amalfa 1.0.3 → 1.0.5

package/README.md

@@ -140,9 +140,9 @@ bun install
 
 5. **Restart Claude Desktop**
 
-**Full setup guide:** [docs/MCP_SETUP.md](docs/MCP_SETUP.md)
+**Full setup guide:** See repository docs for detailed MCP setup
 
-**Package status:** Reserved at https://www.npmjs.com/package/amalfa (v1.0 coming soon)
+**Package:** Available at https://www.npmjs.com/package/amalfa
 
 ---
 
@@ -313,15 +313,14 @@ Amalfa evolved from patterns discovered in the [PolyVis](https://github.com/pjsv
 
 ## Roadmap
 
-### v1.0 (Q1 2026)
+### v1.0 (Released)
 
-- ✅ Package name reserved on npm
+- ✅ Published to npm
 - ✅ Core vision documented
 - ✅ Auto-augmentation design complete
-- 🚧 MVP implementation (in progress)
-- [ ] MCP server functional
-- [ ] Basic semantic search working
-- [ ] Initial release
+- ✅ MCP server functional
+- ✅ Basic semantic search working
+- ✅ Initial release
 
 ### v1.1+ (Future)
 

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "amalfa",
-	"version": "1.0.3",
+	"version": "1.0.5",
 	"description": "Local-first knowledge graph engine for AI agents. Transforms markdown into searchable memory with MCP protocol.",
 	"license": "MIT",
 	"homepage": "https://github.com/pjsvis/amalfa#readme",

package/src/cli.ts

@@ -3,7 +3,7 @@ import { existsSync, statSync } from "node:fs";
 import { join } from "node:path";
 import { spawn } from "node:child_process";
 
-const VERSION = "1.0.3";
+const VERSION = "1.0.5";
 
 // Database path loaded from config (lazy loaded per command)
 let DB_PATH: string | null = null;

package/polyvis.settings.json.bak

@@ -1,38 +0,0 @@
-{
-	"_comment": "⚠️  DEPRECATED: This file will be removed in v2.0. Use amalfa.config.json for user-facing settings. See docs/CONFIG_UNIFICATION.md",
-	"paths": {
-		"database": {
-			"resonance": "public/resonance.db"
-		},
-		"docs": {
-			"root": "docs",
-			"webdocs": "docs/webdocs",
-			"architecture": "docs/architecture",
-			"public": "public/docs"
-		},
-		"sources": {
-			"experience": [
-				{ "path": "debriefs", "name": "Debrief" },
-				{ "path": "playbooks", "name": "Playbook" },
-				{ "path": "briefs", "name": "Brief" },
-				{ "path": "docs", "name": "Docs" }
-			],
-			"persona": {
-				"lexicon": "scripts/fixtures/conceptual-lexicon-ref-v1.79.json",
-				"cda": "scripts/fixtures/cda-ref-v63.json"
-			}
-		}
-	},
-	"graph": {
-		"tuning": {
-			"louvain": {
-				"persona": 0.3,
-				"experience": 0.25
-			}
-		}
-	},
-
-	"schema": {
-		"version": "1.0.0"
-	}
-}

