### 3. Test the CLI Directly
+```bash
+# Show help (all 17 commands)
+node dist/agentdb/cli/agentdb-cli.js --help
+
+# Or use npx (if globally installed)
+npx agentdb --help
+```
+
+### 3. Create a Test Database
+```bash
+# Set database path
+export AGENTDB_PATH=./test-agentdb.db
+
+# Or specify inline for each command
+AGENTDB_PATH=./test-agentdb.db node dist/agentdb/cli/agentdb-cli.js db stats
+```
+
+## Test Each Command Category
+
+### 🧠 Reflexion Memory (Episodic Replay)
+```bash
+# Store an episode with self-critique
+node dist/agentdb/cli/agentdb-cli.js reflexion store \
+  "session-1" \
+  "implement_authentication" \
+  0.95 \
+  true \
+  "Successfully used OAuth2 with JWT tokens" \
+  "User login requirement" \
+  "Working auth system" \
+  1200 \
+  5000
+
+# Store a failed episode
+node dist/agentdb/cli/agentdb-cli.js reflexion store \
+  "session-1" \
+  "implement_authentication" \
+  0.3 \
+  false \
+  "Forgot to validate tokens properly" \
+  "User login requirement" \
+  "Insecure auth" \
+  800 \
+  3000
+
+# Retrieve relevant episodes
+node dist/agentdb/cli/agentdb-cli.js reflexion retrieve \
+  "authentication" \
+  5 \
+  0.5
+
+# Get critique summary from failures
+node dist/agentdb/cli/agentdb-cli.js reflexion critique-summary \
+  "authentication" \
+  true
+
+# Prune old episodes
+node dist/agentdb/cli/agentdb-cli.js reflexion prune 30 0.2
+```
+
+### 🛠️ Skill Library (Lifelong Learning)
+```bash
+# Create a skill manually
+node dist/agentdb/cli/agentdb-cli.js skill create \
+  "jwt_authentication" \
+  "Generate and validate JWT tokens for user authentication" \
+  "function generateJWT(payload) { return jwt.sign(payload, secret); }"
+
+# Search for skills
+node dist/agentdb/cli/agentdb-cli.js skill search \
+  "authentication tokens" \
+  5
+
+# Auto-consolidate episodes into skills
+node dist/agentdb/cli/agentdb-cli.js skill consolidate \
+  3 \
+  0.7 \
+  7
+
+# Prune underperforming skills
+node dist/agentdb/cli/agentdb-cli.js skill prune \
+  3 \
+  0.4 \
+  60
+```
+
+### 🔗 Causal Memory Graph (Intervention-Based)
+```bash
+# Add a causal edge manually
+node dist/agentdb/cli/agentdb-cli.js causal add-edge \
+  "add_unit_tests" \
+  "code_quality_score" \
+  0.25 \
+  0.95 \
+  100
+
+# Create an A/B experiment
+node dist/agentdb/cli/agentdb-cli.js causal experiment create \
+  "test-coverage-vs-bugs" \
+  "test_coverage" \
+  "bug_rate"
+
+# Add observations (treatment group)
+node dist/agentdb/cli/agentdb-cli.js causal experiment add-observation \
+  1 \
+  true \
+  0.15 \
+  '{"coverage": 0.85}'
+
+# Add observations (control group)
+node dist/agentdb/cli/agentdb-cli.js causal experiment add-observation \
+  1 \
+  false \
+  0.35 \
+  '{"coverage": 0.45}'
+
+# Calculate uplift
+node dist/agentdb/cli/agentdb-cli.js causal experiment calculate 1
+
+# Query causal edges
+node dist/agentdb/cli/agentdb-cli.js causal query \
+  "test" \
+  "quality" \
+  0.6 \
+  0.1 \
+  10
+```
+
+### 🔍 Causal Recall (Utility-Based Reranking)
+```bash
+# Retrieve with causal utility and provenance certificate
+node dist/agentdb/cli/agentdb-cli.js recall with-certificate \
+  "implement secure authentication" \
+  10 \
+  0.7 \
+  0.2 \
+  0.1
+```
+
+### 🌙 Nightly Learner (Automated Discovery)
+```bash
+# Discover causal edges from patterns (dry run)
+node dist/agentdb/cli/agentdb-cli.js learner run \
+  3 \
+  0.6 \
+  0.7 \
+  true
+
+# Discover and save edges
+node dist/agentdb/cli/agentdb-cli.js learner run \
+  3 \
+  0.6 \
+  0.7 \
+  false
+
+# Prune low-quality edges
+node dist/agentdb/cli/agentdb-cli.js learner prune \
+  0.5 \
+  0.05 \
+  90
+```
+
+### 📊 Database Stats
+```bash
+# Get comprehensive database statistics
+node dist/agentdb/cli/agentdb-cli.js db stats
+```
+
+## Full Workflow Example
+
+### Scenario: Learning from Authentication Implementation
+
+```bash
+#!/bin/bash
+
+# Set up test database
+export AGENTDB_PATH=./auth-learning.db
+
+# 1. Store successful attempts
+node dist/agentdb/cli/agentdb-cli.js reflexion store \
+  "session-auth-1" "oauth2_implementation" 0.95 true \
+  "Used industry-standard OAuth2 flow" \
+  "Implement secure login" "Working OAuth2 system" 1500 6000
+
+node dist/agentdb/cli/agentdb-cli.js reflexion store \
+  "session-auth-2" "jwt_tokens" 0.90 true \
+  "JWT with proper expiration and refresh tokens" \
+  "Token management" "Secure JWT system" 1200 5500
+
+# 2. Store failed attempts
+node dist/agentdb/cli/agentdb-cli.js reflexion store \
