@kibhq/core

Core engine for kib — the headless knowledge compiler.

This package provides the vault operations, LLM providers, ingest extractors, compile engine, BM25 search, and RAG query engine that power the kib CLI.

Install

npm i @kibhq/core

Most users should install the CLI instead: npm i -g @kibhq/cli

What's inside

| Module | Description | |--------|-------------| | Vault | Filesystem operations, manifest tracking, config management, image asset storage | | Ingest | Extractors for web pages, PDFs, YouTube, GitHub repos, images (vision), and local files | | Compile | LLM-powered compilation from raw sources into structured wiki articles with image references | | Search | BM25 full-text search with English stemming | | Query | RAG engine — retrieves relevant articles and generates cited answers | | Lint | 5 health-check rules (orphan articles, broken links, stale sources, etc.) | | Skills | Skill loader and runner for extensible vault operations | | Daemon | Watch daemon primitives — FIFO queue, folder watchers, auto-compile scheduler, PID management, log rotation, system service installer (launchd/systemd) | | Providers | LLM adapters for Anthropic Claude, OpenAI, and Ollama |

Usage

import {
  resolveVaultRoot,
  loadManifest,
  loadConfig,
  SearchIndex,
  queryVault,
  ingestSource,
  compileVault,
  createProvider,
} from "@kibhq/core";

// Find the vault
const root = resolveVaultRoot();
const config = await loadConfig(root);
const manifest = await loadManifest(root);

// Search
const index = new SearchIndex();
await index.build(root, "all");
const results = index.search("attention mechanisms", { limit: 10 });

// RAG query
const provider = await createProvider(config.provider.default, config.provider.model);
const answer = await queryVault(root, "how does self-attention work?", provider);
console.log(answer.answer);

LLM Providers

| Provider | Env Variable | Default Model | |----------|-------------|---------------| | Anthropic | ANTHROPIC_API_KEY | claude-sonnet-4-6 | | OpenAI | OPENAI_API_KEY | gpt-4o | | Ollama | (auto-detect localhost:11434) | llama3 |

License

MIT