LLM Context (`llms.txt` / `llms-full.txt`)¶

The project ships two plain-text files designed to be fed into an AI coding assistant (Claude, Cursor, Copilot, ChatGPT, …) so that it has accurate, up-to-date context about fastapi-jsonrpc without having to crawl the whole site:

llms.txt — short index. Title, one-paragraph description, and a curated list of links to every documentation page.
llms-full.txt — full reference. Architecture, public API signatures, built-in errors, patterns, testing harness, wire protocol, common pitfalls.

Both files follow the llmstxt.org convention: Markdown, predictable top-level structure, no HTML chrome, no navigation noise. They are served from the documentation site root, so once the docs are deployed they are reachable at stable URLs:

https://smagafurov.github.io/fastapi-jsonrpc/llms.txt
https://smagafurov.github.io/fastapi-jsonrpc/llms-full.txt

When to use which¶

File	Size	Use for
`llms.txt`	small	Navigation hint — let the agent discover which page to fetch next.
`llms-full.txt`	large	Drop-in context for answering a concrete question without any further fetches.

If you are unsure, paste llms-full.txt — it is self-contained.

Usage patterns¶

Claude Code / Claude Desktop¶

Reference the file directly with @ in your prompt, or save it into CLAUDE.md at the root of your project so every session picks it up:

# CLAUDE.md

When working with fastapi-jsonrpc, use the patterns documented at
https://smagafurov.github.io/fastapi-jsonrpc/llms-full.txt

Cursor / Windsurf / Continue¶

Add the URL as a custom documentation source. Most IDE integrations support @Docs or a similar mechanism that fetches and indexes external pages — point it at llms-full.txt.

Raw prompt (ChatGPT, Gemini, local models)¶

Fetch the file once and paste it at the top of your conversation under a # Reference heading:

# Reference
<paste contents of llms-full.txt>

# Task
Implement a JSON-RPC method `transfer(src, dst, amount)` with a typed
`NotEnoughMoney` error using fastapi-jsonrpc.

Programmatic fetch (CI, scripts, agents)¶

import httpx

CTX = httpx.get(
    'https://smagafurov.github.io/fastapi-jsonrpc/llms-full.txt'
).text

What's inside `llms-full.txt`¶

At a glance — the file is structured so an LLM can jump straight to the relevant section:

Architecture — module hierarchy, data flow, key abstractions (API, Entrypoint, MethodRoute, JsonRpcContext).
Public API — signatures of API, Entrypoint, BaseError, context helpers.
Built-in errors — table of JSON-RPC error codes shipped with the library.
Key patterns — declaring methods, dependency injection, per-call vs batch dependencies, middlewares, typed error data.
OpenAPI / OpenRPC — auto-generated schema, per-method Swagger routes, tags, component names.
Testing — pytest plugin (JsonRpcTestClient, jsonrpc_client, auto-validation fixture, opt-out marker), direct method calls.
Sentry integration — FastApiJsonRPCIntegration, transaction naming, batch spans.
Wire protocol — single / batch / notification request and response examples.
Common pitfalls — issues that bite integrators in production.

Keeping the files fresh¶

Both files are hand-written Markdown committed to the repo under docs/. When the library changes — new public API, new built-in error, new fixture — update the corresponding section so that LLMs using the published files see the current reality. Outdated AI context is strictly worse than no context at all: it produces confident wrong answers.

LLM Context (llms.txt / llms-full.txt)¶