AGENTS.md · Bessouat40/RAGLight
AGENTS.md from Bessouat40/RAGLight: 154 lines covering Start here, Repository map, Non-negotiable engineering rules, TDD workflow and more.
Template by Bessouat40
Chalk Labs
Need help deploying AGENTS.md · Bessouat40/RAGLight?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Codex, Cursor, GitHub Copilot, Gemini CLI, OpenCode, Claude Code. Read the repository before you give any tool access to your data.
curl -fsSL https://raw.githubusercontent.com/Bessouat40/RAGLight/main/AGENTS.md -o AGENTS.md
Read it and adapt it to your project before use.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: documentation
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
What's in it
An AGENTS.md file: instructions coding agents (Codex, Cursor, Copilot, Gemini CLI and others) read before working in a repo.
- For
- Codex, Cursor, GitHub Copilot, Gemini CLI, OpenCode, Claude Code
- Length
- 154 lines
- File
AGENTS.md
Sections
- Start here
- Repository map
- Non-negotiable engineering rules
- TDD workflow
- Adding a feature
- Validation commands
- Definition of done
- Working safely in an existing worktree
First lines (shown under the repo's MIT license)
# RAGLight contributor and agent guide This file is the authoritative starting point for humans and AI coding agents working in this repository. Tool-specific instruction files must link here and must not duplicate these rules. RAGLight is a Python library for building retrieval-augmented generation pipelines. Version 4 keeps the core small, makes integrations optional, and exposes stable extension contracts. The v3 CLI, HTTP server, UI, and agentic runtime are not compatibility layers in the v4 tree. ## Start here Read the documents relevant to the change before editing code: | Document | Purpose | | --- | --- | | `docs/architecture.md` | Current v4 layers and dependency rules | | `docs/development.md` | Environment setup and validation commands | | `docs/dependencies.md` | Optional extras, dependency ownership, and package audit | | `docs/testing.md` | TDD workflow and test pyramid | | `docs/integrations.md` | Checklist for adding or changing a provider | | `docs/public-api.md` | Public API policy and compatibility expectations | | `docs/migration-v3-to-v4.md` | Authoritative migration map from legacy APIs | | `docs/error-handling.md` | Error ownership and actionable failures | | `docs/releasing.md` | Package and release validation | | `docs/decisions/` | Architecture decision records | Documentation must distinguish implemented v4 behavior from capabilities that are explicitly planned for a later issue. ## Repository map ```text src/raglight/ contracts/ stable structural extension protocols models/ provider-independent data models pipeline/ classic RAG orchestration integrations/ optional provider adapters tests/ existing unit and component tests tests/architecture/ executable dependency guardrails tests/integration/ cross-component tests tests/e2e/ public workflow tests ``` Do not restore removed v3 packages as compatibility shims. New optional capabilities must be designed on the boundaries in `docs/architecture.md`. ## Non-negotiable engineering rules 1. Use test-driven development for behavior changes: red, green, refactor. 2. Every behavior change has at least a unit test. 3. Every bug fix has a regression test that fails without the fix. 4. Tests run without network access, credentials, or external services unless explicitly marked as opt-in live tests. 5. Core and pipeline code do not depend on concrete integrations. 6. Core extension contracts never depend on concrete provider implementations. 7. Optional packages are imported only within their integration boundary and missing dependencies produce an actionable installation message. 8. Future delivery layers translate inputs and outputs; they do not own RAG business logic. 9. Conversation history belongs to a caller or session, never to a globally shared pipeline. 10. Public API changes require tests, documentation, and migration notes. 11. CI validates source code and never rewrites or commits it. 12. Do not mix unrelated cleanup with a focused behavior change. Do not weaken a guardrail to accommodate a new dependency violation. Change the design or record an explicit architecture decision. ## TDD workflow For each observable behavior: 1. Write the acceptance example in user-visible terms. 2. Identify the layer that owns the behavior. 3. Add the smallest test that expresses it. 4. Run the test and confirm it fails for the expected reason. 5. Implement the minimum code needed to pass. 6. Refactor while keeping the focused test green. 7. Add contract, integration, or end-to-end coverage required by `docs/testing.md`. 8. Run the focused suite, then the complete relevant validation. 9. Update documentation in the same change. Do not write production code first and add superficial tests afterward. TDD is a development discipline; CI cannot reconstruct the red/green history, so code review must verify that tests describe behavior rather than implementation. ## Adding a feature Before implementation, answer these questions in the issue or pull request: - What observable behavior changes? - Which layer owns it? - Is the public Python API affected? - Does it add or change an optional dependency? - Which unit, contract, integration, and end-to-end tests are required? - Which documentation and migration notes must change? Prefer dependency injection and stable contracts over provider-name branches. Keep transport-specific models at the repository edge. See `docs/integrations.md` for provider work. ## Validation commands Use the smallest command during development and the complete gate before finishing: ```bash make test-unit make test-integration make test-e2e make architecture-check make format-check make lint make typecheck make package-check make check ``` See `docs/development.md` for focused test commands and environment details. ## Definition of done A behavior change is complete only when: - acceptance criteria are represented by tests; - unit tests cover the new behavior; - bug fixes include regression tests; - implementations of extension contracts pass shared contract tests; - affected component interactions have integration coverage; - affected public workflows have end-to-end coverage where appropriate; - architecture, formatting, lint, typing, and package checks pass at their currently enforced scope; - documentation and migration guidance are current; - default tests need no secrets, network, or running external service; - the base package remains importable without unrelated optional extras. Documentation-only changes may state why runtime tests are not applicable, but documentation checks and relevant static validation still apply. ## Working safely in an existing worktree - Inspect `git status` before editing. - Treat existing modifications as user-owned unless explicitly told otherwise. - Do not discard, rewrite, stage, or commit unrelated changes. - Re-read shared files immediately before editing them. - Keep commits focused and make generated changes explicit. - Never place credentials, tokens, local databases, or provider responses in tests or fixtures. - If another change prevents validation, report the exact command and failure; do not silently modify unrelated code to make the tree green.
Signals
- Stars
- 673
- Forks
- 102
- Open issues
- 14
- Last commit (this folder)
- Oct 15, 2026 (15 days ago)
- Last release
- Mar 24, 2026 (7 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/Bessouat40/RAGLight
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from Bessouat40
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get AGENTS.md · Bessouat40/RAGLight running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
AGENTS.md · Bessouat40/RAGLight. Agent Mart. https://agentmart.dev/templates/bessouat40-raglight-agents-md (last updated 2026-10-24)