API.md

LNMP SDK API Reference

This document provides a detailed reference for the @lnmp/lnmp TypeScript SDK.

Table of Contents

1. Core & Codec
   • Parser
   • Encoder
   • RecordBuilder
2. Envelope
3. Network
4. Embedding
5. Spatial
6. Sanitization
7. LLB (Explain Mode)
8. SFE (Context Scoring)

---

Core & Codec

Parser

Parses LNMP text format into LnmpRecord objects.

import { Parser } from '@lnmp/lnmp';

const parser = new Parser('F12:i=14532;F7:b=1');
const record = parser.parseRecord();

Encoder

Encodes LnmpRecord objects into canonical LNMP text format.

import { Encoder } from '@lnmp/lnmp';

const encoder = new Encoder();
const text = encoder.encode(record);

RecordBuilder

Fluent API for constructing LnmpRecords.

import { RecordBuilder } from '@lnmp/lnmp';

const record = new RecordBuilder()
    .addField({ fid: 12, value: { type: 'Int', value: 14532 } })
    .addField({ fid: 7, value: { type: 'Bool', value: true } })
    .build();

---

Envelope

Wraps records with operational metadata (timestamp, source, trace ID).

import { Envelope } from '@lnmp/lnmp';

// Wrap a record
const envelope = Envelope.wrap(record, {
    source: 'my-service',
    trace_id: 'abc-123',
    labels: { env: 'prod' }
});

// Unwrap
const innerRecord = Envelope.unwrap(envelope);

// HTTP Headers Interop
const headers = Envelope.toHeaders(envelope);
const restoredMetadata = Envelope.fromHeaders(headers);

---

Network

AI-driven routing and importance scoring.

import { Network } from '@lnmp/lnmp';

const msg = {
    envelope: envelope,
    kind: 'Event',
    priority: 128,
    ttl_ms: 5000
};

// Calculate Importance (0.0 - 1.0)
const score = Network.importance(msg);

// Decide Routing
// Returns: "SendToLLM" | "ProcessLocally" | "Drop"
const decision = Network.decide(msg);

---

Embedding

Vector operations, delta compression, and quantization.

import { Embedding, QuantScheme } from '@lnmp/lnmp';

// Delta Compression
const delta = Embedding.computeDelta(vec1, vec2);
const reconstructed = Embedding.applyDelta(vec1, delta);

// Quantization
// Schemes: QInt8, QInt4, Binary, FP16
const quantized = Embedding.quantize(vec1, QuantScheme.QInt8);
const restored = Embedding.dequantize(quantized);

---

Spatial

Efficient encoding/decoding of spatial frames for hybrid reality.

import { Spatial, FrameMode } from '@lnmp/lnmp';

const frame = {
    header: {
        mode: FrameMode.Absolute,
        sequence_id: 1,
        timestamp: BigInt(Date.now()) * 1000000n, // Nanoseconds (BigInt)
        checksum: 0
    },
    payload: { /* SpatialState */ }
};

const bytes = Spatial.encodeFrame(frame);
const decoded = Spatial.decodeFrame(bytes);

---

Sanitization

Cleans untrusted input to ensure safe parsing.

import { Sanitizer } from '@lnmp/lnmp';

const cleanText = Sanitizer.sanitize('  "Unsafe Input"  ');

---

LLB (Explain Mode)

Generates human-readable explanations for LNMP records using a semantic dictionary.

import { ExplainEncoder } from '@lnmp/lnmp';

const dictionary = {
    12: "user_id",
    7: "is_active"
};

const explanation = ExplainEncoder.encode(record, dictionary);
// Output:
// F7:b=1              # is_active
// F12:i=14532         # user_id

---

SFE (Context Scoring)

Scores contexts (envelopes) to help LLMs prioritize data for RAG or prompt construction.

import { ContextScorer } from '@lnmp/lnmp';

const profile = ContextScorer.score(envelope);

console.log(profile.freshness_score); // 0.0 (stale) to 1.0 (fresh)
console.log(profile.risk_level);      // "Low", "Medium", "High", "Critical"
console.log(profile.importance);      // 0-255
console.log(profile.confidence);      // 0.0 - 1.0