package/src/EnlightenedTriad.ts

@@ -1,146 +0,0 @@
-// --- CONFIGURATION ---
-const PORTS = {
-	SCOUT: 8082, // Phi-3.5 (Fast Entity Extraction)
-	ARCHITECT: 8083, // Llama-3 (Strict Logic/JSON)
-	AUDITOR: 8084, // Olmo-3 (Deep Thinking)
-};
-
-interface AuditResult {
-	passed: boolean;
-	verdict: string;
-	thought_trace: string;
-}
-
-export class EnlightenedTriad {
-	/**
-	 * THE SCOUT (Phi-3.5)
-	 * "The Bouncer"
-	 * Task: Fast, cheap filtering.
-	 */
-	async scout(context: string, extractionTask: string): Promise<string> {
-		return this.callAgent(PORTS.SCOUT, [
-			{
-				role: "system",
-				content:
-					"You are a precise data extractor. Return only the requested data. No filler.",
-			},
-			{
-				role: "user",
-				content: `Context: ${context}\n\nTask: ${extractionTask}`,
-			},
-		]);
-	}
-
-	/**
-	 * THE ARCHITECT (Llama-3 + Accountant Vector)
-	 * "The Builder"
-	 * Task: Structure raw text into strict JSON.
-	 */
-	async architect(unstructuredData: string): Promise<unknown> {
-		const response = await this.callAgent(PORTS.ARCHITECT, [
-			{ role: "system", content: "Output JSON only." },
-			{
-				role: "user",
-				content: `Convert to Causal Graph JSON: ${unstructuredData}`,
-			},
-		]);
-
-		try {
-			// Strip markdown code blocks if present
-			const cleanJson = response
-				.replace(/```json/g, "")
-				.replace(/```/g, "")
-				.trim();
-			return JSON.parse(cleanJson);
-		} catch (_e) {
-			console.error("❌ Architect JSON Error. Raw output:", response);
-			return null;
-		}
-	}
-
-	/**
-	 * THE AUDITOR (Olmo-3-Think)
-	 * "The QA Department"
-	 * Task: Think for 2 minutes, then vote Yes/No.
-	 */
-	async audit(claim: string): Promise<AuditResult> {
-		// High temp for creative debugging/thinking
-		const rawOutput = await this.callAgent(
-			PORTS.AUDITOR,
-			[
-				{
-					role: "user",
-					content: `Analyze the validity of this logical claim. \n\nClaim: "${claim}"\n\nShow your thinking process wrapped in <think> tags like this: <think> ... </think>, then end with 'VERDICT: PASS' or 'VERDICT: FAIL'.`,
-				},
-			],
-			{ temperature: 0.6, max_tokens: 2048 },
-		);
-
-		// --- THE PARSER ---
-		// Olmo uses <think> tags. We split the "Raj" monologue from the Answer.
-		let thoughtTrace = "";
-		let finalAnswer = "";
-
-		const thinkMatch = rawOutput.match(/<think>([\s\S]*?)<\/think>/);
-
-		if (thinkMatch?.[1]) {
-			// case 1: structured output
-			thoughtTrace = thinkMatch[1].trim();
-			finalAnswer = rawOutput.replace(thinkMatch[0], "").trim();
-		} else if (rawOutput.includes("VERDICT:")) {
-			// case 2: implicit separation (fallback)
-			const parts = rawOutput.split("VERDICT:");
-			thoughtTrace = parts[0]?.trim() ?? "";
-			// Reconstruct the verdict part
-			finalAnswer = `VERDICT:${parts.slice(1).join("VERDICT:")}`.trim();
-		} else {
-			// case 3: total failure to structure
-			thoughtTrace = rawOutput;
-			finalAnswer = "VERDICT: FAIL (Parse Error)";
-		}
-
-		const passed = finalAnswer.includes("VERDICT: PASS");
-
-		return {
-			passed,
-			verdict: finalAnswer, // The short answer for UI
-			thought_trace: thoughtTrace, // The monologue for DB/Audit
-		};
-	}
-
-	// --- UTILS (Bun Native Fetch) ---
-	private async callAgent(
-		port: number,
-		messages: { role: string; content: string }[],
-		options: { temperature?: number; max_tokens?: number } = {},
-	): Promise<string> {
-		try {
-			const response = await fetch(
-				`http://127.0.0.1:${port}/v1/chat/completions`,
-				{
-					method: "POST",
-					headers: { "Content-Type": "application/json" },
-					body: JSON.stringify({
-						messages,
-						temperature: options.temperature || 0.1,
-						n_predict: options.max_tokens || 512,
-					}),
-				},
-			);
-
-			if (!response.ok) {
-				throw new Error(
-					`HTTP Error ${response.status}: ${response.statusText}`,
-				);
-			}
-
-			const data = (await response.json()) as {
-				choices: { message: { content: string } }[];
-			};
-			return data.choices?.[0]?.message?.content || "";
-		} catch (_error) {
-			console.error(`❌ Agent at Port ${port} is offline/unreachable.`);
-			return "AGENT_OFFLINE";
-		}
-	}
-}

