DEV Community: Alp Allovi

Introducing Ayeixa Agent Memory Spine: Multi-Tenant Episodic Storage & Replay Compaction

Alp Allovi — Mon, 17 Aug 2026 19:56:20 +0000

Introducing Ayeixa Agent Memory Spine: Multi-Tenant Episodic Storage & Replay Compaction

Long-running agentic workflows require access to past conversational turns, error precedents, and decision logs. In multi-tenant environments, storing and retrieving historical context presents two major risks: cross-tenant state leakage and context-window token overflow.

Ayeixa Agent Memory Spine (@ayeixa/agent-memory-spine) is an open-source TypeScript engine providing tenant and session namespaced in-memory episodic storage, memory indexing, and token-budgeted replay compaction.

GitHub Repository: https://github.com/alpallovy/ayeixa-agent-memory-spine
Status: Pre-Release (v0.1.0-alpha)
License: MIT
NPM Status: Public registry publication is pending; evaluate and build locally.

1. Core Architecture

The system consists of four cooperating components:

SessionIsolationEngine: Enforces tenant/session key isolation (${tenantId}::${sessionId}) and verifies that distinct tenant namespaces share zero episodic records.
EpisodicMemoryStore: Structured in-memory storage for episodic events, tags, timestamps, and importance ratings.
ContextualReplayEngine: Retrieves relevant episodes within an exact token ceiling, formatting them into compact replay prompts.
MemoryCompactor: Applies relevance decay and deduplication to maintain high signal-to-noise ratio within limited context budgets.

2. Implemented Capabilities & Test Verification

Verified with hermetic unit tests:

Namespaced Isolation: Zero cross-tenant data overlap verified via verifyIsolation (tests/isolation.test.ts).
Contextual Replay: Token-budgeted episode reconstruction (tests/replay.test.ts).
Memory Compaction: Deduplication and relevance decay algorithms (tests/compactor.test.ts).

Verification: 3/3 hermetic unit tests passing (0 failures).

3. Local Quick Start

git clone https://github.com/alpallovy/ayeixa-agent-memory-spine.git
cd ayeixa-agent-memory-spine
npm ci
npm run build
npm test

Usage Example

import { SessionIsolationEngine, EpisodicMemoryStore, ContextualReplayEngine } from './src';

const store = new EpisodicMemoryStore();
const isolator = new SessionIsolationEngine(store);

isolator.storeEpisode("tenant-101", "session-A", {
  id: "ep-1",
  role: "developer",
  content: "Fixed regression in authentication token verification",
  tags: ["auth", "security"],
  timestamp: new Date().toISOString(),
  importance: 0.95
});

const replay = new ContextualReplayEngine(store);
const context = replay.buildReplayContext("tenant-101", "session-A", { maxTokenBudget: 500 });
console.log("Replay Context Episodes:", context.episodes);

4. Limitations & Contributing

Pre-release v0.1.0-alpha. Operates as an in-memory store.
Public npm publication is pending.
Feedback and PRs are welcome! Check out open good first issue tags on GitHub.

License: MIT

Introducing Ayeixa MCP Guardian: Capability Fencing & Parameter Sanitization for MCP

Alp Allovi — Mon, 17 Aug 2026 19:55:48 +0000

Introducing Ayeixa MCP Guardian: Capability Fencing & Parameter Sanitization for MCP

Model Context Protocol (MCP) empowers language models to interact directly with databases, filesystems, and CLI utilities. However, exposing tool APIs creates security vulnerabilities: unauthorized directory traversal (../../etc/passwd), destructive shell commands (rm -rf), and untracked actions.

Ayeixa MCP Guardian (@ayeixa/mcp-guardian) is a runtime capability fence and parameter sanitization middleware for Model Context Protocol servers and client tool invocations.

GitHub Repository: https://github.com/alpallovy/ayeixa-mcp-guardian
Status: Pre-Release (v0.1.0-alpha)
License: Apache-2.0
NPM Status: Public registry publication is pending; evaluate and build locally.

1. Core Architecture

MCP Guardian operates as an interceptor middleware:

ToolPermissionFence: Enforces granular Role-Based Access Control (RBAC) and explicit tool allowlists/denylists.
InvocationSanitizer: Sanitizes parameter arguments, blocking directory traversal patterns (../) and dangerous system commands (sudo, rm -rf, chmod 777).
RuntimeAuditLogger: Records every invocation payload, argument set, and permission verdict into a tamper-evident SHA-256 cryptographic hash chain.
GuardianSandbox: Permission-gated execution wrapper evaluating authorization rules and logging audit records before dispatching to tool executors.

2. Implemented Capabilities & Test Verification

Verified with hermetic unit tests:

Permission Fencing: RBAC and denied command enforcement (tests/permission.test.ts).
Audit Ledger: Cryptographic hash chain validation (tests/audit.test.ts).
Execution Wrapper: Permission-gated dispatch and error containment (tests/sandbox.test.ts).

