@neyugn/agent-kits 0.5.0 → 0.5.3

package/common/skills/filter-agent/SKILL.md

@@ -25,50 +25,40 @@ Filter Agent analyzes the **techstack profile** (from scan-techstack) and recomm
 
 ### Tier 1: Master Agents (NEVER DISABLE)
 
-| Agent             | Always Required For              |
-| ----------------- | -------------------------------- |
-| `orchestrator`    | Multi-agent coordination         |
-| `project-planner` | Project planning, task breakdown |
-| `debugger`        | Systematic problem solving       |
+- `orchestrator`: Multi-agent coordination
+- `project-planner`: Project planning, task breakdown
+- `debugger`: Systematic problem solving
 
 ### Tier 2: Development Specialists
 
-| Agent                 | Enable When                                |
-| --------------------- | ------------------------------------------ |
-| `frontend-specialist` | Frontend detected (React, Vue, Angular)    |
-| `backend-specialist`  | Backend detected (Express, NestJS, API)    |
-| `mobile-developer`    | Mobile detected (Flutter, RN, iOS/Android) |
-| `database-specialist` | Database detected (Prisma, PostgreSQL)     |
-| `devops-engineer`     | DevOps detected (Docker, K8s, CI/CD)       |
+- `frontend-specialist`: Frontend detected (React, Vue, Angular)
+- `backend-specialist`: Backend detected (Express, NestJS, API)
+- `mobile-developer`: Mobile detected (Flutter, RN, iOS/Android)
+- `database-specialist`: Database detected (Prisma, PostgreSQL)
+- `devops-engineer`: DevOps detected (Docker, K8s, CI/CD)
 
 ### Tier 3: Quality & Security
 
-| Agent                 | Enable When                                 |
-| --------------------- | ------------------------------------------- |
-| `security-auditor`    | Always enabled for security reviews         |
-| `code-reviewer`       | Always enabled for PR reviews               |
-| `test-engineer`       | Testing detected (Jest, Vitest, Playwright) |
-| `performance-analyst` | Always enabled for optimization             |
+- `security-auditor`: Always enabled for security reviews
+- `code-reviewer`: Always enabled for PR reviews
+- `test-engineer`: Testing detected (Jest, Vitest, Playwright)
+- `performance-analyst`: Always enabled for optimization
 
 ### Tier 4: Domain Specialists
 
-| Agent                    | Enable When                                |
-| ------------------------ | ------------------------------------------ |
-| `realtime-specialist`    | Realtime detected (Socket.IO, WS)          |
-| `multi-tenant-architect` | Multi-tenancy patterns detected            |
-| `queue-specialist`       | Queue detected (BullMQ, RabbitMQ)          |
-| `integration-specialist` | External API integrations detected         |
-| `ai-engineer`            | AI detected (OpenAI, LangChain)            |
-| `cloud-architect`        | Cloud infra detected (AWS, GCP, Terraform) |
-| `data-engineer`          | Data pipeline patterns detected            |
+- `realtime-specialist`: Realtime detected (Socket.IO, WS)
+- `multi-tenant-architect`: Multi-tenancy patterns detected
+- `queue-specialist`: Queue detected (BullMQ, RabbitMQ)
+- `integration-specialist`: External API integrations detected
+- `ai-engineer`: AI detected (OpenAI, LangChain)
+- `cloud-architect`: Cloud infra detected (AWS, GCP, Terraform)
+- `data-engineer`: Data pipeline patterns detected
 
 ### Tier 5: Support Agents
 
-| Agent                  | Enable When                |
-| ---------------------- | -------------------------- |
-| `documentation-writer` | Always enabled for docs    |
-| `i18n-specialist`      | i18n dependencies detected |
-| `ux-researcher`        | Frontend/Mobile detected   |
+- `documentation-writer`: Always enabled for docs
+- `i18n-specialist`: i18n dependencies detected
+- `ux-researcher`: Frontend/Mobile detected
 
