Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/coeusyk/inference-xWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/rules/coeusyk/inference-x/backend-python)<a href="https://agentmods.dev/rules/coeusyk/inference-x/backend-python"><img src="https://agentmods.dev/badge/rules/coeusyk/inference-x/backend-python/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/rules/coeusyk/inference-x/backend-python"><img src="https://agentmods.dev/badge/rules/coeusyk/inference-x/backend-python.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00424 | $0.00424 |
| Opus 5.5 | $0.00170 | $0.00170 |
| Sonnet 5.5 | $0.00085 | $0.00085 |
| Haiku 4.5 | $0.00042 | $0.00042 |
Grade A, and why
backend-python scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
Backend Python rules
Stack assumptions
- Python 3.11+
- FastAPI for HTTP APIs
- Pydantic for schemas
- vLLM as the first engine implementation
- Development on Windows via WSL2 Ubuntu
Code style
- Use type hints on all public functions, methods, and class attributes where practical.
- Prefer small classes and focused functions.
- Keep route handlers limited to validation, dependency resolution, and response formatting.
- Put orchestration in services and engine-specific logic in engine modules.
- Prefer explicit imports over wildcard imports.
- Raise domain-specific exceptions or HTTP errors deliberately; do not swallow exceptions.
FastAPI conventions
- Define routes under
src/inference_x/api/routes/. - Define shared dependencies in
src/inference_x/api/deps.py. - Define error mapping in
src/inference_x/api/errors.py. - Keep endpoint contracts stable once introduced.
- Prefer versioned endpoints under
/v1.
Schema conventions
- Put request and response models in
src/inference_x/schemas/. - Separate transport schemas from engine internals when complexity grows.
- Keep field names aligned with OpenAI-compatible formats where intended.
- Validate config-derived values before use.
Engine conventions
- Define a base engine interface in
src/inference_x/engines/base.py. - Implement vLLM-specific behavior in
src/inference_x/engines/vllm_engine.py. - Keep prompt formatting and request adaptation outside route handlers.
- Expose health-check capability from each engine implementation.
Testing conventions
- Unit tests go in
tests/unit/. - Integration tests go in
tests/integration/. - Contract tests go in
tests/contract/. - Mock vLLM where possible in unit tests; reserve real-engine execution for explicit integration tests.
Operational rules
- Read model and server settings from YAML or environment-backed settings.
- Do not hardcode model identifiers, ports, file paths, or GPU values in core logic.
- Log structured events for engine startup, request handling, latency, and failures.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 53 lines · 424 tokens per session scan A 256822926a3e
backend-python is a cursor rule published in the GitHub repository coeusyk/inference-x (2 stars, last pushed 18d ago), licensed MIT. It adds 424 tokens to every session, about $0.0017 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other cursor rules, from other repositories
go-api-handlers
Use authorization functions for apps.
fastapi-pydantic
You are an expert in Python, FastAPI, and scalable API development.
python
Python best practices and patterns for modern software development with Flask and SQLite.
aiohttp
This guide defines definitive best practices for using aiohttp, focusing on efficient, robust, and maintainable asynchronous HTTP client code.
django
Definitive guidelines for writing maintainable, performant, and secure Django applications, emphasizing modern best practices, clear code organization, and efficient patterns.
02-python-standards
Python coding standards, libraries, and concurrency safety.