Skip to main content

Introduction

Adist provides powerful AI-powered features through integration with multiple Large Language Model (LLM) providers. You can use cloud-based providers like Anthropic Claude and OpenAI, or run models locally with Ollama.

Available LLM Providers

Adist supports three LLM providers:

Anthropic Claude

Cloud-based AI using Claude 3 models (Opus, Sonnet, Haiku)

OpenAI

Cloud-based AI using GPT-4o, GPT-4 Turbo, and GPT-3.5 Turbo

Ollama

Run AI models locally with no API costs

LLM-Powered Features

Adist uses LLMs to provide several intelligent features:

Document Summarization

Generate comprehensive summaries of your project files to help you understand large codebases quickly. Each file summary includes:
  • Purpose and functionality
  • Key components (classes, functions, modules)
  • Notable patterns and important details

Question Answering

Ask specific questions about your codebase and get AI-powered answers based on semantic search and context analysis:
adist query "How does authentication work?"

Interactive Chat

Have natural conversations about your project with persistent context:
adist chat
The chat session maintains:
  • Conversation history within the session
  • Context awareness across multiple questions
  • Automatic retrieval of relevant documents

Streaming Responses

All AI interactions support two modes:
Shows a loading spinner while generating responses with full code highlighting:
adist query "Explain the authentication system"

Smart Context Management

Adist implements intelligent context optimization to improve response quality and reduce costs:

Topic-Based Caching

The system automatically:
  • Identifies query topics using AI
  • Caches relevant context for 30 minutes
  • Reuses cached context for related queries
  • Merges related contexts when appropriate

Query Complexity Analysis

Each query is analyzed for complexity (low, medium, high) based on:
  • Word count
  • Technical terms
  • Code snippets
  • Comparison indicators
Context allocation is dynamically adjusted based on complexity.

Document Relevance Scoring

Documents are scored based on:
  • Code blocks and examples
  • Comments and documentation
  • Function and class definitions
  • Query complexity requirements
Higher-scoring documents receive more space in the context window.

Configuration

Switch between LLM providers using the configuration command:
adist llm-config
This interactive command allows you to:
  • Select your preferred LLM provider
  • Choose specific models (Claude 3 Opus/Sonnet/Haiku, GPT-4o/GPT-4 Turbo/GPT-3.5 Turbo, or local Ollama models)
  • Configure API URLs (for Ollama)

Markdown Formatting

All LLM responses are formatted using proper Markdown with:
  • Headers (#, ##, ###)
  • Bold (**text**) and italic (*text*)
  • Code blocks with syntax highlighting (language)
  • Inline code references (`code`)
  • Lists (bullet and numbered)
Syntax highlighting is automatically applied when code blocks specify the language (e.g., javascript, python, ```typescript).

Cost Tracking

Adist tracks API costs for cloud-based providers:
  • Anthropic Claude: $3 per million tokens (Claude 3 Sonnet)
  • OpenAI: 10/millioninputtokens,10/million input tokens, 30/million output tokens (GPT-4o)
  • Ollama: Free (runs locally)
Costs are displayed after each operation when using cloud providers.

Next Steps

Anthropic Setup

Configure Anthropic Claude for cloud-based AI

OpenAI Setup

Configure OpenAI GPT models

Local Setup

Run Ollama locally for free AI features

Build docs developers (and LLMs) love