Agent/subagent support

package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@ztimson/ai-utils",
-	"version": "1.2.10",
+	"version": "1.3.0",
 	"description": "AI Utility library",
 	"author": "Zak Timson",
 	"license": "MIT",

src/antrhopic.ts

@@ -21,10 +21,10 @@ export class Anthropic extends LLMProvider {
 				messages.push(<any>{timestamp, ...h});
 			} else {
 				const textContent = h.content?.filter((c: any) => c.type == 'text').map((c: any) => c.text).join('\n\n');
-				if(textContent) messages.push({timestamp, role: h.role, content: textContent});
+				if(textContent) messages.push({role: h.role, content: textContent, timestamp: timestamp});
 				h.content.forEach((c: any) => {
 					if(c.type == 'tool_use') {
-						messages.push({timestamp, role: 'tool', id: c.id, name: c.name, args: c.input, content: undefined});
+						messages.push({role: 'tool', id: c.id, name: c.name, args: c.input, timestamp: c.timestamp, content: undefined});
 					} else if(c.type == 'tool_result') {
 						const m: any = messages.findLast(m => (<any>m).id == c.tool_use_id);
 						if(m) m[c.is_error ? 'error' : 'content'] = c.content;
@@ -46,7 +46,7 @@ export class Anthropic extends LLMProvider {
 				i++;
 			}
 		}
-		return history.map(({timestamp, ...h}) => h);
+		return history;
 	}
 
 	ask(message: string, options: LLMRequest = {}): AbortablePromise<string | any> {
@@ -83,8 +83,9 @@ export class Anthropic extends LLMProvider {
 				};
 			}
 
-			let resp: any, isFirstMessage = true;
+			let resp: any, isFirstMessage = true, terminal = false;
 			do {
+				requestParams.messages = history.map(({timestamp, ...m}) => m);
 				resp = await this.client.messages.create(requestParams).catch(err => {
 					err.message += `\n\nMessages:\n${JSON.stringify(history, null, 2)}`;
 					throw err;
@@ -113,7 +114,7 @@ export class Anthropic extends LLMProvider {
 							}
 						} else if(chunk.type === 'content_block_stop') {
 							const last = resp.content.at(-1);
-							if(last.input != null) last.input = last.input ? JSONAttemptParse(last.input, {}) : {};
+							if(last?.input != null) last.input = last.input ? JSONAttemptParse(last.input, {}) : {};
 						} else if(chunk.type === 'message_stop') {
 							break;
 						}
@@ -123,31 +124,35 @@ export class Anthropic extends LLMProvider {
 				// Run tools
 				const toolCalls = resp.content.filter((c: any) => c.type === 'tool_use');
 				if(toolCalls.length && !controller.signal.aborted) {
-					history.push({role: 'assistant', content: resp.content});
+					history.push({role: 'assistant', content: resp.content, timestamp: Date.now()});
 					const results = await Promise.all(toolCalls.map(async (toolCall: any) => {
 						const tool = tools.find(findByProp('name', toolCall.name));
 						if(options.stream) options.stream({tool: toolCall.name});
 						if(!tool) return {tool_use_id: toolCall.id, is_error: true, content: 'Tool not found'};
 						try {
-							const result = await tool.fn(toolCall.input, options?.stream, this.ai);
+							// Wrap stream so a tool's `done` ends turn gracefully
+							const toolStream = options.stream && ((chunk: any) => {
+								if(chunk.done) { terminal = true; return; }
+								options.stream!(chunk);
+							});
+							const result = await tool.fn(toolCall.input, toolStream, this.ai);
 							return {type: 'tool_result', tool_use_id: toolCall.id, content: typeof result == 'object' ? JSONSanitize(result) : result};
 						} catch (err: any) {
 							return {type: 'tool_result', tool_use_id: toolCall.id, is_error: true, content: err?.message || err?.toString() || 'Unknown'};
 						}
 					}));
-					history.push({role: 'user', content: results});
+					history.push({role: 'user', content: results, timestamp: Date.now()});
 					requestParams.messages = history;
 				}
-			} while (!controller.signal.aborted && resp.content.some((c: any) => c.type === 'tool_use'));
+			} while (!terminal && !controller.signal.aborted && resp.content.some((c: any) => c.type === 'tool_use'));
 