package/src/JIT_Triad.ts

@@ -1,137 +0,0 @@
-import { type Subprocess, spawn } from "bun";
-import { EnlightenedTriad } from "./EnlightenedTriad"; // Re-using your existing interface
-
-// --- CONFIGURATION ---
-const BASE_DIR = `${import.meta.dir}/../experiments/enlightenment`; // Adjust relative to src/
-const BIN_PATH = `${BASE_DIR}/llama.cpp/build/bin/llama-server`;
-const VECTORS_DIR = `${BASE_DIR}/vectors`;
-
-const AGENTS = {
-	SCOUT: {
-		port: 8082,
-		model: `${VECTORS_DIR}/Phi-3.5-mini-instruct-Q4_K_M.gguf`,
-		ctx: 4096,
-		vector: null,
-	},
-	ARCHITECT: {
-		port: 8083,
-		model: `${VECTORS_DIR}/Meta-Llama-3-8B-Instruct-Q4_K_M.gguf`,
-		ctx: 8192,
-		vector: `${VECTORS_DIR}/enlightenment_vector_v2.gguf`,
-		scale: "-0.3",
-	},
-	AUDITOR: {
-		port: 8084,
-		model: `${VECTORS_DIR}/Olmo-3-7B-Think-Q4_K_M.gguf`,
-		ctx: 8192,
-		vector: null,
-	},
-};
-
-let currentProcess: Subprocess | null = null;
-
-// --- LIFECYCLE MANAGER ---
-async function bootAgent(role: keyof typeof AGENTS) {
-	if (currentProcess) {
-		console.log("♻️  Freeing VRAM (Stopping previous agent)...");
-		currentProcess.kill();
-		await new Promise((r) => setTimeout(r, 1000)); // Cool down
-	}
-
-	const config = AGENTS[role];
-	console.log(`🚀 Booting ${role} on Port ${config.port}...`);
-
-	const args = [
-		BIN_PATH,
-		"-m",
-		config.model,
-		"--port",
-		config.port.toString(),
-		"--ctx-size",
-		config.ctx.toString(),
-		"--n-gpu-layers",
-		"99", // Metal
-		"--log-disable", // Keep console clean
-	];
-
-	if (config.vector) {
-		args.push("--control-vector-scaled", `${config.vector}:${config.scale}`);
-	}
-
-	// Spawn via Bun
-	currentProcess = spawn(args, {
-		stdout: "ignore", // Silence server logs
-		stderr: "ignore",
-	});
-
-	// Wait for Health Check
-	process.stdout.write("   Waiting for neural activity...");
-	for (let i = 0; i < 30; i++) {
-		try {
-			const res = await fetch(`http://127.0.0.1:${config.port}/health`);
-			if (res.ok) {
-				console.log(" Online! 🟢");
-				return;
-			}
-		} catch (_e) {}
-		await new Promise((r) => setTimeout(r, 500));
-		process.stdout.write(".");
-	}
-	throw new Error(`${role} failed to start.`);
-}
-
-async function shutdown() {
-	if (currentProcess) {
-		console.log("\n🛑 Shutting down final agent...");
-		currentProcess.kill();
-	}
-}
-
-// --- THE DEMO FLOW ---
-async function runOptimizedPipeline() {
-	const triad = new EnlightenedTriad();
-
-	try {
-		console.log("⚡️ SYSTEM ONLINE: Engaging OPTIMIZED Intelligence...\n");
-		const rawLog =
-			"2025-12-19 14:02:11 [CRITICAL] Connection refused at 192.168.1.5 (DB_SHARD_04). Latency 4005ms.";
-
-		// --- STEP 1: SCOUT ---
-		await bootAgent("SCOUT");
-		console.log("\n--- 🕵️ SCOUT TASK ---");
-		const scoutResult = await triad.scout(
-			rawLog,
-			"Extract IP and Error Message. No notes.",
-		);
-		console.log(`>> Output: ${scoutResult}`);
-
-		// --- STEP 2: ARCHITECT ---
-		await bootAgent("ARCHITECT");
-		console.log("\n--- 📐 ARCHITECT TASK ---");
-		const architectResult = await triad.architect(scoutResult);
-		console.log(`>> Output:`, JSON.stringify(architectResult, null, 2));
-
-		// --- STEP 3: AUDITOR ---
-		await bootAgent("AUDITOR");
-		console.log("\n--- 🧠 AUDITOR TASK ---");
-		const claim = `The error 'Connection refused' caused the high latency.`;
-		const auditResult = await triad.audit(claim);
-
-		console.log(
-			`\n📝 THOUGHT TRACE:\n${auditResult.thought_trace.substring(0, 300)}...`,
-		);
-		console.log(`\n⚖️ VERDICT: ${auditResult.passed ? "✅ PASS" : "❌ FAIL"}`);
-	} catch (error) {
-		console.error("\n💥 Pipeline Error:", error);
-	} finally {
-		await shutdown();
-	}
-}
-
-// Handle Ctrl+C
-process.on("SIGINT", async () => {
-	await shutdown();
-	process.exit(0);
-});
-
-runOptimizedPipeline();

package/src/data/experience/test_doc_1.md

@@ -1,2 +0,0 @@
-# Test Doc 1
-This is a test document for verification.