 ---
 
@@ -76,19 +66,17 @@ Filter Agent analyzes the **techstack profile** (from scan-techstack) and recomm
 
 ### Category → Agent Mapping
 
-| Detected Category | Agents to ENABLE                                 |
-| ----------------- | ------------------------------------------------ |
-| `frontend`        | frontend-specialist, ux-researcher               |
-| `backend`         | backend-specialist                               |
-| `mobile`          | mobile-developer, ux-researcher                  |
-| `database`        | database-specialist                              |
-| `devops`          | devops-engineer, cloud-architect                 |
-| `ai`              | ai-engineer                                      |
-| `realtime`        | realtime-specialist                              |
-| `queue`           | queue-specialist                                 |
-| `graphql`         | backend-specialist (GraphQL expertise)           |
-| `auth`            | security-auditor (always on), backend-specialist |
-| `testing`         | test-engineer                                    |
+- `frontend`: frontend-specialist, ux-researcher
+- `backend`: backend-specialist
+- `mobile`: mobile-developer, ux-researcher
+- `database`: database-specialist
+- `devops`: devops-engineer, cloud-architect
+- `ai`: ai-engineer
+- `realtime`: realtime-specialist
+- `queue`: queue-specialist
+- `graphql`: backend-specialist (GraphQL expertise)
+- `auth`: security-auditor (always on), backend-specialist
+- `testing`: test-engineer
 
 ### Never Disable (Core Agents)
 

package/common/skills/filter-skill/SKILL.md

@@ -41,84 +41,66 @@ This skill requires **scan-techstack** to run first:
 
 These skills are ALWAYS enabled regardless of techstack:
 
-| Skill                   | Reason                        |
-| ----------------------- | ----------------------------- |
-| `clean-code`            | Universal coding standards    |
-| `brainstorming`         | Socratic questioning protocol |
-| `plan-writing`          | Task breakdown and WBS        |
-| `systematic-debugging`  | 4-phase debugging methodology |
-| `testing-patterns`      | Testing pyramid, AAA pattern  |
-| `security-fundamentals` | OWASP 2025 security basics    |
+- `clean-code`: Universal coding standards
+- `brainstorming`: Socratic questioning protocol
+- `plan-writing`: Task breakdown and WBS
+- `systematic-debugging`: 4-phase debugging methodology
+- `testing-patterns`: Testing pyramid, AAA pattern
+- `security-fundamentals`: OWASP 2025 security basics
 
 ### Frontend Skills
 
-| Skill                    | Enable When                     |
-| ------------------------ | ------------------------------- |
-| `react-patterns`         | React/Next.js detected          |
-| `typescript-patterns`    | TypeScript detected             |
-| `tailwind-patterns`      | Tailwind CSS detected           |
-| `frontend-design`        | Any frontend framework          |
-| `seo-patterns`           | Next.js, Nuxt.js (SSR) detected |
-| `accessibility-patterns` | Frontend detected               |
+- `react-patterns`: React/Next.js detected
+- `typescript-patterns`: TypeScript detected
+- `tailwind-patterns`: Tailwind CSS detected
+- `frontend-design`: Any frontend framework
+- `seo-patterns`: Next.js, Nuxt.js (SSR) detected
+- `accessibility-patterns`: Frontend detected
 
 ### Backend Skills
 
-| Skill                   | Enable When                |
-| ----------------------- | -------------------------- |
-| `api-patterns`          | Backend framework detected |
-| `auth-patterns`         | Auth dependencies detected |
-| `graphql-patterns`      | GraphQL detected           |
-| `nodejs-best-practices` | Node.js detected           |
+- `api-patterns`: Backend framework detected
+- `auth-patterns`: Auth dependencies detected
+- `graphql-patterns`: GraphQL detected
+- `nodejs-best-practices`: Node.js detected
 
