Skip to content

CHAT_ORCHESTRATION.md

Latest commit

 

History

History
364 lines (279 loc) · 8.26 KB

CHAT_ORCHESTRATION.md

File metadata and controls

364 lines (279 loc) · 8.26 KB

Chat Orchestration - Natural Language Commands

🎯 Overview

The chat now intelligently detects your intent and automatically invokes the right functionality - no slash commands required!

✨ What's New

Previously, you HAD to use slash commands:

❌ Old Way (required):
> /scanrepo
> /contextload
> /swarm
> /make

Now, just talk naturally:

✅ New Way (natural):
> scan this repository
> load context
> enable swarm mode
> analyze this app

🧠 Intent Detection

The chat router detects 8 types of intents:

1. Scan Intent

Triggers: "scan repo", "scan codebase", "scan this", "scan project"

Example:

You: scan this repository
App: 🔍 Scanning repository...
     [Runs /scanrepo automatically]

2. Analyze Intent

Triggers: "analyze app", "analyze code", "code analysis", "review code"

Example:

You: analyze this app
App: 🔍 Analyzing codebase with AI...
     [Scans if needed, then runs AI analysis]

3. Context Intent

Triggers: "load context", "save context", "show context", "list context"

Example:

You: load context
App: 📂 Loading saved context...
     [Runs /contextload automatically]

You: show context
App: 📋 Showing context...
     [Displays context state]

4. Swarm Intent

Triggers: "use swarm", "enable swarm", "swarm mode", "activate swarm"

Example:

You: enable swarm mode
App: 🐝 Swarm Mode Configuration
     Current Status: OFF

     To use swarm mode for analysis:
       1. Swarm is now enabled
       2. Add context: run scan or add files
       3. Use /make with your prompt
       4. Swarm will activate for large contexts

     ✓ Swarm mode enabled!

5. Status Intent

Triggers: "show status", "check status", "current status"

Example:

You: show status
App: 📊 Current Status

     Platform:   termux
     Provider:   gemini
     Model:      2.5-flash
     Repository: AIworkbench
     Swarm:      🐝 ON (auto, 5 workers)

6. Edit Intent

Triggers: "implement", "add", "fix", "update", "modify", "change", "edit"

Example:

You: add a dark mode toggle
App: 🤖 Detected code change request. Analyzing repository...
     [Routes to smart_edit automatically]

7. Generate Intent

Triggers: "generate", "scaffold", "boilerplate", "build from scratch"

Example:

You: generate a login page
App: 🤖 Starting code generation workflow...
     [Enters /make mode automatically]

8. Chat Intent (Questions)

Triggers: "how", "what", "why", "explain", questions with "?"

Example:

You: how does swarm mode work?
App: [Provides explanation via AI chat]

🚀 Powerful Workflows

Workflow 1: Quick Analysis

You: scan this repo

App: [Scans all files]
     Context saved: 45 source files + analysis

You: analyze this codebase for improvements

App: [Uses swarm if enabled]
     🐝 Swarm Mode Execution
     Processing 5 chunks with 5 parallel workers
     [Complete analysis provided]

Workflow 2: Enable Swarm & Analyze

You: enable swarm mode

App: ✓ Swarm mode enabled!

You: use swarm to analyze the app

App: 🔍 Analyzing codebase with AI...
     [Scans if needed]
     🐝 Swarm Mode Execution
     [Multi-agent analysis]

Workflow 3: Context Management

You: scan this project

App: [Scans files]
     Context saved: 30 files

You: show context

App: Context: 0 in-memory, 30 saved + scan

You: load context

App: ✅ Loaded 30 files into memory

You: analyze the architecture

App: [Analyzes with loaded context]

🎯 Best Practices

1. Natural Language Works

✅ "scan this repo"
✅ "enable swarm"
✅ "analyze the codebase"
✅ "load my context"
✅ "show current status"

2. Still Use Slash Commands When Specific

✅ /make     (when you want the make menu)
✅ /tweak    (when you want tweak mode specifically)
✅ /context  (when you want the context menu)

3. Questions Stay in Chat

✅ "How does swarm work?"      → AI explanation
✅ "What is context?"          → AI explanation
✅ "Why use swarm mode?"       → AI explanation

