ragwatch AGENTS.md

ragwatch AGENTS.md is an instructions file for Codex, OpenCode from soumendrak/ragwatch. It costs 2,930 tokens per session, scanned A, original, MIT.

Project instructions for RAGWatch, a Python toolkit that records and scores retrieval-augmented generation traces. RAG is a pattern where a language model answers using information retrieved from documents.

In plain words
What is it for?
Use them when implementing or reviewing RAGWatch SDK features, OpenTelemetry instrumentation, retrieval scoring, or LangGraph integration.
Why use it?
They explain the project's architecture, instrumentation limits, quality scores, and supported retrieval workflows before changes are made.

Instructions file for CodexOpenCode

Install

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.

agentmods
npx agentmods add instructions/soumendrak/ragwatch/agents-md
Clone the repo
git clone --depth 1 https://github.com/soumendrak/ragwatch

Made for: Codex, OpenCode.

Wrote 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.

agentmods badge for ragwatch AGENTS.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/soumendrak/ragwatch/agents-md.svg)](https://agentmods.dev/instructions/soumendrak/ragwatch/agents-md)
Your own site
<a href="https://agentmods.dev/instructions/soumendrak/ragwatch/agents-md"><img src="https://agentmods.dev/badge/instructions/soumendrak/ragwatch/agents-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 2,930 This file is loaded in full into every session.
When invoked 2,930 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.02930 $0.02930
Opus 5 $0.01465 $0.01465
Sonnet 5 $0.00586 $0.00586
Haiku 4.5 $0.00293 $0.00293

Measured 4d ago against content hash 1c41f6cfe5d2, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

ragwatch AGENTS.md 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 4d 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.

AGENTS.md · 214 lines

How it starts

The opening of the file, as written. The whole thing — 214 lines — stays where its author put it; the contents beside it link to each section on GitHub.

RAGWatch — AGENTS.md

Version: v0.3 (enterprise hardening + runtime wiring) Tagline: "Quality scores in your RAG traces — computed, not just recorded"

Project Overview

  • Name: RAGWatch
  • Purpose: OpenTelemetry-native RAG observability Python SDK with semantic quality scores
  • Key Differentiator: chunk_relevance_score + user.feedback_scorecomputed semantic quality
  • Instrumentation: Explicit decorators only (use OpenLLMetry for auto-instrumentation)
  • RAG Types: Linear RAG (v0.1), multi-stage via composed decorators
  • Efficiency: ~1-5 ms overhead per request
  • Scope: SDK only. Users bring their own OTel backend.
  • Install: uv add ragwatch (recommended) or pip install ragwatch (core); uv add ragwatch --extra langgraph or pip install ragwatch[langgraph]

Architecture

Instrumentation Strategy

Explicit decorators only — no auto-instrumentation in v0.1.

Layer Instrumentation Reason
LLM calls None (use OpenLLMetry) They do it better; compose via OTel
Embedding generation Explicit Query embedding stored for relevance score
Vector search / retrieval Explicit chunk_relevance_score computed here
User feedback Explicit record_feedback() is user-initiated
Agent nodes / workflows Explicit User decides span boundaries

Linear RAG Stages

Stage Span Name Key Attributes
Embedding ragwatch.embedding.generate model.name, embedding.dimensions, duration_ms
Vector Search ragwatch.retrieval.search top_k, chunks.returned, chunk.relevance_score
Response ragwatch.response.emit response.length, user.feedback_score

chunk_relevance_score Plumbing (single-process only)

  1. embedding.py computes query embedding → stores in OTel Context (not baggage)
  2. context.py manages thread-local storage for query embedding
  3. retrieval.py reads context → computes cosine similarity → sets chunk.relevance_score
  4. Supported dimensions: up to 512-dim

Read the full file on GitHub · 214 lines

Changes

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.

  1. 4d ago First seen · 214 lines · 2,930 tokens per session scan A 1c41f6cfe5d2

Subscribe to this mod's changes

ragwatch AGENTS.md is an instructions file published in the GitHub repository soumendrak/ragwatch (5 stars, last pushed 4mo ago), licensed MIT. It adds 2,930 tokens to every session, about $0.0146 per session on Opus 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.