Skip to main content

​
Overview

The chat command starts an interactive session where you can have natural conversations with AI about your project. Unlike single queries, chat maintains conversation history and context across multiple questions.

​
Starting a Chat Session

adist chat
You’ll see: 
Chat Session Started
Project: my-awesome-project
Type "/help" to see available commands

You:
Enable streaming for real-time responses:
adist chat --stream

​
How Chat Works

​
Context-Aware Conversations

For each message, the chat system:
  1. Searches for relevant blocks: Finds code blocks matching your question
  2. Includes conversation history: Maintains context from previous messages
  3. Generates AI response: Uses LLM with full project context
  4. Displays formatted output: Shows code with syntax highlighting
  5. Tracks costs: Monitors API usage per message

​
Example Chat Session

You: how does the indexer work?

Assistant:
The indexer uses a block-based approach...
[detailed explanation with code snippets]

Message cost: $0.0031

You: what file types does it support?

Assistant:
The indexer supports JavaScript, TypeScript, Python...
[continues with context from previous answer]

Message cost: $0.0018 (using cached context)

​
Slash Commands

Chat mode includes built-in commands:

​
/help

Show all available commands: 
You: /help

Available Commands:
/help - Show available commands
/exit - Exit the chat session
/reset - Reset the chat history
/clear - Clear the terminal screen
/cost - Show total cost of the session
/debug - Toggle debug information display
/reindex - Reindex the project with summaries

​
/exit

End the chat session: 
You: /exit

Chat Session Ended
Total cost: $0.0234

​
/reset

Clear conversation history and start fresh: 
You: /reset
✓ Chat history reset
This removes all previous messages from context. The AI won’t remember earlier parts of the conversation.

​
/clear

Clear the terminal screen while keeping chat history: 
You: /clear

Chat Session
Project: my-awesome-project

​
/cost

Show total session cost: 
You: /cost

Total session cost: $0.0156

​
/debug

Toggle debug information on/off: 
You: /debug
✓ Debug information disabled
Debug info shows:
  • Number of documents and blocks found
  • Document tree with file structure
  • Block types and line numbers

​
/reindex

Reindex the project without leaving chat: 
You: /reindex

Reindexing project with block-based indexing and summaries...
This will take a few moments...

✓ Project reindexed successfully with block-based indexing!

Chat session resumed.

​
Response Modes

Without the --stream flag:
  • Shows a loading spinner while AI generates response
  • Applies full syntax highlighting to code blocks
  • Better formatted markdown output
  • Cleaner reading experience
adist chat

​
Streaming Mode

With the --stream flag:
  • Shows responses in real-time as they’re generated
  • Faster time-to-first-token
  • May have limited code highlighting
  • Good for long responses
adist chat --stream

​
Debug Information

When debug mode is enabled, you’ll see detailed search results: 
Debug Info:
Found 2 document(s) with 5 relevant blocks

Document tree:
└── src
    ├── commands
    │   └── chat.ts (3 blocks)
    │       ├── function: chatCommand (lines 27-595)
    │       ├── function: handleSlashCommand (lines 119-177)
    │       └── interface: Message (lines 11-14)
    └── utils
        └── block-indexer.ts (2 blocks)
            ├── class: BlockSearchEngine (lines 259-465)
            └── function: searchBlocks (lines 263-284)

​
Summary Support

Chat mode integrates with project summaries:

​
Explicit Summary Requests

You: summary
If summaries were generated during indexing, you’ll get the project summary immediately without an LLM call. Questions like these automatically use summaries when available: 
You: what is this project about?
You: give me an overview
You: describe the architecture
Generate summaries with:
adist reindex --summarize

​
Context Caching

Chat uses intelligent context caching:
  • Cache Key: Based on project ID
  • Cache Contents: Project structure and summaries
  • Benefits:
    • Faster subsequent messages
    • Lower API costs
    • Consistent context
Message cost: $0.0021 (using cached context)

​
Query Complexity

The system analyzes query complexity: 
You: what is this file?
Message cost: $0.0008 · complexity: low
Complexity affects:
  • Token usage
  • Response detail level
  • API cost

​
Code Highlighting

Chat automatically detects and highlights code: 
export const chatCommand = new Command('chat')
  .description('Start an interactive chat session about your project')
  .option('--stream', 'Enable streaming responses')
  .action(async (options) => {
    // Chat implementation
  });
Supported languages include JavaScript, TypeScript, Python, Go, Java, and many more.

​
Best Practices

​
Start Broad, Then Narrow

1

Get an overview

You: what does this project do?
2

Dive into specifics

You: how is user authentication implemented?
3

Explore details

You: show me the JWT token validation code

​
Use Follow-up Questions

The AI remembers context: 
You: how does the indexer work?
Assistant: [explains indexing process]

You: what about summarization?
Assistant: [explains with context from previous answer]

You: does it support custom file types?
Assistant: [continues conversation naturally]

​
Reset When Changing Topics

Use /reset when switching to unrelated topics: 
You: [asking about authentication]
...
You: /reset
✓ Chat history reset
You: [now asking about database schema]

​
Troubleshooting

​
Project Not Indexed

✘ Project has no indexed files.
Run "adist reindex" to index your project files.
Solution: Index your project first: 
adist reindex

​
No Project Selected

✘ No project is currently selected.
Run "adist init" or "adist switch" first.
Solution: Initialize or switch to a project: 
adist init my-project
# or
adist switch my-project

​
LLM Configuration Error

✘ LLM Error: No LLM provider available
To configure an LLM provider, run "adist llm-config"
Solution: Configure your LLM provider: 
adist llm-config

​
Missing Summaries Warning

⚠️ Project does not have summaries.
Run "adist reindex --summarize" to generate summaries for better context.
Summaries are optional but improve response quality: 
adist reindex --summarize

​
Performance Tips

​
Reduce Costs

  1. Use Ollama: Free local LLM, no API costs 
    adist llm-config
# Select Ollama
  2. Enable Summaries: Better context with fewer tokens 
    adist reindex --summarize
  3. Keep Questions Focused: Simpler queries = lower costs

​
Improve Response Speed

  1. Use Streaming Mode: See responses faster 
    adist chat --stream
  2. Reindex Regularly: Keep indexes up-to-date 
    adist reindex
  3. Use Context Caching: Automatic after first query

​
Comparison: Chat vs Query

FeatureChatQuery
Conversation history✅ Yes❌ No
Multiple questions✅ Yes❌ Single
Slash commands✅ Yes❌ No
Cost tracking✅ Per session✅ Per query
Interactive✅ Yes❌ One-shot
Debug toggle✅ Yes✅ Always on
Use chat for exploration and learning. Use query for quick one-off questions.

​
Next Steps

Search Commands

Learn about search and query commands

LLM Configuration

Configure your AI provider

Build docs developers (and LLMs) love

Get started for freeTalk to us