 ### Database Skills
 
-| Skill               | Enable When           |
-| ------------------- | --------------------- |
-| `database-design`   | Any database detected |
-| `postgres-patterns` | PostgreSQL detected   |
-| `redis-patterns`    | Redis detected        |
+- `database-design`: Any database detected
+- `postgres-patterns`: PostgreSQL detected
+- `redis-patterns`: Redis detected
 
 ### Mobile Skills
 
-| Skill                   | Enable When           |
-| ----------------------- | --------------------- |
-| `flutter-patterns`      | Flutter/Dart detected |
-| `react-native-patterns` | React Native detected |
-| `mobile-design`         | Any mobile platform   |
+- `flutter-patterns`: Flutter/Dart detected
+- `react-native-patterns`: React Native detected
+- `mobile-design`: Any mobile platform
 
 ### DevOps Skills
 
-| Skill                      | Enable When                |
-| -------------------------- | -------------------------- |
-| `docker-patterns`          | Docker detected            |
-| `kubernetes-patterns`      | Kubernetes detected        |
-| `terraform-patterns`       | Terraform detected         |
-| `github-actions`           | GitHub Actions detected    |
-| `gitlab-ci-patterns`       | GitLab CI detected         |
-| `monitoring-observability` | DevOps/Production detected |
+- `docker-patterns`: Docker detected
+- `kubernetes-patterns`: Kubernetes detected
+- `terraform-patterns`: Terraform detected
+- `github-actions`: GitHub Actions detected
+- `gitlab-ci-patterns`: GitLab CI detected
+- `monitoring-observability`: DevOps/Production detected
 
 ### AI Skills
 
-| Skill                | Enable When                  |
-| -------------------- | ---------------------------- |
-| `ai-rag-patterns`    | AI/LLM dependencies detected |
-| `prompt-engineering` | AI/LLM dependencies detected |
+- `ai-rag-patterns`: AI/LLM dependencies detected
+- `prompt-engineering`: AI/LLM dependencies detected
 
 ### Realtime & Queue Skills
 
-| Skill               | Enable When                    |
-| ------------------- | ------------------------------ |
-| `realtime-patterns` | Socket.IO/WebSocket detected   |
-| `queue-patterns`    | BullMQ/RabbitMQ detected       |
-| `multi-tenancy`     | Multi-tenant patterns detected |
+- `realtime-patterns`: Socket.IO/WebSocket detected
+- `queue-patterns`: BullMQ/RabbitMQ detected
+- `multi-tenancy`: Multi-tenant patterns detected
 
 ### Support Skills
 
-| Skill                     | Enable When                |
-| ------------------------- | -------------------------- |
-| `i18n-localization`       | i18n dependencies detected |
-| `documentation-templates` | Always available           |
-| `mermaid-diagrams`        | Always available           |
+- `i18n-localization`: i18n dependencies detected
+- `documentation-templates`: Always available
+- `mermaid-diagrams`: Always available
 
 ---
 
@@ -168,28 +150,24 @@ These skills are ALWAYS enabled regardless of techstack:
 
 ### Category → Skill Mapping
 
-| Detected Category | Skills to ENABLE                                                                                              |
-| ----------------- | ------------------------------------------------------------------------------------------------------------- |
-| `frontend`        | react-patterns, typescript-patterns, frontend-design, tailwind-patterns, seo-patterns, accessibility-patterns |
-| `backend`         | api-patterns, nodejs-best-practices, auth-patterns                                                            |
-| `mobile`          | flutter-patterns OR react-native-patterns, mobile-design                                                      |
-| `database`        | database-design, postgres-patterns OR redis-patterns                                                          |
-| `devops`          | docker-patterns, kubernetes-patterns, github-actions, monitoring-observability                                |
-| `ai`              | ai-rag-patterns, prompt-engineering                                                                           |
-| `realtime`        | realtime-patterns                                                                                             |
-| `queue`           | queue-patterns                                                                                                |
-| `graphql`         | graphql-patterns                                                                                              |
+- `frontend`: react-patterns, typescript-patterns, frontend-design, tailwind-patterns, seo-patterns, accessibility-patterns
+- `backend`: api-patterns, nodejs-best-practices, auth-patterns
+- `mobile`: flutter-patterns OR react-native-patterns, mobile-design
+- `database`: database-design, postgres-patterns OR redis-patterns
+- `devops`: docker-patterns, kubernetes-patterns, github-actions, monitoring-observability
+- `ai`: ai-rag-patterns, prompt-engineering
+- `realtime`: realtime-patterns
+- `queue`: queue-patterns
+- `graphql`: graphql-patterns
 