+  "session-auth-3" "session_storage" 0.35 false \
+  "Insecure session storage in localStorage" \
+  "Session management" "Security vulnerability" 800 3000
+
+# 3. Create a skill from successful pattern
+node dist/agentdb/cli/agentdb-cli.js skill create \
+  "secure_oauth2_jwt" \
+  "OAuth2 flow with JWT token management" \
+  "const auth = { oauth2: true, jwt: true, refresh: true }"
+
+# 4. Add causal edge
+node dist/agentdb/cli/agentdb-cli.js causal add-edge \
+  "add_token_refresh" "session_security" 0.40 0.92 50
+
+# 5. Query for authentication guidance
+node dist/agentdb/cli/agentdb-cli.js reflexion retrieve \
+  "secure authentication" 5 0.8
+
+# 6. Get critique summary of what NOT to do
+node dist/agentdb/cli/agentdb-cli.js reflexion critique-summary \
+  "authentication" true
+
+# 7. Search for applicable skills
+node dist/agentdb/cli/agentdb-cli.js skill search \
+  "oauth jwt tokens" 3
+
+# 8. Check database stats
+node dist/agentdb/cli/agentdb-cli.js db stats
+```
+
+## Verify Installation
+
+### Check Binary Links
+```bash
+# Should show the CLI binary path
+which agentdb
+
+# Or check npm bin
+npm bin agentdb
+```
+
+### Run from Package
+```bash
+# If you've run npm install -g or npm link
+agentdb --help
+```
+
+## Environment Variables
+
+```bash
+# Database path (default: ./agentdb.db)
+export AGENTDB_PATH=/path/to/your/database.db
+
+# Example with custom path
+AGENTDB_PATH=~/my-agent-memory.db node dist/agentdb/cli/agentdb-cli.js db stats
+```
+
+## Expected Output Examples
+
+### Successful Episode Storage
+```
+✓ Stored episode #1 (session: session-1, task: implement_authentication)
+  Reward: 0.95 | Success: true | Latency: 1200ms | Tokens: 5000
+```
+
+### Skill Search Results
+```
+🔍 Found 3 skills for "authentication"
+
+#1: jwt_authentication (success rate: 0.90, uses: 5)
+    Generate and validate JWT tokens for user authentication
+
+#2: oauth2_flow (success rate: 0.85, uses: 3)
+    Complete OAuth2 authorization code flow
+```
+
+### Database Stats
+```
+AgentDB Statistics
+
+Episodes:          15
+Skills:            8
+Causal Edges:      12
+Experiments:       3
+Certificates:      5
+
+Total Size:        2.4 MB
+```
+
+## Troubleshooting
+
+### Issue: "Cannot find module"
+```bash
+# Rebuild the project
+npm run build
+
+# Check dist folder exists
+ls dist/agentdb/cli/agentdb-cli.js
+```
+
+### Issue: "Database is locked"
+```bash
+# Close any open database connections
+# Or use a different database path
+export AGENTDB_PATH=./test2-agentdb.db
+```
+
+### Issue: "Permission denied"
+```bash
+# Make CLI executable
+chmod +x dist/agentdb/cli/agentdb-cli.js
+```
+
+## Advanced Testing
+
+### Test with Programmatic API
+```typescript
+import { AgentDBCLI } from './dist/agentdb/cli/agentdb-cli.js';
+
+const cli = new AgentDBCLI('./test.db');
+
+// Store episode
+await cli.reflexionStore({
+  sessionId: 'test-1',
+  task: 'example',
+  reward: 0.9,
+  success: true,
+  critique: 'Good approach'
+});
+
+// Retrieve episodes
+await cli.reflexionRetrieve({
+  task: 'example',
+  k: 5
+});
+```
+
+### Integration with Your Project
+```typescript
+// In your agent code
+import { AgentDBCLI } from 'agentic-flow/agentdb';
+
+const memory = new AgentDBCLI();
+
+// Learn from task outcomes
+async function learnFromTask(task, outcome) {
+  await memory.reflexionStore({
+    sessionId: getCurrentSession(),
+    task: task.name,
+    reward: outcome.score,
+    success: outcome.passed,
+    critique: outcome.feedback,
+    input: task.input,
+    output: outcome.result,
+    latencyMs: outcome.duration,
+    tokensUsed: outcome.tokens
+  });
+}
+
+// Retrieve similar past experiences
+async function recallSimilar(task) {
+  return await memory.reflexionRetrieve({
+    task: task.name,
+    k: 5,
+    minReward: 0.7
+  });
+}
+```
+
+## Performance Benchmarks
+
+```bash
+# Time a full workflow
+time bash full-workflow-example.sh
+
+# Check database performance
+node dist/agentdb/cli/agentdb-cli.js db stats
+```
+
+## Next Steps
+
+1. ✅ Test basic commands (reflexion, skill)
+2. ✅ Test causal features (edges, experiments)
+3. ✅ Run nightly learner for discovery
+4. ✅ Verify causal recall with certificates
+5. ✅ Check database stats
+6. 🚀 Integrate into your agent workflows
+
+## Resources
+
+- **AgentDB Controllers**: `/src/agentdb/controllers/`
+- **CLI Source**: `/src/agentdb/cli/agentdb-cli.ts`
+- **Tests**: `/src/agentdb/tests/frontier-features.test.ts`
+- **Binary**: `/dist/agentdb/cli/agentdb-cli.js`