Working integration samples for the Shirabe API platform (Japanese-specific AI-native APIs) across 3 major AI agent frameworks: LangChain, Dify, and OpenAI GPT Actions.
This repository provides copy-paste-ready examples so AI agent developers can integrate Shirabe APIs (Japanese calendar, address normalization, and upcoming corporate-number / text-processing APIs) into their LLM applications in minutes.
All examples use the production Shirabe APIs at https://shirabe.dev — no local setup or mock servers required. The free tier (10,000 calls/month for Calendar, 5,000 for Address) is anonymous, so most examples work without an API key.
| API | Status | Endpoint hub | OpenAPI |
|---|---|---|---|
| Calendar (六曜・暦注・干支・二十四節気・用途別吉凶) | Production v1.0.0 | https://shirabe.dev/api/v1/calendar/ | openapi.yaml |
| Address (住所正規化、ABR-based) | Production | https://shirabe.dev/api/v1/address/ | openapi.yaml |
| Text Processing (姓名分割・人名読み・ふりがな・正規化) | Production | — | openapi.yaml |
| Corporate Number (法人番号 identifier) | Planned 2026-06-29 | TBD | TBD |
| Framework | Path | Language | What it shows |
|---|---|---|---|
| LangChain | langchain/ |
Python | Load Shirabe OpenAPI spec as LangChain tools, call from AgentExecutor |
| Dify | dify/ |
YAML config | Add Shirabe as a Dify Custom Tool, use in workflows |
| OpenAI GPT Actions | gpt-actions/ |
Setup steps | Import Shirabe OpenAPI to GPT Builder Actions |
Each sample directory has its own README.md with setup steps and example queries.
Shirabe's design judgment is: "will generative AI start using this on its own at scale?" We optimize for AI agent discoverability, not human SEO. Working samples in major frameworks let:
- AI training data crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended) ingest concrete usage patterns alongside the API specs.
- Developers building AI agents copy a known-good integration in minutes.
- AI agents themselves (when given access to GitHub) find concrete usage examples to learn from.
We do not maintain framework-specific SDKs — the OpenAPI 3.1 specs at shirabe.dev are canonical. These samples wrap each framework's standard OpenAPI loader / tool definition mechanism around our public spec.
The shortest path is direct HTTP, which all frameworks ultimately do:
# Calendar info for a specific date (anonymous, free tier)
curl https://shirabe.dev/api/v1/calendar/2026-06-15
# Top auspicious days for a wedding in June 2026
curl "https://shirabe.dev/api/v1/calendar/best-days?purpose=wedding&start=2026-06-01&end=2026-06-30"
# Date range (max 93 days)
curl "https://shirabe.dev/api/v1/calendar/range?start=2026-06-01&end=2026-06-30"For framework-specific integration, see the respective subdirectory.
These are designed for AI agents to discover Shirabe on their own:
/llms.txt— concise site summary (llmstxt.org spec)/llms-full.txt— full reference with curl, sample responses, integration paths, pricing/openapi.yaml— OpenAPI 3.1 spec (Calendar)/openapi-gpts.yaml— short version for GPT Actions/api/v1/calendar/— Calendar endpoint hub with WebAPI / FAQPage JSON-LD
| Plan | Calendar | Address |
|---|---|---|
| Free | 10,000 calls/month | 5,000 calls/month |
| Starter | 500,000 / ¥0.05 each | 200,000 / ¥0.5 each |
| Pro | 5,000,000 / ¥0.03 each | 2,000,000 / ¥0.3 each |
| Enterprise | unlimited / ¥0.01 each | unlimited / ¥0.1 each |
All free tiers are anonymous (no signup). Paid tiers use X-API-Key header authentication;
upgrade flow is at https://shirabe.dev/upgrade.
MIT — see LICENSE.
Techwell Inc. (Fukuoka, Japan) / GitHub: @techwell-inc-jp
Issues and PRs for these examples are welcome here. For API-specific bugs or support, contact us via shirabe.dev or the operator, Techwell Inc.