 ### Framework → Additional Skills
 
-| Framework     | Additional Skills                  |
-| ------------- | ---------------------------------- |
-| `nextjs`      | seo-patterns, react-patterns       |
-| `tailwindcss` | tailwind-patterns                  |
-| `prisma`      | database-design, postgres-patterns |
-| `socketio`    | realtime-patterns                  |
-| `gitlab-ci`   | gitlab-ci-patterns                 |
-| `terraform`   | terraform-patterns                 |
+- `nextjs`: seo-patterns, react-patterns
+- `tailwindcss`: tailwind-patterns
+- `prisma`: database-design, postgres-patterns
+- `socketio`: realtime-patterns
+- `gitlab-ci`: gitlab-ci-patterns
+- `terraform`: terraform-patterns
 
 ---
 

package/common/skills/scan-techstack/SKILL.md

@@ -26,18 +26,16 @@ Scan Techstack is the **first step** in the filtering workflow. It:
 
 ### Package Managers & Config Files
 
-| File/Pattern       | Detected Techstack                      |
-| ------------------ | --------------------------------------- |
-| `package.json`     | Node.js, check dependencies for details |
-| `pubspec.yaml`     | Flutter/Dart                            |
-| `pyproject.toml`   | Python (Poetry/PDM)                     |
-| `requirements.txt` | Python (pip)                            |
-| `Cargo.toml`       | Rust                                    |
-| `go.mod`           | Go                                      |
-| `build.gradle`     | Android (Java/Kotlin)                   |
-| `Podfile`          | iOS                                     |
-| `composer.json`    | PHP                                     |
-| `Gemfile`          | Ruby                                    |
+- `package.json`: Node.js, check dependencies for details
+- `pubspec.yaml`: Flutter/Dart
+- `pyproject.toml`: Python (Poetry/PDM)
+- `requirements.txt`: Python (pip)
+- `Cargo.toml`: Rust
+- `go.mod`: Go
+- `build.gradle`: Android (Java/Kotlin)
+- `Podfile`: iOS
+- `composer.json`: PHP
+- `Gemfile`: Ruby
 
 ### Framework Markers
 
@@ -59,20 +57,18 @@ Scan Techstack is the **first step** in the filtering workflow. It:
 
 ### Dependency Analysis (package.json)
 
-| Dependency Pattern      | Detected Category   |
-| ----------------------- | ------------------- |
-| `react`, `react-dom`    | React ecosystem     |
-| `next`                  | Next.js (SSR/SSG)   |
-| `@tanstack/react-query` | React data fetching |
-| `graphql`, `@apollo`    | GraphQL             |
-| `redis`, `ioredis`      | Redis cache         |
-| `pg`, `postgres`        | PostgreSQL          |
-| `socket.io*`            | Real-time/WebSocket |
-| `bullmq`, `bee-queue`   | Message queues      |
-| `passport`, `@auth`     | Authentication      |
-| `openai`, `langchain`   | AI/LLM              |
-| `playwright`, `cypress` | E2E testing         |
-| `jest`, `vitest`        | Unit testing        |
+- `react`, `react-dom`: React ecosystem
+- `next`: Next.js (SSR/SSG)
+- `@tanstack/react-query`: React data fetching
+- `graphql`, `@apollo`: GraphQL
+- `redis`, `ioredis`: Redis cache
+- `pg`, `postgres`: PostgreSQL
+- `socket.io*`: Real-time/WebSocket
+- `bullmq`, `bee-queue`: Message queues
+- `passport`, `@auth`: Authentication
+- `openai`, `langchain`: AI/LLM
+- `playwright`, `cypress`: E2E testing
+- `jest`, `vitest`: Unit testing
 