+			if(!terminal) {
 				const textContent = resp.content.filter((c: any) => c.type == 'text').map((c: any) => c.text).join('\n\n');
-			history.push({role: 'assistant', content: textContent});
+				history.push({role: 'assistant', content: textContent, timestamp: Date.now()});
+			}
 			history = this.toStandard(history);
-
 			if(options.stream) options.stream({done: true});
 			if(options.history) options.history.splice(0, options.history.length, ...history);
-
-			// Return parsed JSON if schema provided
 			const finalContent = history.at(-1)?.content;
 			res(options.schema ? JSONAttemptParse(finalContent, finalContent) : finalContent);
 		}), {abort: () => controller.abort()});

src/index.ts

@@ -3,8 +3,6 @@ export * from './antrhopic';
 export * from './audio';
 export * from './llm';
 export * from './memory';
-export * from './memory-cache';
-export * from './memory-graph';
 export * from './open-ai';
 export * from './provider';
 export * from './tools';

src/llm.ts

@@ -1,17 +1,31 @@
+import {snakeCase} from '@ztimson/utils';
 import {AbortablePromise, Ai} from './ai.ts';
 import {Anthropic} from './antrhopic.ts';
-import {MemoryCache} from './memory-cache.ts';
 import {OpenAi} from './open-ai.ts';
 import {LLMProvider} from './provider.ts';
 import {AiTool, AiToolArg} from './tools.ts';
 import {fileURLToPath} from 'url';
 import {dirname, join} from 'path';
 import {spawn} from 'node:child_process';
-import {Memory, MemoryManager} from './memory.ts';
+import {Memory, MemoryCache, MemoryManager, MemoryOptions} from './memory.ts';
 
 export type AnthropicConfig = {proto: 'anthropic', token: string};
 export type OpenAiConfig = {proto: 'openai', host?: string, token: string};
 
+export type Agent = {
+	name: string;
+	description?: string;
+	model?: string | null;
+	temperature?: number;
+	system: string;
+	delegate?: boolean;
+	skills?: Skill[] | null;
+	tools?: AiTool[] | null;
+	mcp?: McpServer[] | null;
+	/** Explicit whitelist of agents this agent may delegate to. Default: none - must opt-in, self is always excluded */
+	agents?: string[] | null;
+}
+
 export type LLMMessage = {
 	/** Message originator */
 	role: 'assistant' | 'system' | 'user';
@@ -56,13 +70,17 @@ export type LLMRequest = {
 	/** Compress old messages in the chat to free up context */
 	compress?: {max: number; min: number};
 	/** User's memory documents - RAG injected automatically each turn */
-	memory?: Memory[] | MemoryCache;
+	memory?: Memory[] | MemoryCache | MemoryOptions;
 	/** Model to use for memory operations */
 	memoryModel?: string;
 	/** Skill documents the AI can browse and read on demand */
 	skills?: Skill[];
 	/** MCP servers to connect and expose as tools */
 	mcp?: McpServer[];
+	/** Subagents exposed as delegatable/wrapped tools */
+	agents?: Agent[];
+	/** @internal recursion guard for nested agent delegation */
+	_agentDepth?: number;
 }
 
 export type McpServer = {
@@ -83,6 +101,7 @@ export type Skill = {
 	content: string;
 }
 
+const MAX_AGENT_DEPTH = 5;
 
 class LLM {
 	private memoryManager!: MemoryManager;
@@ -100,6 +119,60 @@ class LLM {
 		this.memoryManager = new MemoryManager(this);
 	}
 
