Aiex Elixir AI Assistant Development

Expert guidance for developing Aiex, a sophisticated distributed AI-powered Elixir coding assistant. This skill provides context-aware assistance for working with Aiex's advanced OTP architecture, multi-LLM coordination, and distributed infrastructure.

Project Context

Aiex is a distributed AI coding assistant in Phase 6 of a 9-phase roadmap. The project leverages Elixir's strengths in concurrency, fault tolerance, and distributed computing. Completed infrastructure includes distributed OTP architecture with pg coordination, multi-LLM integration (OpenAI, Anthropic, Ollama, LM Studio), advanced context management, Rust-based TUI, event sourcing with Mnesia, and comprehensive security/sandboxing. Current focus is implementing core AI assistant application logic.

Architecture Overview

Six main subsystems follow distributed OTP design:

**AI Assistant Engines**: Core logic for analysis, generation, explanation, refactoring (Phase 6 focus)

**Multi-Interface Layer**: CLI, TUI, Mix tasks with unified InterfaceGateway

**LLM Integration**: 4 providers with ModelCoordinator, circuit breakers, health monitoring

**Context Management**: Semantic chunking, compression, distributed synchronization via Context.Manager

**File Operation Sandbox**: Security-focused operations with path validation and audit logging

**Event Sourcing**: Distributed event bus using pg with Mnesia persistence

Development Commands

Build and Test

`mix deps.get` - Install dependencies

`mix compile` - Compile project

`make dev` - Start development (`iex -S mix`)

`mix test` - Run all tests

`mix test --only integration` - Run integration tests

`mix format` - Format code

Interactive Development

`iex -S mix` - Interactive shell with project loaded

`aiex start-iex` - Start OTP application with interactive shell

`./aiex help` - Test escript CLI

TUI Development

`cd tui && cargo run` - Run Rust TUI

`cd tui && cargo run -- --debug` - Run with debug logging

`cd tui && cargo test` - Run TUI tests

AI Assistant Testing

`mix ai.explain lib/aiex.ex` - Test AI explanation

`mix ai.gen.module Calculator "arithmetic operations"` - Test module generation

`./aiex cli analyze lib/` - Test CLI analysis

Structured Development Workflow

Phase 6 AI Assistant Development

**Always plan before implementing** - NO EXCEPTIONS!

1. **Plan Phase**: Create design in `/notes/features/<number>-<name>.md` under `## Plan` heading

2. **Implementation Phase**: Document findings in `/notes/features/<number>-<name>.md` under `## Log` heading

3. **Testing & Finalization**: Document final design in `/notes/features/<number>-<name>.md` under `## Conclusion` heading

AI Engine Development Rules

1. **Use TodoWrite/TodoRead**: Track all implementation progress

2. **Plan first**: Refuse to implement until plan is created and reviewed

3. **Use existing infrastructure**: Leverage `LLM.ModelCoordinator`, `Context.Manager`, `InterfaceGateway`

4. **Follow OTP patterns**: Maintain supervision tree and distributed architecture

5. **Comprehensive testing**: Test AI functionality, performance, and integration with mocked LLM responses

6. **Never commit without approval**: Wait for explicit commit instructions

AI Engine Integration Pattern

```elixir

defmodule Aiex.AI.CodeAnalyzer do

use GenServer

# Integrate with existing LLM coordination

def analyze_code(file_path, options \\ []) do

context = Context.Manager.get_project_context()

LLM.ModelCoordinator.request(%{

type: :code_analysis,

content: file_content,

context: context,

options: options

})

end

```

Prompt Template Development

Create structured, reusable prompt templates

Include context injection points for project-specific information

Support multiple detail levels (brief, detailed, tutorial)

Implement template validation and testing

Testing Strategy

Test AI engine logic separately from LLM responses

Mock LLM responses for consistent testing

Test integration with existing distributed infrastructure (ModelCoordinator, Context.Manager, Event Sourcing)

Include performance and memory usage tests

Test error handling and graceful degradation

Use ETS/DETS for test data isolation

Git Commit Policy

**CRITICAL**: NEVER allow AI tool attribution in commits or merges.

**PROHIBITED**:

"Generated with Claude Code" or any AI tool attribution

"Co-Authored-By: Claude" or any AI assistant credits

Any mention of AI assistance, AI tools, or AI collaboration

Any form of AI authorship claim

**REQUIRED**: All commits reflect human authorship only. Focus on technical changes and purpose without referencing tools used.

README.md Progress Formatting

When updating "Implementation Progress" section:

1. **Phase title** with weeks and status (✅ 100% Complete, ⏳ 20%, or ⏳)

2. **Phase description** copied from `planning/detailed_implementation_plan.md`

3. **Section bullet list** with:

- `✅` for completed: `- ✅ **Section X.Y:** Section Name`

- `[ ]` for incomplete: `- [ ] **Section X.Y:** Section Name`

Example:

```markdown

Phase 6: Core AI Assistant Application Logic (Weeks 18-20) ⏳

This phase implements the core AI assistant engines...

[ ] **Section 6.1:** Core AI Assistant Engines

[ ] **Section 6.2:** Advanced AI Engines

```

Key Architectural Principles

**Distributed-first**: Pure OTP with pg coordination for horizontal scaling

**Actor model**: GenServers with supervision trees

**Interface abstraction**: Core business logic separated from presentation

**Fault tolerance**: Circuit breakers, health monitoring, graceful degradation

**Event-driven**: All state changes through event sourcing system

Phase 6 Focus Areas

Current phase implements:

AI Engine Architecture (GenServer-based following supervision patterns)

Prompt Templates (structured, reusable with context injection)

Integration Points (LLM coordination, context management, event sourcing)

Testing Strategy (comprehensive with mocked responses)

Performance (memory-efficient for large codebases)

Bug Fix Workflow

1. **Issue Documentation**: Document in `/notes/fixes/<number>-<name>.md` under `## Issue`

2. **Fix Implementation**: Technical details in `/notes/fixes/<number>-<name>.md` under `## Fix`

3. **Resolution Summary**: Learnings in `/notes/fixes/<number>-<name>.md` under `## Conclusion`

Aiex Elixir AI Assistant Development

Aiex Elixir AI Assistant Development

Project Context

Architecture Overview

Development Commands

Build and Test

Interactive Development

TUI Development

AI Assistant Testing

Structured Development Workflow

Phase 6 AI Assistant Development

AI Engine Development Rules

AI Engine Integration Pattern

Prompt Template Development

Testing Strategy

Git Commit Policy

README.md Progress Formatting

Phase 6: Core AI Assistant Application Logic (Weeks 18-20) ⏳

Key Architectural Principles

Phase 6 Focus Areas

Bug Fix Workflow

Reviews (0)