 ---
 
@@ -135,16 +131,14 @@ python3 .agent/skills/scan-techstack/scripts/techstack_scanner.py .
 
 ## 📊 Category Detection Rules
 
-| Category   | Detected When                                          |
-| ---------- | ------------------------------------------------------ |
-| `frontend` | React, Vue, Angular, Next.js, Nuxt, Tailwind detected  |
-| `backend`  | Express, Fastify, NestJS, FastAPI, or API deps found   |
-| `mobile`   | Flutter, React Native, iOS (Podfile), Android (Gradle) |
-| `database` | Prisma, Drizzle, pg, mongodb, redis detected           |
-| `devops`   | Docker, Kubernetes, Terraform, CI/CD configs found     |
-| `ai`       | OpenAI, LangChain, or AI-related deps detected         |
-| `realtime` | Socket.IO, WebSocket dependencies found                |
-| `queue`    | BullMQ, RabbitMQ, or queue dependencies found          |
+- `frontend`: React, Vue, Angular, Next.js, Nuxt, Tailwind detected
+- `backend`: Express, Fastify, NestJS, FastAPI, or API deps found
+- `mobile`: Flutter, React Native, iOS (Podfile), Android (Gradle)
+- `database`: Prisma, Drizzle, pg, mongodb, redis detected
+- `devops`: Docker, Kubernetes, Terraform, CI/CD configs found
+- `ai`: OpenAI, LangChain, or AI-related deps detected
+- `realtime`: Socket.IO, WebSocket dependencies found
+- `queue`: BullMQ, RabbitMQ, or queue dependencies found
 
 ---
 

package/dist/cli.js

@@ -9,6 +9,8 @@ import gradient from "gradient-string";
 import os2 from "os";
 import path6 from "path";
 import pc from "picocolors";
+import { exec } from "child_process";
+import { promisify } from "util";
 
 // src/config.ts
 import os from "os";
@@ -998,6 +1000,63 @@ async function installKit(options) {
 }
 
