L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Développement · SKILL.md

sdk

sdk

Use Headroom compression APIs, SharedContext, relevance and image helpers, and the TypeScript SDK in application code.

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: sdk description: "Use Headroom compression APIs, SharedContext, relevance and image helpers, and the TypeScript SDK in application code." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

Headroom SDK

Use this sub-skill when the user wants to call Headroom from Python or TypeScript code, embed compression in an application, or understand the SDK-level APIs that do not require operating the proxy or durable memory workflows.

Route here for

  • Python direct compression: compress, compress_spreadsheet, CompressConfig, CompressResult, HeadroomClient, ChatCompletions, Messages, and validate_setup/metrics helpers.
  • SharedContext for compressed inter-agent handoffs.
  • Relevance and image helpers: headroom.relevance, headroom.image, and optional extras that affect those APIs.
  • TypeScript SDK usage from sdk/typescript/, including compress, HeadroomClient, simulate, retrieve, SharedContext, hooks, format conversion, and path helpers.
  • App/framework recipes: OpenAI, Anthropic, LiteLLM, Vercel AI SDK, LangChain, MCP tool outputs, spreadsheet/tabular compression, and other local integration patterns.

Route elsewhere

  • Proxy base URL, headroom proxy, and agent wrapping/setup: use ../proxy-wrap/SKILL.md.
  • Memory persistence, MCP registration, learn, and recovery: use ../memory/SKILL.md.
  • Install/update/deployment lifecycle and broad operator diagnostics: use ../ops/SKILL.md.

Fast start

Python local compression:

from headroom import compress

result = compress([
    {"role": "user", "content": "Explain this file"},
], optimize=False)

Python configuration or app wrapper:

from headroom import CompressConfig
from headroom.client import HeadroomClient

TypeScript SDK:

import { compress, HeadroomClient, SharedContext } from "headroom-ai";

For a safe smoke check, read the bundled helper and run the small no-network validation it documents:

python scripts/compress_smoke.py --json

References and helpers

  • references/api-reference.md captures the verified Python API shapes and important defaults.
  • references/typescript-sdk.md summarizes the TS exports and app-facing examples.
  • references/workflows.md gives concrete Python and TypeScript application recipes.
  • references/troubleshooting.md covers missing extras, proxy assumptions, ONNX/HF/image issues, and type/runtime mismatches.
  • scripts/compress_smoke.py runs a safe no-network Python compression smoke.
  • scripts/tabular_compression_demo.py is the bundled table/spreadsheet demo adapted from the repo example.

Safety rules

  • Do not promise live proxy behavior when optimize=False or when the user has not configured a running proxy.
  • Avoid claiming GPU or cloud behavior unless the user explicitly selected the required extras or backend and the environment was verified.
  • Do not put private environment paths, tokens, or provider credentials in examples.
  • If a feature depends on an optional extra, say so explicitly and show the smallest import path or smoke check that proves it.
Skills enregistrées