package/src/data/experience/test_doc_2.md

@@ -1,2 +0,0 @@
-# Test Doc 2
-This is a second test document for verification.

package/src/demo-triad.ts

@@ -1,45 +0,0 @@
-import { EnlightenedTriad } from "./EnlightenedTriad";
-
-async function runPipeline() {
-	const triad = new EnlightenedTriad();
-
-	console.log("\n⚡️ SYSTEM ONLINE: Engaging Heterogeneous Intelligence...\n");
-
-	// --- STEP 1: RAW INPUT ---
-	const rawLog =
-		"2025-12-19 14:02:11 [CRITICAL] Connection refused at 192.168.1.5 (DB_SHARD_04). Latency 4005ms.";
-	console.log(`📄 INPUT: "${rawLog}"\n`);
-
-	// --- STEP 2: SCOUT (Extraction) ---
-	console.log("--- 🕵️ SCOUT (Phi-3.5) ---");
-	const scoutResult = await triad.scout(
-		rawLog,
-		"Extract the IP address and the specific Error Message.",
-	);
-	console.log(`>> Output: ${scoutResult}\n`);
-
-	// --- STEP 3: ARCHITECT (Structure) ---
-	console.log("--- 📐 ARCHITECT (Llama-3) ---");
-	// Feed Scout's output into Architect
-	const architectResult = await triad.architect(scoutResult);
-	console.log(`>> Output (JSON):`, JSON.stringify(architectResult, null, 2));
-	console.log("");
-
-	// --- STEP 4: AUDITOR (Verification) ---
-	console.log("--- 🧠 AUDITOR (Olmo-3) ---");
-	// Feed a claim based on the structure to the Auditor
-	const claim = `The error 'Connection refused' at 192.168.1.5 caused the high latency.`;
-	const auditResult = await triad.audit(claim);
-
-	console.log(`\n📝 THOUGHT TRACE (The 'Raj' Monologue):`);
-	// Truncate for console readability
-	console.log(
-		`${auditResult.thought_trace.substring(0, 300)}... [truncated] ...`,
-	);
-
-	console.log(
-		`\n⚖️ FINAL VERDICT: ${auditResult.passed ? "✅ PASS" : "❌ FAIL"}`,
-	);
-}
-
-runPipeline();

package/src/gardeners/BaseGardener.ts

@@ -1,55 +0,0 @@
-import type { ResonanceDB } from "@src/resonance/db";
-import { getLogger, type Logger } from "@src/utils/Logger";
-
-export interface Candidate {
-	nodeId: string;
-	filePath: string;
-	content: string;
-	type: string;
-}
-
-export abstract class BaseGardener {
-	protected db: ResonanceDB;
-	protected log: Logger;
-
-	constructor(db: ResonanceDB) {
-		this.db = db;
-		// We use a generic name initially, subclasses can override or we rely on 'name' property later
-		// Actually, we can't access 'this.name' safely in constructor if it's a property.
-		// Let's use "Gardener" as the component.
-		this.log = getLogger("Gardener");
-	}
-
-	abstract name: string;
-
-	/**
-	 * Finds candidates that need attention.
-	 */
-	abstract scan(limit: number): Promise<Candidate[]>;
-
-	/**
-	 * Applies changes to a single candidate.
-	 */
-	abstract cultivate(candidate: Candidate): Promise<void>;
-
-	/**
-	 * Main loop.
-	 */
-	public async run(limit: number = 10) {
-		this.log.info({ gardener: this.name }, "🌿 Gardener starting...");
-		const candidates = await this.scan(limit);
-		this.log.info(
-			{ gardener: this.name, count: candidates.length },
-			"Found candidates",
-		);
-
-		for (const candidate of candidates) {
-			this.log.debug(
-				{ gardener: this.name, nodeId: candidate.nodeId },
-				"Processing candidate",
-			);
-			await this.cultivate(candidate);
-		}
-		this.log.info({ gardener: this.name }, "✅ Gardener finished");
-	}
-}