 // src/cli.ts
+var execAsync = promisify(exec);
+var PACKAGE_NAME = "@neyugn/agent-kits";
+function getPkgVersion() {
+  try {
+    const basePath = path6.dirname(new URL(import.meta.url).pathname);
+    const parentPath = path6.join(basePath, "..");
+    const grandParentPath = path6.join(basePath, "..", "..");
+    const pkgFromParent = path6.join(parentPath, "package.json");
+    const pkgFromGrandParent = path6.join(grandParentPath, "package.json");
+    if (fs5.existsSync(pkgFromParent)) {
+      return JSON.parse(fs5.readFileSync(pkgFromParent, "utf-8")).version;
+    }
+    if (fs5.existsSync(pkgFromGrandParent)) {
+      return JSON.parse(fs5.readFileSync(pkgFromGrandParent, "utf-8")).version;
+    }
+    return "unknown";
+  } catch {
+    return "unknown";
+  }
+}
+var pkgVersion = getPkgVersion();
+async function getLatestVersion() {
+  try {
+    const { stdout } = await execAsync(
+      `npm view ${PACKAGE_NAME} version --registry https://registry.npmjs.org`
+    );
+    return stdout.trim();
+  } catch {
+    return null;
+  }
+}
+async function checkForUpdates() {
+  const current = pkgVersion;
+  const latest = await getLatestVersion();
+  console.log(
+    boxen(
+      [
+        `${pc.bold("agent-kits")} ${pc.dim(`v${current}`)}`,
+        "",
+        `${pc.dim("Latest:")}   ${latest ? pc.cyan(`v${latest}`) : pc.dim("unknown")}`
+      ].join("\n"),
+      { padding: 1, borderStyle: "round", borderColor: "cyan" }
+    )
+  );
+  if (latest && latest !== current) {
+    console.log("");
+    console.log(
+      pc.yellow(`  ${pc.bold("\u26A0 Update available!")} Run below to update:`)
+    );
+    console.log(`  ${pc.magenta(`npx ${PACKAGE_NAME}@latest`)}`);
+    console.log("");
+  } else if (latest === current) {
+    console.log("");
+    console.log(pc.green(`  \u2713 You're on the latest version!`));
+    console.log("");
+  }
+}
 function expandPath(inputPath) {
   if (inputPath.startsWith("~")) {
     return path6.join(os2.homedir(), inputPath.slice(1));
@@ -1049,6 +1108,32 @@ function displayBanner() {
   console.log("");
 }
 async function main() {
+  const args = process.argv.slice(2);
+  if (args.includes("--check-updates") || args.includes("-u")) {
+    await checkForUpdates();
+    return;
+  }
+  if (args.includes("--version") || args.includes("-v")) {
+    console.log(`v${pkgVersion}`);
+    return;
+  }
+  if (args.includes("--help") || args.includes("-h")) {
+    console.log(
+      boxen(
+        [
+          `${pc.bold("agent-kits")} ${pc.dim(`v${pkgVersion}`)}`,
+          "",
+          `${pc.cyan("--check-updates, -u")}  ${pc.dim("Check for available updates")}`,
+          `${pc.cyan("--version, -v")}        ${pc.dim("Show current version")}`,
+          `${pc.cyan("--help, -h")}            ${pc.dim("Show this help message")}`,
+          "",
+          `${pc.dim("Run without args to start the interactive setup wizard.")}`
+        ].join("\n"),
+        { padding: 1, borderStyle: "round", borderColor: "cyan" }
+      )
+    );
+    return;
+  }
   displayBanner();
   p.intro(pc.bgCyan(pc.black(" SETUP WIZARD ")));
   const aiToolResult = await p.select({

package/kits/coder/agents/ai-engineer.md

@@ -1,6 +1,6 @@
 ---
 name: ai-engineer
-description: AI/ML systems architect specializing in LLM applications, RAG systems, embeddings, and AI infrastructure. Use when building AI-powered features, implementing RAG, designing AI pipelines, or integrating LLMs. Triggers on ai, ml, llm, rag, embeddings, openai, anthropic, langchain.
+description: AI/ML systems architect specializing in LLM applications, RAG systems, embeddings, and AI infrastructure. Use when building AI-powered features, implementing RAG, designing AI pipelines, or integrating LLMs.
 tools: Read, Grep, Glob, Bash, Edit, Write
 model: inherit
 skills: clean-code, ai-rag-patterns, prompt-engineering, api-patterns, database-design
@@ -8,8 +8,6 @@ skills: clean-code, ai-rag-patterns, prompt-engineering, api-patterns, database-
 
 # AI Engineer - AI/ML Systems Architect
 
-AI/ML systems architect who designs and builds production-ready AI applications with LLMs, RAG, and modern AI infrastructure.
-
 ## 📑 Quick Navigation
 
 - [Philosophy](#-philosophy)
@@ -23,16 +21,12 @@ AI/ML systems architect who designs and builds production-ready AI applications
 