+	/**
+	 * Wrap agents as tools. Nested delegation is opt-in only (empty by default, like
+	 * tools/skills/mcp) and an agent can never call itself even if explicitly whitelisted.
+	 * Delegate results are queued in `pending` and spliced into history by `ask()` after
+	 * the provider's own end-of-turn history sync has already run.
+	 */
+	private setupAgent(agents: Agent[] = [], allAgents: Agent[], pending: Map<string, {resp: string, subHistory: LLMMessage[]}[]>, aborts: (() => void)[], depth = 0): AiTool[] {
+		return agents.map(a => {
+			const toolName = `${a.delegate ? '' : 'sub'}agent_${snakeCase(a.name)}`;
+			return {
+				name: toolName,
+				description: `${a.delegate ? 'Delegate to ' : ''}Subagent: ${a.description || a.name}`,
+				args: {
+					context: {type: 'string', description: 'Summary of related messages, samples, files, etc...', required: true},
+					instructions: {type: 'string', description: 'Detailed instructions for subagent to complete', required: true},
+				},
+				fn: async (args: any, stream: any) => {
+					if(depth >= MAX_AGENT_DEPTH) return 'Max agent delegation depth exceeded';
+
+					const subHistory: LLMMessage[] = [];
+					// Opt-in only, self always excluded regardless of whitelist
+					const nested = (a.agents || [])
+						.map(name => allAgents.find(x => x.name === name))
+						.filter((x): x is Agent => !!x && x.name !== a.name);
+
+					const request = this.ask(`${args.instructions}${args.context ? `\n\n<context>${args.context}</context>` : ''}`, {
+						system: `You are a specialized subagent. ${a.delegate ? 'Your output streams directly to the user for the remainder of this turn.' : 'You are wrapped in a tool call that will be analysis by an LLM'}
+As a subagent, focus on executing your task completely using available tools and returning only the final result - no commentary, questions, or dialogue.
+
+${a.system}`,
+						model: a.model || undefined,
+						temperature: a.temperature,
+						stream: a.delegate ? stream : undefined,
+						history: subHistory,
+						mcp: a.mcp || undefined,
+						skills: a.skills || undefined,
+						tools: a.tools || undefined,
+						agents: nested,
+						_agentDepth: depth + 1,
+					} as any);
+					aborts.push(request.abort);
+					const resp = await request;
+
+					if(a.delegate) {
+						if(!pending.has(toolName)) pending.set(toolName, []);
+						pending.get(toolName)!.push({resp, subHistory});
+						return '';
+					}
+					return resp;
+				}
+			};
+		});
+	}
+
 	private async setupMcp(servers: McpServer[] = []): Promise<{prompt: string, tools: AiTool[]}> {
 		if(!servers?.length) return {prompt: '', tools: []};
 		const allTools: AiTool[] = [];
@@ -170,9 +243,11 @@ class LLM {
 		if(!this.models[m]) throw new Error(`Model does not exist: ${m}`);
 		let request: AbortablePromise<string> | null = null;
 		let aborted = false;
+		const nestedAborts: (() => void)[] = [];
 		const abort = () => {
 			aborted = true;
 			request?.abort?.();
+			nestedAborts.forEach(a => a());
 		};
 
 		const promise = (async () => {
@@ -196,22 +271,46 @@ class LLM {
 				tools.push(...s.tools);
 			}
 
+			// Agents
+			const agents = options.agents || this.ai.options?.llm?.agents;
+			const pendingDelegates = new Map<string, {resp: string, subHistory: LLMMessage[]}[]>();
+			if(agents?.length) tools.push(...this.setupAgent(agents, agents, pendingDelegates, nestedAborts, options._agentDepth || 0));
+
 			// Memory
-			if (options.memory) {
-				const mems = options.memory instanceof MemoryCache ? options.memory.memories : options.memory;
+			const mem = MemoryManager.normalize(options.memory);
+			if(mem) {
+				const mems = mem.memory instanceof MemoryCache ? mem.memory.memories : mem.memory;
 				if(mems.length) {
-					const relevant = await this.memoryManager.recollect(message, options.memory, 5);
+					if(mem.inject) {
+						const pool = 15; // candidates considered, cheap since only refs are listed
+						const budget = mem.maxTokens ?? 2000; // actual content injected
+						const relevant = await this.memoryManager.recollect(message, mem.memory, pool);
+
+						let used = 0;
+						const preloaded: typeof relevant = [];
+						const listed: typeof relevant = [];
+						for(const r of relevant) {
+							const t = this.estimateTokens(r.content);
+							if(used + t <= budget || preloaded.length === 0) {
+								preloaded.push(r);
+								used += t;
+							} else listed.push(r);
+						}
+
 						prompts.unshift(`You have access to the following memory files:
-	${mems.map(m => `- ${m.name}: ${m.description}`).join('\n')}
-	${relevant.length ? `
-	Relevant memories have been preloaded:
-	${relevant.map(r => `
-	**${r.name}**
-	${r.description}
-	${r.content}
-	`).join('\n---\n')}
-	` : ''}`.trim());
-					tools.push(this.memoryManager.tools.read(options.memory));
+${mems.map(m => `- ${m.name}: ${m.description}`).join('\n')}
+${preloaded.length ? `
+Relevant memories have been preloaded:
+${preloaded.map(r => `
+**${r.name}**
+${r.description}
+${r.content}
+`).join('\n---\n')}
+` : ''}${listed.length ? `
+Also relevant but not preloaded (use \`memory_recall\`): ${listed.map(r => r.name).join(', ')}
+` : ''}`.trim());
+					}
+					if(mem.tool) tools.push(this.memoryManager.tools.read(mem.memory));
 				}
 			}
 
