Skip to content

AGENTS.md

Latest commit

 

History

History
159 lines (110 loc) · 7.01 KB

AGENTS.md

File metadata and controls

159 lines (110 loc) · 7.01 KB

AGENTS.md

Repository Expectations

  • use ripgrep (through rg command) in place of grep.
  • Document public utilities in docs/ when you change behavior.
  • Whenever you import typing, use: import typing as t
  • Use as much typing as possible on your function definitions
  • Use python native type hinting whenever possible, e.g. list[dict]
  • Always use named-arguments when calling functions or methods.
  • Document function and methods using numpy-style docs.
  • To install libraries to the project, use uv add

Development workflows

  • When asking me questions in Plan Mode, give me extensive information to make an informed choice: define the impact of each invidual choice and give context about the question before asking it.

  • Run tests (using pytest) for every code change but not when changing code comments or documentation-related stuff. If the pytest command does not work due to missing imports, try activating the environment first with source .venv/bin/activate

  • Run pre-commits (using prek run -a) for every code change including code comments or documentation.

pre-commits to check for:

  • pre-commit-hooks/ruff-check/ruff-format: syntax/style related, they autofix most of the time
  • markdownlint-cli: syntax/style related, does not autofix.
  • ty-check: type hinting, does not autofix
  • bandit/detect-secrets: security-related, does not autofix but can have false flags.

  • When updating static docs, always rebuild in strict mode after your changes. In that case, no need to run tests with pytest.

  • When pursuing a complex task, break it down as simpler tasks and make atomic commits to facilitate code review. Do the atomic commits yourself using git commit -m. When the task is done, include atomic commit names in your recap to streamline your approach. Always follow good practices for atomic commits.

Component index

End-to-end flow (high level)

  1. Callers wrap a function or client instance with batchify, which creates a Batcher and installs hooks.
  2. BatchingContext activates a context variable holding the Batcher when its async context is entered.
  3. HTTP hooks intercept supported requests and enqueue them into the Batcher.
  4. The Batcher batches pending requests, submits them, and resolves per-request futures.
  5. Provider adapters normalize URLs and decode batch API results back into HTTP responses.

See the component pages above for detailed behavior and extension points.

Overall Coding Principles

Whenever asked to do something, strictly follow these principles in your actions.

1. Think Before Coding

Don't assume. Don't hide confusion. Surface tradeoffs.

Before implementing:

  • State your assumptions explicitly. If uncertain, ask.
  • If multiple interpretations exist, present them - don't pick silently.
  • If a simpler approach exists, say so. Push back when warranted.
  • If something is unclear, stop. Name what's confusing. Ask.

2. Simplicity First

Minimum code that solves the problem. Nothing speculative.

  • No features beyond what was asked.
  • No abstractions for single-use code.
  • No "flexibility" or "configurability" that wasn't requested.
  • No error handling for impossible scenarios.
  • If you write 200 lines and it could be 50, rewrite it.

Ask yourself: "Would a senior engineer say this is overcomplicated?" If yes, simplify.

Add later if needed: Caching (when performance matters), validation (when bad data appears), merging (when requirement emerges).

3. Surgical Changes

Touch only what you must. Clean up only your own mess.

When editing existing code:

  • Don't "improve" adjacent code, comments, or formatting.
  • Don't refactor things that aren't broken.
  • Match existing style, even if you'd do it differently.
  • If you notice unrelated dead code, mention it - don't delete it.

When your changes create orphans:

  • Remove imports/variables/functions that YOUR changes made unused.
  • Don't remove pre-existing dead code unless asked.

The test: Every changed line should trace directly to the user's request.

4. Goal-Driven Execution

Define success criteria. Loop until verified.

Transform tasks into verifiable goals:

  • "Add validation" → "Write tests for invalid inputs, then make them pass"
  • "Fix the bug" → "Write a test that reproduces it, then make it pass"
  • "Refactor X" → "Ensure tests pass before and after"

For multi-step tasks, state a brief plan:

1. [Step] → verify: [check]
2. [Step] → verify: [check]
3. [Step] → verify: [check]

Strong success criteria let you loop independently. Weak criteria ("make it work") require constant cl

Here is real-world code examples demonstrating the four principles. Each example shows what LLMs commonly do wrong and how to fix it.

Anti-Patterns Summary

Principle Anti-Pattern Fix
Think Before Coding Silently assumes file format, fields, scope List assumptions explicitly, ask for clarification
implicity First Strategy pattern for single discount calculation One function until complexity is actually needed
Surgical Changes Reformats quotes, adds type hints while fixing bug Only change lines that fix the reported issue
Goal-Driven "I'll review and improve the code" "Write test for bug X → make it pass → verify no regressions"

Key Insight

The "overcomplicated" examples aren't obviously wrong—they follow design patterns and best practices. The problem iming**: they add complexity before it's needed, which:

  • Makes code harder to understand
  • Introduces more bugs
  • Takes longer to implement
  • Harder to test

The "simple" versions are:

  • Easier to understand
  • Faster to implement
  • Easier to test
  • Can be refactored later when complexity is actually needed

Good code is code that solves today's problem simply, not tomorrow's problem prematurely.