@planckspace/core

Shared types, constants, and utilities consumed by the backend, CLI, extension, and frontend.

Exports

Types

TelemetryPayload

Session metadata captured by the scraper and extension. Contains only metadata — never prompt or response text.

| Field | Type | Description | |---|---|---| | sessionId | string | Unique session identifier | | tool | "claude_code" \| "cursor" \| "windsurf" \| "antigravity" | AI coding tool used | | model | string | Model name (e.g. claude-sonnet-4-6) | | startedAt / endedAt | string | ISO 8601 timestamps | | inputTokens / outputTokens | number | Token counts | | cacheReadTokens / cacheWriteTokens | number | Prompt cache token counts | | costUsd | number | Computed dollar cost | | repoName | string | Repository name | | gitBranch | string \| null | Current branch | | gitAuthorEmail | string \| null | Git author email | | outcome | "shipped" \| "abandoned" \| "errored" \| "partial" \| "unknown" | Session outcome | | turnCount | number | Number of conversation turns | | filesTouchedCount | number | Number of files modified | | source | "scraper" \| "extension" | Data collection source | | workspaceId | string | Planckspace workspace ID |

PlanckConfig

Schema for ~/.planckspace/config.json, shared by the CLI and extension.

{
  apiUrl: string;
  token: string | null;
  workspaceId: string | null;
  workspaceName: string | null;
  pausedWorkspaces: string[];
}

SkillRecRequest / SkillRecResponse

Request/response types for the skill recommendation API.

type SkillRecRequest = {
  workspaceId: string;
  detectedStack: string[];
  recentActivityHints: string[];
};

type SkillRecResponse = {
  recommendations: Array<{
    skillName: string;
    reason: string;
    estimatedMonthlySavingsUsd: number;
  }>;
};

Constants

MODEL_PRICING

Per-million-token pricing keyed by model name. Fields: input, output, cacheRead, cacheWrite.

Note: Verify current pricing at docs.claude.com before shipping.

Functions

computeCostUsd(usage, model): number

Computes dollar cost from token counts using MODEL_PRICING. Returns 0 for unknown models.

import { computeCostUsd } from "@planckspace/core";

const cost = computeCostUsd(
  { inputTokens: 10_000, outputTokens: 2_000, cacheReadTokens: 5_000, cacheWriteTokens: 0 },
  "claude-sonnet-4-6"
);

Development

npm install
npm run build   # compiles to dist/
npm test        # runs unit tests

Publishing

npm publish     # runs prepublishOnly (build) then publishes to npm

The package is scoped to @planckspace and published with "access": "public". Ensure you are authenticated (npm login) and have publish rights to the planckspace-dev org before running npm publish.