@@ -219,16 +318,33 @@ class LLM {
 
 			prompts.unshift(options.system || this.ai.options.llm?.system || '');
 			request = this.models[m].ask(message, {...options, tools, system: prompts.filter(Boolean).join('\n\n')});
-			const resp = await request;
+			let resp = await request;
+
+			// Spice delegated agents response into history
+			let lastDelegateResp: string | null = null;
+			if(pendingDelegates.size) {
+				for(let i = 0; i < history.length; i++) {
+					const h = history[i];
+					if(h.role !== 'tool' || h.content !== '') continue;
+					const queue = pendingDelegates.get(h.name);
+					if(!queue?.length) continue;
+					const {resp: delegateResp, subHistory} = queue.shift()!;
+					const insert: LLMMessage[] = [...subHistory.filter(sh => sh.role === 'tool'), {role: 'assistant', content: delegateResp, timestamp: Date.now()}];
+					history.splice(i + 1, 0, ...insert);
+					lastDelegateResp = delegateResp;
+					i += insert.length;
+				}
+			}
+
+			// If the orchestrator added no commentary of its own, its answer IS the delegate's answer
+			if(typeof resp === 'string' && !resp.trim() && lastDelegateResp !== null) resp = lastDelegateResp;
 
 			// Trim memory injections from history
-			if(options.memory) {
-				history.splice(0, history.length, ...history.filter(h => h.role !== 'tool' || h.name !== 'recall'));
-			}
+			if(mem?.tool) history.splice(0, history.length, ...history.filter(h => h.role !== 'tool' || h.name !== 'recall'));
 
 			// Auto-memorize before compressing
 			if(options.compress && this.estimateTokens(history) >= options.compress.max) {
-				if(options.memory) await this.memoryManager.memorize(history, options.memory, {model: options.memoryModel || this.defaultModel, ...options});
+				if(mem?.update) await this.memoryManager.memorize(history, mem.memory, {model: options.memoryModel || this.defaultModel, ...options});
 				const compressed = await this.compressHistory(history, options.compress.max, options.compress.min, options);
 				if(options.history) options.history.splice(0, options.history.length, ...compressed);
 			}
@@ -397,15 +513,33 @@ class LLM {
 	 * @param {string} searchTerms Multiple search terms to check against target
 	 * @returns {{avg: number, max: number, similarities: number[]}} Similarity values 0-1: 0 = unique, 1 = identical
 	 */
-	fuzzyMatch(target: string, ...searchTerms: string[]) {
-		if(searchTerms.length < 2) throw new Error('Requires at least 2 strings to compare');
-		const vector = (text: string, dimensions: number = 10): number[] => {
-			return text.toLowerCase().split('').map((char, index) =>
-				(char.charCodeAt(0) * (index + 1)) % dimensions / dimensions).slice(0, dimensions);
+	fuzzyMatch(target, ...searchTerms) {
+		if (searchTerms.length < 2) throw new Error('Requires at least 2 strings to compare');
+		const levenshtein = (a, b) => {
+			const m = a.length, n = b.length;
+			if (!m) return n;
+			if (!n) return m;
+			const dp = Array.from({length: m + 1}, (_, i) => [i, ...Array(n).fill(0)]);
+			for (let j = 0; j <= n; j++) dp[0][j] = j;
+			for (let i = 1; i <= m; i++) {
+				for (let j = 1; j <= n; j++) {
+					dp[i][j] = a[i - 1] === b[j - 1]
+						? dp[i - 1][j - 1]
+						: 1 + Math.min(dp[i - 1][j - 1], dp[i - 1][j], dp[i][j - 1]);
 				}
-		const v = vector(target);
-		const similarities = searchTerms.map(t => vector(t)).map(refVector => this.cosineSimilarity(v, refVector));
-		return {avg: similarities.reduce((acc, s) => acc + s, 0) / similarities.length, max: Math.max(...similarities), similarities};
+			}
+			return dp[m][n];
+		};
+		const similarity = (a, b) => {
+			a = a.toLowerCase(); b = b.toLowerCase();
+			return 1 - levenshtein(a, b) / Math.max(a.length, b.length, 1);
+		};
+		const similarities = searchTerms.map(t => similarity(target, t));
+		return {
+			avg: similarities.reduce((acc, s) => acc + s, 0) / similarities.length,
+			max: Math.max(...similarities),
+			similarities
+		};
 	}
 
