@dtoolkit/core

Shared types and schemas for the dtoolkit ecosystem

Foundation package for the dtoolkit ecosystem. Exports TypeScript types, Zod schemas, and shared utilities used by all other @dtoolkit/* packages.

Install

npm install @dtoolkit/core

What's inside

Adapter types

The universal adapter contract used by all @dtoolkit/adapter-* packages:

Export	Description
`Adapter`	Interface: `{ provider, execute(request), stream?(request) }`
`AdapterRequest`	Request shape: prompt, model, maxTurns, sessionId, options
`AdapterResult`	Result shape: text, sessionId, costUsd, durationMs, usage, model, raw
`AdapterUsage`	Token counts: inputTokens, outputTokens, totalTokens
`AdapterStreamEvent`	Stream event union: text delta, tool use, result, error

import type { Adapter, AdapterRequest, AdapterResult } from "@dtoolkit/core";

Data types and schemas

Export	Description
`Entity` / `EntitySchema`	Entity type with PARA category and Zod schema
`EntityType` / `ParaCategory`	Enums for entity classification
`Fact` / `FactSchema`	Fact type with tier config and `computeTier()`
`Tier` / `TierConfig`	Memory tier logic: HOT, WARM, COLD
`ContextBlock` / `ContextBlockSchema`	Context injection block schema
`HistoryEntry` / `SessionInfo`	Prompt history and session types
`TemplateDefinition` / `TemplateVariable`	YAML prompt template types

import { EntitySchema, computeTier, Tier } from "@dtoolkit/core";

const entity = EntitySchema.parse(data);
const tier = computeTier(fact, { hotDays: 7, hotMinAccess: 10, warmDays: 30 });
// Tier.HOT | Tier.WARM | Tier.COLD

Telemetry types

Types used by @dtoolkit/dops and the adapter packages for telemetry extraction:

Export	Description
`Target`	Hook target interface for CLI integrations
`TelemetryExtractor`	Interface for extracting sessions from CLI transcripts
`ParsedSession` / `ParsedTokenUsage` / `ParsedToolCall`	Parsed telemetry data shapes
`TranscriptEntry`	Raw transcript entry type
`IngestState`	State tracking for incremental ingestion

Streaming utilities

Export	Description
`LineBuffer`	Newline-delimited buffer for JSONL stream parsing

Dcontext helpers

Shared helpers for reading/writing the dcontext section in instruction files (CLAUDE.md, AGENTS.md, etc.):

Export	Description
`writeDcontextMdSection(content, section)`	Write/replace a dcontext section in a markdown file
`removeDcontextMdSection(content)`	Remove the dcontext section from a markdown file
`DCONTEXT_MARKER` / `DBRAIN_START` / `DBRAIN_END`	Section delimiters

File utilities

Export	Description
`isTextFile(path)`	Check if a file is text (not binary)
`embedTextFiles(files)`	Embed text files into a prompt-friendly format
`detectMimeType(path)`	Detect MIME type from extension
`InputFile`	File attachment type for adapter requests

Other utilities

Export	Description
`diceCoefficient(a, b)`	Dice coefficient string similarity (0–1)

License

MIT