package/src/llm/EnlightenedProvider.ts

@@ -1,95 +0,0 @@
-// src/llm/EnlightenedProvider.ts
-
-export interface Message {
-	role: "system" | "user" | "assistant";
-	content: string;
-}
-
-export interface EnlightenmentConfig {
-	port?: number;
-	modelAlias?: string;
-	temperature?: number;
-	maxTokens?: number;
-}
-
-/**
- * THE ENLIGHTENED PROVIDER
- * A specialized adapter for the "Kirkcaldy Accountant" vector-steered model.
- * Default Port: 8083
- */
-export class EnlightenedProvider {
-	private baseUrl: string;
-	private modelAlias: string;
-	private defaultTemp: number;
-
-	constructor(config: EnlightenmentConfig = {}) {
-		this.baseUrl = `http://127.0.0.1:${config.port || 8083}/v1`;
-		this.modelAlias = config.modelAlias || "enlightened-llama";
-		this.defaultTemp = config.temperature || 0.1; // Keep it cold for logic
-	}
-
-	/**
-	 * THE RATIONALITY CHECK
-	 * Pings the server to ensure the Enlightenment engine is online.
-	 */
-	async isOnline(): Promise<boolean> {
-		try {
-			const response = await fetch(`${this.baseUrl}/models`);
-			return response.ok;
-		} catch {
-			return false;
-		}
-	}
-
-	/**
-	 * THE THINK METHOD
-	 * Sends the prompt to the vector-clamped model.
-	 */
-	async think(messages: Message[]): Promise<string> {
-		try {
-			const response = await fetch(`${this.baseUrl}/chat/completions`, {
-				method: "POST",
-				headers: {
-					"Content-Type": "application/json",
-					Authorization: "Bearer sk-dummy-key", // Required by protocol, ignored by server
-				},
-				body: JSON.stringify({
-					model: this.modelAlias,
-					messages: messages,
-					temperature: this.defaultTemp,
-					max_tokens: 1024,
-					stream: false,
-				}),
-			});
-
-			if (!response.ok) {
-				throw new Error(
-					`Enlightenment Error: ${response.status} ${response.statusText}`,
-				);
-			}
-
-			const data = (await response.json()) as {
-				choices: { message: { content: string } }[];
-			};
-			return data?.choices?.[0]?.message?.content?.trim() || "";
-		} catch (error) {
-			console.error("🏴󠁧󠁢󠁳󠁣󠁴󠁿 The Philosopher is silent (Connection Error).", error);
-			throw error;
-		}
-	}
-
-	/**
-	 * SPECIALIST METHOD: DE-FLUFF
-	 * A pre-configured routine to strip buzzwords from text.
-	 */
-	async defluff(inputText: string): Promise<string> {
-		return this.think([
-			{
-				role: "system",
-				content:
-					"You are a ruthless editor. Rewrite the following text to be concise, factual, and free of corporate buzzwords. Return ONLY the rewritten text.",
-			},
-			{ role: "user", content: inputText },
-		]);
-	}
-}