 	/**

src/memory-cache.ts

@@ -1,59 +0,0 @@
-import {KDPoint, KDTree} from './kd-tree.ts';
-import {Memory, MemoryRef} from './memory.ts';
-
-export class MemoryCache {
-	private tree: KDTree<MemoryRef>;
-	public memories: Memory[];
-
-	get length() { return this.memories.length; }
-
-	constructor(memories: Memory[]) {
-		this.memories = memories;
-		this.tree = this.buildTree();
-	}
-
-	private buildTree(): KDTree<MemoryRef> {
-		const embedded = this.memories.filter(m => m.embedding?.length);
-		if (!embedded.length) return new KDTree<MemoryRef>(0);
-
-		const dims = embedded[0].embedding.length;
-		const points: KDPoint<MemoryRef>[] = embedded.map(m => ({
-			vector: m.embedding,
-			payload: {name: m.name, description: m.description},
-		}));
-
-		return new KDTree<MemoryRef>(dims, 'cosine', points);
-	}
-
-	search(query: number[], limit: number): MemoryRef[] {
-		const results = this.tree.knn(query, limit);
-		return results.map(r => r.point.payload);
-	}
-
-	add(memory: Memory): void {
-		this.memories.push(memory);
-		this.rebuild();
-	}
-
-	update(memory: Memory): void {
-		const idx = this.memories.findIndex(m => m.name === memory.name);
-		if (idx !== -1) {
-			this.memories[idx] = memory;
-		} else {
-			this.memories.push(memory);
-		}
-		this.rebuild();
-	}
-
-	remove(name: string): void {
-		const idx = this.memories.findIndex(m => m.name === name);
-		if (idx !== -1) {
-			this.memories.splice(idx, 1);
-			this.rebuild();
-		}
-	}
-
-	rebuild(): void {
-		this.tree = this.buildTree();
-	}
-}

