Interactive AI Console

The Interactive AI Console is the first AI-native development surface inside Aksara Studio. Developers type natural-language commands — like "explain the User model" — and the console automatically detects intent, builds context, routes to the correct AI Flow, executes via the runtime, and returns a structured result.

Quick Start

Studio UI

1. Open Aksara Studio → click AI Console in the sidebar (or press Ctrl+I / Cmd+I).
2. Type a command in the input box, e.g. explain the User model.
3. Press Enter to send.
4. Review the AI response and follow suggested next actions.

CLI

aksara ai chat "explain the User model"
aksara ai chat "review GET /api/users" --format json
aksara ai chat "suggest indexes" --provider <provider> --model <model-name>

Use any provider and model combination that your AI setup supports.

HTTP API

# Send a console query
curl -X POST http://localhost:8000/studio/ai/console \
  -H "Content-Type: application/json" \
  -d '{"message": "explain the User model"}'

# Get autocomplete suggestions
curl "http://localhost:8000/studio/ai/console/suggest?q=explain"

Architecture

User input
    │
    ▼
┌─────────────────┐
│  Intent Router   │  — rule-based pattern matching, no LLM call
│  (intent_router) │
└────────┬────────┘
         │ IntentMatch(intent, flow_type, action_key, confidence, context)
         ▼
┌─────────────────┐
│ Context Builder  │  — enriches with registry data (models, routes, etc.)
│ (console_context)│
└────────┬────────┘
         │ enriched context dict
         ▼
┌─────────────────┐
│  Console Engine  │  — orchestrates the full pipeline
│ (console_engine) │
└────────┬────────┘
         │
    ┌────┴────┐
    ▼         ▼
┌────────┐ ┌──────────┐
│AI Flow │ │ Runtime   │  — builds prompt pack → executes via connector
│Builder │ │(run_pack) │
└────────┘ └──────────┘

Supported Intents

The console recognises 11 intents, one for each AI Flow action:

Intent	Flow Type	Example Command
explain_model	model	"Explain the User model"
suggest_constraints	model	"Suggest constraints for Order"
refactor_suggestions	model	"Refactor the Product model"
review_endpoint	route	"Review GET /api/users"
harden_permissions	route	"Harden permissions on POST /api/admin"
generate_examples	route	"Generate example requests for GET /api/products"
explain_plan	query	"Explain the query plan"
suggest_indexes	query	"Suggest indexes"
rewrite_suggestions	query	"Rewrite query for performance"
explain_migration	migration	"Explain migration impact"
safe_rollout_plan	migration	"Safe rollout plan for migration"
diagnostic_prioritize	diagnostic	"Prioritize diagnostic issues"

Response Shape

All console responses follow this structure:

{
  "ok": true,
  "intent": "explain_model",
  "flow_type": "model",
  "action_key": "explain_model",
  "confidence": 0.91,
  "extracted_context": {"model_name": "User"},
  "prompt_pack": { "..." },
  "execution": {
    "ok": true,
    "provider": "configured-provider",
    "model": "configured-model",
    "response": "The User model has 5 fields...",
    "tokens": {"prompt": 20, "completion": 50, "total": 70},
    "elapsed_ms": 1234.5
  },
  "suggestions": ["suggest_constraints", "refactor_suggestions"],
  "elapsed_ms": 1500.0,
  "error": null,
  "error_code": null
}

Safety

The AI Console follows Aksara's core safety rule:

AI suggests, developer confirms.

The console will never:

• Auto-modify source code
• Execute migrations
• Apply database changes
• Introduce autonomous agents

All output is analysis, suggestions, and explanations that the developer reviews before taking any action.

Command Suggestions

The console provides autocomplete suggestions as you type. The suggestion system is entirely local — no LLM calls are made for suggestions.

Suggestions are available via:

• Studio UI: typing in the console input field
• API: GET /studio/ai/console/suggest?q=prefix

Keyboard Shortcuts

Shortcut	Action
Ctrl+I / Cmd+I	Open AI Console
Enter	Send message
↑ / ↓	Navigate history
Escape	Close suggestions

Configuration

The AI Console uses the same provider configuration as the AI Hub. No additional configuration is needed — if AI Hub is configured with a provider, the console will use it automatically.

To override the provider or model per-request:

aksara ai chat "explain the User model" --provider anthropic --model claude-3-5-sonnet

Or via the API:

{
  "message": "explain the User model",
  "provider_override": "anthropic",
  "model_override": "claude-3-5-sonnet"
}