package/src/services/README.md

@@ -1,56 +0,0 @@
-# PolyVis Services
-
-This directory contains the service wrappers for the PolyVis backend infrastructure, including the AI model servers.
-
-## Service Lifecycle Management
-
-All services in this directory must implement the `ServiceLifecycle` pattern (`src/utils/ServiceLifecycle.ts`). This ensures consistent behavior for:
-- **Startup**: checking for zombie processes, stale PID files, and creating log files.
-- **Process Management**: standard `start`, `stop`, `restart`, `status` commands.
-- **Zombie Defense**: Integration with the `ZombieDefense` system to prevent rogue processes.
-
-## Available Services
-
-| Service | Script | Port | Description |
-| :--- | :--- | :--- | :--- |
-| **Olmo-3** | `olmo3.ts` | `8084` | The "Auditor" model. Runs with DeepSeek reasoning format for heavy verification tasks. |
-| **Phi-3.5** | `phi.ts` | `8082` | The "Scout" model. Fast, lightweight model for initial queries. |
-| **Llama-3** | `llama.ts` | `8083` | The "Architect" / "Accountant". Steered via Control Vector (Scale: -0.11) for professional, structured output. |
-| **Llama-3-UV** | `llamauv.ts` | `8085` | Unvectored Llama-3. Runs the raw base model for comparison/baseline purposes. |
-
-## Usage
-
-All services can be managed via `bun run`.
-
-```bash
-# Start a service
-bun run <service_name> start
-
-# Stop a service
-bun run <service_name> stop
-
-# Check status
-bun run <service_name> status
-
-# Restart
-bun run <service_name> restart
-```
-
-### Master Status CLI
-
-To view the status of ALL running services:
-
-```bash
-bun run servers
-```
-
-## Adding a New Service
-
-1.  **Duplicate** an existing wrapper (e.g., `src/services/llamauv.ts`).
-2.  **Configure**:
-    *   Update `PORT`, `name`, `pidFile`, and `logFile`.
-    *   Set the correct binary path and arguments in `runServer()`.
-3.  **Register**:
-    *   Add a script to `package.json`.
-    *   Add to the whitelist in `src/utils/ZombieDefense.ts`.
-    *   Add to the status list in `scripts/cli/servers.ts`.

package/src/services/llama.ts

@@ -1,59 +0,0 @@
-import { join } from "node:path";
-import { ServiceLifecycle } from "../utils/ServiceLifecycle";
-
-// --- Configuration ---
-const BASE_DIR = join(import.meta.dir, "../../experiments/enlightenment");
-const BIN_PATH = join(BASE_DIR, "llama.cpp/build/bin/llama-server");
-const MODEL_PATH = join(
-	BASE_DIR,
-	"vectors/Meta-Llama-3-8B-Instruct-Q4_K_M.gguf",
-);
-const VECTOR_PATH = join(BASE_DIR, "vectors/enlightenment_vector_v2.gguf");
-
-const PORT = 8083;
-
-const args = process.argv.slice(2);
-const command = args[0] || "serve";
-
-// --- Service Lifecycle ---
-
-const lifecycle = new ServiceLifecycle({
-	name: "Llama-3",
-	pidFile: ".llama.pid",
-	logFile: ".llama.log",
-	entryPoint: "src/services/llama.ts",
-});
-
-// --- Server Logic ---
-
-async function runServer() {
-	console.log(`🚀 Starting Llama-3 Server on port ${PORT}...`);
-	console.log(`   🧠 Applying Control Vector: The Accountant (-0.11)`);
-
-	const cmd = [
-		BIN_PATH,
-		"-m",
-		MODEL_PATH,
-		"--port",
-		PORT.toString(),
-		"--ctx-size",
-		"8192",
-		"--n-gpu-layers",
-		"99", // Offload to GPU/Metal
-		"--control-vector-scaled",
-		`${VECTOR_PATH}:-0.11`, // The Accountant (Calibrated: Optimal Strength)
-		"--log-disable",
-	];
-
-	const serverProcess = Bun.spawn(cmd, {
-		stdout: "inherit",
-		stderr: "inherit",
-	});
-
-	// Wait for process to exit
-	await serverProcess.exited;
-}
-
-// --- Dispatch ---
-
-await lifecycle.run(command, runServer);

package/src/services/llamauv.ts

@@ -1,56 +0,0 @@
-import { join } from "node:path";
-import { ServiceLifecycle } from "../utils/ServiceLifecycle";
-
-// --- Configuration ---
-const BASE_DIR = join(import.meta.dir, "../../experiments/enlightenment");
-const BIN_PATH = join(BASE_DIR, "llama.cpp/build/bin/llama-server");
-const MODEL_PATH = join(
-	BASE_DIR,
-	"vectors/Meta-Llama-3-8B-Instruct-Q4_K_M.gguf",
-);
-
-const PORT = 8085;
-
-const args = process.argv.slice(2);
-const command = args[0] || "serve";
-
-// --- Service Lifecycle ---
-
-const lifecycle = new ServiceLifecycle({
-	name: "Llama-3-UV",
-	pidFile: ".llamauv.pid",
-	logFile: ".llamauv.log",
-	entryPoint: "src/services/llamauv.ts",
-});
-
-// --- Server Logic ---
-
-async function runServer() {
-	console.log(`🚀 Starting Llama-3-UV (Unvectored) Server on port ${PORT}...`);
-	console.log(`   🧠 Mode: RAW (No Control Vector)`);
-
-	const cmd = [
-		BIN_PATH,
-		"-m",
-		MODEL_PATH,
-		"--port",
-		PORT.toString(),
-		"--ctx-size",
-		"8192",
-		"--n-gpu-layers",
-		"99", // Offload to GPU/Metal
-		"--log-disable",
-	];
-
-	const serverProcess = Bun.spawn(cmd, {
-		stdout: "inherit",
-		stderr: "inherit",
-	});
-
-	// Wait for process to exit
-	await serverProcess.exited;
-}
-
-// --- Dispatch ---
-
-await lifecycle.run(command, runServer);