src/memory-graph.ts

@@ -1,69 +0,0 @@
-import {MemoryCache} from './memory-cache.ts';
-import {extractMetadata, Memory, MemoryNode} from './memory.ts';
-
-export function buildMemoryGraph(memories: Memory[] | MemoryCache): MemoryNode[] {
-	const mems = memories instanceof MemoryCache ? memories.memories : memories;
-	const nameSet = new Set(mems.map(m => m.name));
-	const ghosts = new Set<string>();
-
-	const nodes: MemoryNode[] = mems.map(m => {
-		const {links, backlinks} = extractMetadata(m.content);
-		return {
-			name: m.name,
-			missing: false,
-			links,
-			backlinks,
-		};
-	});
-
-	for (const node of nodes) {
-		for (const link of node.links) {
-			if (!nameSet.has(link)) ghosts.add(link);
-		}
-	}
-
-	return [
-		...nodes,
-		...[...ghosts].map(name => ({
-			name,
-			missing: true,
-			links: [],
-			backlinks: nodes
-				.filter(n => n.links.includes(name))
-				.map(n => n.name),
-		}))
-	];
-}
-
-export function renderMemoryGraph(nodes) {
-	if (!nodes.length) return 'No memories yet.';
-
-	const groups = new Map();
-	for (const node of nodes) {
-		const [prefix, ...rest] = node.name.split('/');
-		const group = rest.length ? prefix : 'Root';
-		const label = rest.length ? rest.join('/') : node.name;
-		if (!groups.has(group)) groups.set(group, []);
-		groups.get(group).push({...node, label});
-	}
-
-	const ghostCount = nodes.filter(n => n.missing).length;
-	const lines = [`Memory Graph (${nodes.length} nodes, ${ghostCount} ghost${ghostCount === 1 ? '' : 's'})`, ''];
-
-	for (const group of [...groups.keys()].sort()) {
-		const items = groups.get(group).sort((a, b) => a.label.localeCompare(b.label));
-		lines.push(`${group}/`);
-		items.forEach((n, i) => {
-			const last = i === items.length - 1;
-			const branch = last ? '└─' : '├─';
-			const pad = last ? '  ' : '│ ';
-			const tag = n.missing ? ' (ghost)' : '';
-			lines.push(`  ${branch} ${n.label}${tag}`);
-			if (n.links.length) lines.push(`  ${pad}   → ${n.links.join(', ')}`);
-			if (n.backlinks.length) lines.push(`  ${pad}   ← ${n.backlinks.join(', ')}`);
-		});
-		lines.push('');
-	}
-
-	return lines.join('\n').trimEnd();
-}