4. Commands Are Detected

✅ "scan this"                 → Runs /scanrepo
✅ "analyze app"               → Runs analysis workflow
✅ "enable swarm"              → Enables swarm mode

🔍 How It Works

Detection Order (Priority)

  1. Scan keywords → Repository scan
  2. Context keywords → Context operations
  3. Swarm keywords → Swarm configuration
  4. Status keywords → Status display
  5. Edit keywords → Code editing
  6. Generate keywords → Code generation
  7. Analyze keywords → AI analysis
  8. Question keywords → AI chat
  9. Default → AI chat

Intent Detection Examples

"scan this repo"
→ Matches: "scan.*repo"
→ Intent: scan
→ Action: auto_scan_repo()
→ Result: Repository scanned

"load context"
→ Matches: "load.*context"
→ Intent: context
→ Action: auto_handle_context() → cmd_contextload()
→ Result: Context loaded

"enable swarm mode"
→ Matches: "enable swarm"
→ Intent: swarm
→ Action: auto_handle_swarm() → Enables swarm
→ Result: Swarm enabled

"analyze this app"
→ Matches: "analyze.*app"
→ Intent: analyze
→ Action: auto_analyze_code() → Scans + analyzes
→ Result: AI analysis provided

📋 Complete Command Reference

Natural Language → Action Mapping

What You Say What Happens Equivalent Slash Command
"scan this repo" Scans repository /scanrepo
"scan repository" Scans repository /scanrepo
"scan codebase" Scans repository /scanrepo
"analyze app" Scans + AI analysis /scanrepo → /make
"analyze code" Scans + AI analysis /scanrepo → /make
"load context" Loads saved context /contextload
"save context" Saves context /contextsave
"show context" Shows context state /contextshow
"clear context" Clears context /contextclear
"enable swarm" Enables swarm mode /swarm → Enable
"use swarm" Enables swarm mode /swarm → Enable
"swarm mode" Shows swarm menu /swarm
"show status" Shows current status /status
"check status" Shows current status /status
"implement X" Code editing (smart_edit)
"add X" Code editing (smart_edit)
"fix X" Code editing (smart_edit)
"generate X" Code generation /make
"create X" Code editing/generation (context-dependent)

🎓 Pro Tips

1. Combine Operations

You: enable swarm, then scan and analyze this repo

App: [Enables swarm]
     [Scans repository]
     [Runs AI analysis with swarm]

2. Context-Aware Operations

You: scan this

App: [Scans current directory]

You: now analyze it

App: [Uses just-scanned context for analysis]

3. Chained Workflows

You: load context

App: [Loads 30 files]

You: analyze for security issues

App: [Analyzes loaded files for security]

🐛 Troubleshooting

Intent Not Detected?

If the chat doesn't detect your intent, try:

  1. Use more specific keywords from the triggers list
  2. Use the slash command directly
  3. Rephrase your request

Examples:

❌ "I want to see the repo"
✅ "scan this repo"

❌ "turn on that swarm thing"
✅ "enable swarm mode"

❌ "check what's going on"
✅ "show status"

Still Goes to AI Chat?

If your command went to AI chat instead of routing:

  1. Check if you used a question word (how, what, why)
  2. Try removing question marks
  3. Use imperative form (commands)

Examples:

❌ "Can you scan the repo?"        → Goes to chat
✅ "scan the repo"                 → Runs scan

❌ "How do I enable swarm?"        → Goes to chat
✅ "enable swarm"                  → Enables swarm

🎉 Summary

You asked: "Why not possible for chat to use all functionality?!"

Answer: It IS now! Just talk naturally:

  • "scan this" → Scans
  • "analyze app" → Analyzes
  • "enable swarm" → Enables swarm
  • "load context" → Loads context
  • "show status" → Shows status

No slash commands needed! The chat is now truly intelligent. 🚀

For more information, see:

  • docs/SWARM_MODE_GUIDE.md - Swarm mode details
  • docs/CRITICAL_FIXES.md - Recent bug fixes
  • QUICKSTART.md - Getting started guide