Verification: 6/6 hermetic unit tests passing (0 failures).

3. Local Quick Start

git clone https://github.com/alpallovy/ayeixa-mcp-guardian.git
cd ayeixa-mcp-guardian
npm ci
npm run build
npm test

Usage Example

import { GuardianSandbox, ToolPermissionFence, RuntimeAuditLogger } from './src';

const fence = new ToolPermissionFence({
  allowedTools: ['read_file', 'list_dir'],
  blockedCommands: ['rm -rf', 'sudo', 'chmod 777']
});
const audit = new RuntimeAuditLogger();
const sandbox = new GuardianSandbox(fence, audit);

const verdict = await sandbox.evaluateAndExecute({
  toolName: 'read_file',
  arguments: { path: './src/index.ts' },
  role: 'developer'
}, async (args) => {
  return "File content safely read.";
});

console.log("Allowed:", verdict.allowed);

4. Limitations & Contributing

Pre-release v0.1.0-alpha. Provides application-level permission gating and parameter sanitization.
Public npm publication is pending.
We encourage security developers and researchers to contribute. See open good first issue tags on GitHub.

License: Apache-2.0

Introducing Ayeixa Sentience Eval: Catching Raw JSON Leaks & Robotic Clichés in Agent Dialogue

Alp Allovi — Mon, 17 Aug 2026 19:55:15 +0000

Introducing Ayeixa Sentience Eval: Catching Raw JSON Leaks & Robotic Clichés in Agent Dialogue

When autonomous agents convert structured internal states and tool execution payloads into natural language dialogue, formatting breakdowns can occur: unparsed JSON delimiters leak into output text, robotic conversational clichés degrade user experience, and assertions deviate from context.

Ayeixa Sentience Eval (@ayeixa/sentience-eval) is a standalone assertion harness for testing, benchmarking, and guarding behavioral cleanliness in autonomous conversational agents.

GitHub Repository: https://github.com/alpallovy/ayeixa-sentience-eval
Status: Pre-Release (v0.1.0-alpha)
License: MIT
NPM Status: Public registry publication is pending; evaluate and build locally.

1. Core Architecture

Sentience Eval is composed of four inspection engines:

RawJsonLeakDetector: Identifies raw JSON structures, unbalanced brace delimiters, and unrendered system delimiters matching configured regex patterns.
HallucinationDetector: Compares response claims against ground-truth context to flag ungrounded assertions.
BehavioralCleanlinessHarness: Enforces conversational naturalness rules by detecting forbidden robotic clichés, prompt leaks, and meta-tokens.
SentienceBenchmarkRunner: Executes evaluation batches and computes a weighted cleanliness index (0–100).

2. Implemented Capabilities & Test Verification

Verified with hermetic unit tests:

Raw JSON Detection: Detects tested unescaped JSON state and delimiter leaks (tests/json-leak.test.ts).
Cleanliness Suite: Flags tested cliché markers and delimiter imbalances (tests/cleanliness.test.ts).
Benchmark Runner: Weighted penalty scoring calculation (tests/benchmark.test.ts).

Verification: 6/6 hermetic unit tests passing (0 failures).

3. Local Quick Start

git clone https://github.com/alpallovy/ayeixa-sentience-eval.git
cd ayeixa-sentience-eval
npm ci
npm run build
npm test

Usage Example

import { SentienceBenchmarkRunner } from './src';

const runner = new SentienceBenchmarkRunner();
const evaluation = runner.evaluate({
  output: "Here is your task: {\"internalState\": true, \"status\": \"RUNNING\"}",
  groundTruthContext: ["The user requested task status."]
});

console.log("Cleanliness Score (0-100):", evaluation.cleanlinessScore);
console.log("Violations Detected:", evaluation.violations);

4. Limitations & Contributing

Pre-release v0.1.0-alpha. Focuses on structural and lexical patterns covered by its rule set.
Public npm publication is pending.
Contributions welcome! Check out good first issue tags on GitHub.

License: MIT

Introducing Ayeixa FuelLite: Free-Tier Token Arbitrage & Dynamic Failover Router

Alp Allovi — Mon, 17 Aug 2026 19:54:27 +0000

Introducing Ayeixa FuelLite: Free-Tier Token Arbitrage & Dynamic Failover Router

Continuous autonomous agent execution loops consume substantial token volumes. In multi-tenant environments, unmonitored model calls quickly trigger rate limits (HTTP 429) or excessive API billing.

Ayeixa FuelLite (@ayeixa/fuellite) is an open-source TypeScript library for zero-cost multi-provider token arbitrage, quota tracking, and resilient failover.

GitHub Repository: https://github.com/alpallovy/ayeixa-fuellite
Status: Pre-Release (v0.1.0-alpha)
License: MIT
NPM Status: Public registry publication is pending; evaluate and build locally.