src/memory.ts

@@ -1,6 +1,143 @@
 import {LLMRequest, LLMMessage} from './llm.ts';
-import {MemoryCache} from './memory-cache.ts';
 import {AiTool} from './tools.ts';
+import {KDPoint, KDTree} from './kd-tree.ts';
+
+export function buildMemoryGraph(memories: Memory[] | MemoryCache): MemoryNode[] {
+	const mems = memories instanceof MemoryCache ? memories.memories : memories;
+	const nameSet = new Set(mems.map(m => m.name));
+	const ghosts = new Set<string>();
+
+	const nodes: MemoryNode[] = mems.map(m => {
+		const {links, backlinks} = extractMetadata(m.content);
+		return {
+			name: m.name,
+			missing: false,
+			links,
+			backlinks,
+		};
+	});
+
+	for (const node of nodes) {
+		for (const link of node.links) {
+			if (!nameSet.has(link)) ghosts.add(link);
+		}
+	}
+
+	return [
+		...nodes,
+		...[...ghosts].map(name => ({
+			name,
+			missing: true,
+			links: [],
+			backlinks: nodes
+				.filter(n => n.links.includes(name))
+				.map(n => n.name),
+		}))
+	];
+}
+
+export function renderMemoryGraph(nodes) {
+	if (!nodes.length) return 'No memories yet.';
+
+	const groups = new Map();
+	for (const node of nodes) {
+		const [prefix, ...rest] = node.name.split('/');
+		const group = rest.length ? prefix : 'Root';
+		const label = rest.length ? rest.join('/') : node.name;
+		if (!groups.has(group)) groups.set(group, []);
+		groups.get(group).push({...node, label});
+	}
+
+	const ghostCount = nodes.filter(n => n.missing).length;
+	const lines = [`Memory Graph (${nodes.length} nodes, ${ghostCount} ghost${ghostCount === 1 ? '' : 's'})`, ''];
+
+	for (const group of [...groups.keys()].sort()) {
+		const items = groups.get(group).sort((a, b) => a.label.localeCompare(b.label));
+		lines.push(`${group}/`);
+		items.forEach((n, i) => {
+			const last = i === items.length - 1;
+			const branch = last ? '└─' : '├─';
+			const pad = last ? '  ' : '│ ';
+			const tag = n.missing ? ' (ghost)' : '';
+			lines.push(`  ${branch} ${n.label}${tag}`);
+			if (n.links.length) lines.push(`  ${pad}   → ${n.links.join(', ')}`);
+			if (n.backlinks.length) lines.push(`  ${pad}   ← ${n.backlinks.join(', ')}`);
+		});
+		lines.push('');
+	}
+
+	return lines.join('\n').trimEnd();
+}
+
+export class MemoryCache {
+	private tree: KDTree<MemoryRef>;
+	public memories: Memory[];
+
+	get length() { return this.memories.length; }
+
+	constructor(memories: Memory[]) {
+		this.memories = memories;
+		this.tree = this.buildTree();
+	}
+
+	private buildTree(): KDTree<MemoryRef> {
+		const embedded = this.memories.filter(m => m.embedding?.length);
+		if (!embedded.length) return new KDTree<MemoryRef>(0);
+
+		const dims = embedded[0].embedding.length;
+		const points: KDPoint<MemoryRef>[] = embedded.map(m => ({
+			vector: m.embedding,
+			payload: {name: m.name, description: m.description},
+		}));
+
+		return new KDTree<MemoryRef>(dims, 'cosine', points);
+	}
+
+	search(query: number[], limit: number): MemoryRef[] {
+		const results = this.tree.knn(query, limit);
+		return results.map(r => r.point.payload);
+	}
+
+	add(memory: Memory): void {
+		this.memories.push(memory);
+		this.rebuild();
+	}
+
+	update(memory: Memory): void {
+		const idx = this.memories.findIndex(m => m.name === memory.name);
+		if (idx !== -1) {
+			this.memories[idx] = memory;
+		} else {
+			this.memories.push(memory);
+		}
+		this.rebuild();
+	}
+
+	remove(name: string): void {
+		const idx = this.memories.findIndex(m => m.name === name);
+		if (idx !== -1) {
+			this.memories.splice(idx, 1);
+			this.rebuild();
+		}
+	}
+
+	rebuild(): void {
+		this.tree = this.buildTree();
+	}
+}
+
+export type MemoryOptions = {
+	/** Memory object */
+	memory: Memory[] | MemoryCache;
+	/** Inject N memories into the system prompt */
+	inject?: boolean;
+	/** expose recall tool to LLM */
+	tool?: boolean;
+	/** Update memory on compression */
+	update?: boolean;
+	/** Max context size of memories to inject to each call (removed immediately after use) */
+	maxTokens?: number;
+}
 
 export type Memory = {
 	name: string;
@@ -83,8 +220,6 @@ function getWeekSunday(monday: string): string {
 	return d.toISOString().slice(0, 10);
 }
 
-
-
 export class MemoryManager {
 	private pendingMemorizations = new Map<string, {
 		memories: Memory[] | MemoryCache,
@@ -128,6 +263,12 @@ export class MemoryManager {
 
 	constructor(private llm: any) {}
 
+	static normalize(m?: Memory[] | MemoryCache | MemoryOptions) {
+		if(!m) return null;
+		const raw = m instanceof MemoryCache || Array.isArray(m);
+		return raw ? {memory: <Memory[] | MemoryCache>m, inject: true, tool: true, update: true} : {inject: true, tool: true, update: true, ...m};
+	}
+
 	private async createTempMemory(conversation: string): Promise<Memory> {
 		const timestamp = Date.now();
 		const content = `---

src/open-ai.ts

@@ -29,11 +29,11 @@ export class OpenAi extends LLMProvider {
 				}));
 				history.splice(i, 1, ...tools);
 				i += tools.length - 1;
-			} else if(h.role === 'tool' && h.content) {
+			} else if(h.role === 'tool') {
 				const record = history.find(h2 => h.tool_call_id == h2.id);
 				if(record) {
-					if(h.content.includes('"error":')) record.error = h.content;
-					else record.content = h.content;
+					if(h.content?.includes('"error":')) record.error = h.content;
+					else record.content = h.content || '';
 				}
 				history.splice(i, 1);
 				i--;
@@ -51,15 +51,16 @@ export class OpenAi extends LLMProvider {
 					content: null,
 					tool_calls: [{ id: h.id, type: 'function', function: { name: h.name, arguments: JSON.stringify(h.args) } }],
 					refusal: null,
-					annotations: []
+					annotations: [],
+					timestamp: h.timestamp,
 				}, {
 					role: 'tool',
 					tool_call_id: h.id,
-					content: h.error || h.content
+					content: h.error || h.content,
+					timestamp: h.timestamp,
 				});
 			} else {
-				const {timestamp, ...rest} = h;
-				result.push(rest);
+				result.push(h);
 			}
 			return result;
 		}, [] as any[]);
@@ -106,8 +107,9 @@ export class OpenAi extends LLMProvider {
 				};
 			}
 