package/src/services/olmo3.ts

@@ -1,58 +0,0 @@
-import { join } from "node:path";
-import { ServiceLifecycle } from "../utils/ServiceLifecycle";
-
-// --- Configuration ---
-const BASE_DIR = join(import.meta.dir, "../../experiments/enlightenment");
-const BIN_PATH = join(BASE_DIR, "llama.cpp/build/bin/llama-server");
-const MODEL_PATH = join(BASE_DIR, "vectors/Olmo-3-7B-Think-Q4_K_M.gguf");
-
-const PORT = 8084;
-
-const args = process.argv.slice(2);
-const command = args[0] || "serve";
-
-// --- Service Lifecycle ---
-
-const lifecycle = new ServiceLifecycle({
-	name: "Olmo-3",
-	pidFile: ".olmo3.pid",
-	logFile: ".olmo3.log",
-	entryPoint: "src/services/olmo3.ts",
-});
-
-// --- Server Logic ---
-
-async function runServer() {
-	console.log(`🚀 Starting Olmo-3 Server on port ${PORT}...`);
-
-	const cmd = [
-		BIN_PATH,
-		"-m",
-		MODEL_PATH,
-		"--port",
-		PORT.toString(),
-		"--ctx-size",
-		"8192", // Reduced context for stability on Metal
-		"--n-gpu-layers",
-		"99", // Offload to GPU/Metal
-		"--jinja", // Jinja2 template support
-		"--reasoning-format",
-		"deepseek", // Separate partial thinking
-		"-fa",
-		"on", // Flash Attention
-		"--temp",
-		"0.6",
-	];
-
-	const serverProcess = Bun.spawn(cmd, {
-		stdout: "inherit",
-		stderr: "inherit",
-	});
-
-	// Wait for process to exit
-	await serverProcess.exited;
-}
-
-// --- Dispatch ---
-
-await lifecycle.run(command, runServer);

package/src/services/phi.ts

@@ -1,52 +0,0 @@
-import { join } from "node:path";
-import { ServiceLifecycle } from "../utils/ServiceLifecycle";
-
-// --- Configuration ---
-const BASE_DIR = join(import.meta.dir, "../../experiments/enlightenment");
-const BIN_PATH = join(BASE_DIR, "llama.cpp/build/bin/llama-server");
-const MODEL_PATH = join(BASE_DIR, "vectors/Phi-3.5-mini-instruct-Q4_K_M.gguf");
-
-const PORT = 8082;
-
-const args = process.argv.slice(2);
-const command = args[0] || "serve";
-
-// --- Service Lifecycle ---
-
-const lifecycle = new ServiceLifecycle({
-	name: "Phi-3.5",
-	pidFile: ".phi.pid",
-	logFile: ".phi.log",
-	entryPoint: "src/services/phi.ts",
-});
-
-// --- Server Logic ---
-
-async function runServer() {
-	console.log(`🚀 Starting Phi-3.5 Server on port ${PORT}...`);
-
-	const cmd = [
-		BIN_PATH,
-		"-m",
-		MODEL_PATH,
-		"--port",
-		PORT.toString(),
-		"--ctx-size",
-		"4096",
-		"--n-gpu-layers",
-		"99", // Offload to GPU/Metal
-		"--log-disable", // Keep clean for Phi
-	];
-
-	const serverProcess = Bun.spawn(cmd, {
-		stdout: "inherit",
-		stderr: "inherit",
-	});
-
-	// Wait for process to exit
-	await serverProcess.exited;
-}
-
-// --- Dispatch ---
-
-await lifecycle.run(command, runServer);