1. Core Architecture

FuelLite provides four synchronized modules:

CostEstimator: Deterministic numeric cost estimation using token counts and configured provider rates, with reported USD values rounded to six decimal places.
ProviderArbitrageRouter: Ranks configured model providers, prioritizing providers configured as free-tier (isFreeTier: true) before falling back to lower-cost paid endpoints.
TokenBudgetTracker: Tracks cumulative tenant and session token budgets with hard-stop limits and threshold warning alarms.
FailoverExecutor: Transparently handles provider outages and rate limits with exponential backoff ($2^n \times \text{base}$) and cooldown recovery.

2. Implemented Capabilities & Test Verification

Verified with hermetic unit tests:

Cost Estimation: Six-decimal USD calculation across token counts (tests/cost.test.ts).
Arbitrage Routing: Prioritizing configured free-tier providers and lowest cost (tests/arbitrage.test.ts).
Budget Tracking: Hard quota enforcement and spend recording (tests/budget.test.ts).
Failover Logic: HTTP 429/503 retry backoff and provider hopping (tests/failover.test.ts).

Verification: 6/6 hermetic unit tests passing (0 failures).

3. Local Quick Start

git clone https://github.com/alpallovy/ayeixa-fuellite.git
cd ayeixa-fuellite
npm ci
npm run build
npm test

Usage Example

import { CostEstimator, ProviderArbitrageRouter, TokenBudgetTracker } from './src';

const estimator = new CostEstimator();
const cost = estimator.calculateCost({
  provider: 'gemini-flash',
  inputTokens: 1500,
  outputTokens: 400,
  isFreeTier: true
});
console.log("Total Cost (USD):", cost.totalCostUsd); // 0.0

const tracker = new TokenBudgetTracker({ maxBudgetUsd: 5.0 });
tracker.recordSpend("tenant-alpha", cost.totalCostUsd, 1900);
console.log("Remaining Budget:", tracker.getRemainingBudget("tenant-alpha"));

4. Limitations & Contributing

Pre-release v0.1.0-alpha. Provider rates are statically configured.
Public npm publication is pending.
Feedback and PRs welcome! See open good first issue items on GitHub.

License: MIT

Introducing Ayeixa OpenCoordinator: Multi-Agent DAG Orchestration for TypeScript

Alp Allovi — Mon, 17 Aug 2026 19:54:22 +0000

Introducing Ayeixa OpenCoordinator: Multi-Agent DAG Orchestration for TypeScript

As multi-agent engineering workflows scale, managing dependent subagents requires deterministic dependency resolution, cycle prevention, and capability-aware task routing.

Ayeixa OpenCoordinator (@ayeixa/open-coordinator) is a lightweight multi-agent orchestration spine developed in TypeScript under the Apache-2.0 license.

GitHub Repository: https://github.com/alpallovy/ayeixa-open-coordinator
Status: Pre-Release (v0.1.0-alpha)
NPM Status: Public registry publication is pending; evaluate and build locally.

1. Core Architecture

OpenCoordinator is structured around four decoupled architectural primitives:

IntentParser: Extracts domain context, action classifications (CODE, TEST, DEPLOY, RESEARCH), and capability constraints from natural language task directives.
TaskGraphEngine: Constructs an in-memory Directed Acyclic Graph (DAG), executes topological sort, detects cycles deterministically, and groups independent tasks into concurrent execution stages.
ExecutionKernel: Implements a state machine managing task lifecycles (PENDING, RUNNING, COMPLETED, FAILED, BLOCKED) with bounded transitions.
AgentRouter: Matches tasks to subagents based on declared capabilities and current availability.

2. Implemented Capabilities & Test Verification

All capabilities are implemented and verified with 100% hermetic unit tests:

DAG Resolution: Topological sorting and execution stage resolution (tests/dag.test.ts).
Intent Parsing: Action and entity extraction (tests/intent.test.ts).
Router Matching: Capability scoring (tests/router.test.ts).
Kernel State: State transitions and error containment (tests/kernel.test.ts).

Verification: 9/9 hermetic unit tests passing (0 failures).

3. Local Quick Start

git clone https://github.com/alpallovy/ayeixa-open-coordinator.git
cd ayeixa-open-coordinator
npm ci
npm run build
npm test

Usage Example

import { IntentParser, TaskGraphEngine, ExecutionKernel } from './src';

const parser = new IntentParser();
const intent = parser.parse("Refactor auth database module to support SQLite");

const dag = new TaskGraphEngine();
dag.addTask({ id: "task-1", title: "Refactor auth db", taskType: intent.taskType, dependencies: [] });
dag.addTask({ id: "task-2", title: "Add auth tests", taskType: "TEST", dependencies: ["task-1"] });

const stages = dag.resolveExecutionStages();
console.log("Execution stages:", stages);

4. Limitations & Contributing

Current version is v0.1.0-alpha. Graph storage operates in-memory.
Public npm publication is pending.
Feedback, issues, and contributions are warmly welcomed! Check out good first issue tags on GitHub.

License: Apache-2.0