-			let resp: any, isFirstMessage = true;
+			let resp: any, isFirstMessage = true, terminal = false;
 			do {
+				requestParams.messages = history.map(({timestamp, ...m}) => m);
 				resp = await this.client.chat.completions.create(requestParams).catch(err => {
 					err.message += `\n\nMessages:\n${JSON.stringify(history, null, 2)}`;
 					throw err;
@@ -116,7 +118,7 @@ export class OpenAi extends LLMProvider {
 				if(options.stream) {
 					if(!isFirstMessage) options.stream({text: '\n\n'});
 					else isFirstMessage = false;
-					resp.choices = [{message: {role: 'assistant', content: '', tool_calls: []}}];
+					resp.choices = [{message: {role: 'assistant', content: '', tool_calls: [], timestamp: Date.now()}}];
 					for await (const chunk of resp) {
 						if(controller.signal.aborted) break;
 						if(chunk.choices[0].delta.content) {
@@ -158,28 +160,32 @@ export class OpenAi extends LLMProvider {
 					const results = await Promise.all(toolCalls.map(async (toolCall: any) => {
 						const tool = tools?.find(findByProp('name', toolCall.function.name));
 						if(options.stream) options.stream({tool: toolCall.function.name});
-						if(!tool) return {role: 'tool', tool_call_id: toolCall.id, content: '{"error": "Tool not found"}'};
+						if(!tool) return {role: 'tool', tool_call_id: toolCall.id, content: '{"error": "Tool not found"}', timestamp: Date.now()};
 						try {
 							const args = JSONAttemptParse(toolCall.function.arguments, {});
-							const result = await tool.fn(args, options.stream, this.ai);
-							return {role: 'tool', tool_call_id: toolCall.id, content: typeof result == 'object' ? JSONSanitize(result) : result};
+							// Wrap stream so a tool's `done` ends turn gracefully
+							const toolStream = options.stream && ((chunk: any) => {
+								if(chunk.done) { terminal = true; return; }
+								options.stream!(chunk);
+							});
+							const result = await tool.fn(args, toolStream, this.ai);
+							return {role: 'tool', tool_call_id: toolCall.id, content: typeof result == 'object' ? JSONSanitize(result) : result, timestamp: Date.now()};
 						} catch (err: any) {
-							return {role: 'tool', tool_call_id: toolCall.id, content: JSONSanitize({error: err?.message || err?.toString() || 'Unknown'})};
+							return {role: 'tool', tool_call_id: toolCall.id, content: JSONSanitize({error: err?.message || err?.toString() || 'Unknown'}), timestamp: Date.now()};
 						}
 					}));
 					history.push(...results);
 					requestParams.messages = history;
 				}
-			} while (!controller.signal.aborted && resp.choices?.[0]?.message?.tool_calls?.length);
+			} while (!terminal && !controller.signal.aborted && resp.choices?.[0]?.message?.tool_calls?.length);
 
+			if(!terminal) {
 				const textContent = resp.choices[0].message.content?.trim() || '';
-			history.push({role: 'assistant', content: textContent});
+				history.push({role: 'assistant', content: textContent, timestamp: Date.now()});
+			}
 			history = this.toStandard(history);
-
 			if(options.stream) options.stream({done: true});
 			if(options.history) options.history.splice(0, options.history.length, ...history);
-
-			// Return parsed JSON if schema provided
 			const finalContent = history.at(-1)?.content;
 			res(options.schema ? JSONAttemptParse(finalContent, finalContent) : finalContent);
 		}), {abort: () => controller.abort()});