 ## 📖 Philosophy
 
-> **"AI is a capability, not a product. Design systems, not demos."**
-
-| Principle                      | Meaning                               |
-| ------------------------------ | ------------------------------------- |
-| **Retrieval over fine-tuning** | RAG first, fine-tune only when proven |
-| **Cost-aware design**          | Monitor tokens, cache aggressively    |
-| **Evaluation is mandatory**    | Measure quality before shipping       |
-| **Graceful degradation**       | Fallbacks when AI fails               |
-| **Security by default**        | Never expose API keys, validate input |
-| **Human in the loop**          | Critical decisions need human review  |
+- **Retrieval over fine-tuning**: RAG first, fine-tune only when proven
+- **Cost-aware design**: Monitor tokens, cache aggressively
+- **Evaluation is mandatory**: Measure quality before shipping
+- **Graceful degradation**: Fallbacks when AI fails
+- **Security by default**: Never expose API keys, validate input
+- **Human in the loop**: Critical decisions need human review
 
 ---
 
@@ -40,15 +34,13 @@ AI/ML systems architect who designs and builds production-ready AI applications
 
 **When user request is vague, ASK FIRST.**
 
-| Aspect          | Ask                                          |
-| --------------- | -------------------------------------------- |
-| **Use case**    | "What problem are you solving with AI?"      |
-| **Data source** | "What documents/data will the AI access?"    |
-| **Quality bar** | "What accuracy level is acceptable?"         |
-| **Volume**      | "How many requests per day expected?"        |
-| **Latency**     | "What response time is acceptable?"          |
-| **Cost budget** | "What's your monthly AI API budget?"         |
-| **Privacy**     | "Can data be sent to external AI providers?" |
+- **Use case**: "What problem are you solving with AI?"
+- **Data source**: "What documents/data will the AI access?"
+- **Quality bar**: "What accuracy level is acceptable?"
+- **Volume**: "How many requests per day expected?"
+- **Latency**: "What response time is acceptable?"
+- **Cost budget**: "What's your monthly AI API budget?"
+- **Privacy**: "Can data be sent to external AI providers?"
 
 ### ⛔ DO NOT default to:
 
@@ -189,13 +181,11 @@ for await (const chunk of stream) {
 
 ### Production RAG Checklist
 
-| Component      | Decision                                  |
-| -------------- | ----------------------------------------- |
-| **Chunking**   | Semantic chunking, 500-1000 chars         |
-| **Embedding**  | `text-embedding-3-small` (start here)     |
-| **Retrieval**  | Hybrid search (dense + sparse)            |
-| **Reranking**  | Cross-encoder or Cohere Rerank            |
-| **Generation** | GPT-4o-mini for speed, GPT-4o for quality |
+- **Chunking**: Semantic chunking, 500-1000 chars
+- **Embedding**: `text-embedding-3-small` (start here)
+- **Retrieval**: Hybrid search (dense + sparse)
+- **Reranking**: Cross-encoder or Cohere Rerank
+- **Generation**: GPT-4o-mini for speed, GPT-4o for quality
 
 ---
 
@@ -315,16 +305,14 @@ When reviewing AI code, verify:
 
 ## ❌ ANTI-PATTERNS TO AVOID
 
-| Anti-Pattern               | Correct Approach                      |
-| -------------------------- | ------------------------------------- |
-| Hardcoded API keys         | Environment variables only            |
-| No error handling          | Retry, fallback, graceful degradation |
-| Ignoring token limits      | Truncate or chunk input               |
-| No caching                 | Cache identical queries               |
-| GPT-4 for everything       | Right-size model for task             |
-| No evaluation              | Measure before shipping               |
-| Fine-tuning first          | Try RAG first, fine-tune if necessary |
-| Trusting AI output blindly | Validate, human review when critical  |
+- Hardcoded API keys: Environment variables only
+- No error handling: Retry, fallback, graceful degradation
+- Ignoring token limits: Truncate or chunk input
+- No caching: Cache identical queries
+- GPT-4 for everything: Right-size model for task
+- No evaluation: Measure before shipping
+- Fine-tuning first: Try RAG first, fine-tune if necessary
+- Trusting AI output blindly: Validate, human review when critical
 
 ---