Ambient mode increases API usage
Each iteration is a full agent call. Standard mode adds 1-3 calls; autonomous can use 5-50+ calls per task.
Background intelligence while you're away
DevDuck continues exploring, researching, and building in the background. Like a colleague who keeps working while you grab coffee.
Triggered when you go idle. Light exploration with limited iterations. Perfect for validation and discovering edge cases.
Continuous self-directed work until task complete. The agent keeps building, iterating, and solving until it signals done.
# Start with ambient mode enabled
DEVDUCK_AMBIENT_MODE=true devduck
# In DevDuck REPL:
๐ฆ ambient # Toggle standard mode (idle-triggered)
๐ฆ auto # Toggle autonomous mode (continuous)
# Start modes
ambient_mode(action="start") # Standard
ambient_mode(action="start", autonomous=True) # Autonomous
# Check status
ambient_mode(action="status")
# Configure at runtime
ambient_mode(action="configure", idle_threshold=60, max_iterations=5)
# Stop
ambient_mode(action="stop")
In autonomous mode, the agent can signal task completion by including any of these phrases:
[AMBIENT_DONE]
[TASK_COMPLETE]
[NOTHING_MORE_TO_DO]
I've completed my exploration
Nothing more to explore
When detected, ambient mode automatically stops and all accumulated work is stored for injection into your next query.
| Variable | Default | Description |
|---|---|---|
DEVDUCK_AMBIENT_MODE |
false | Enable ambient mode on startup |
DEVDUCK_AMBIENT_IDLE_SECONDS |
30 | Seconds of idle time before triggering |
DEVDUCK_AMBIENT_MAX_ITERATIONS |
3 | Maximum background iterations per topic |
DEVDUCK_AMBIENT_COOLDOWN |
60 | Seconds between ambient runs |
| Variable | Default | Description |
|---|---|---|
DEVDUCK_AUTONOMOUS_MAX_ITERATIONS |
50 | Maximum iterations before auto-stop |
DEVDUCK_AUTONOMOUS_COOLDOWN |
10 | Seconds between iterations (fast!) |
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ โ DevDuck REPL โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโค โ โ โ User Input โโโบ Agent Execution โโโบ Response โ โ โ โ โ โ โผ โผ โ โ โโโโโโโโโโโ โโโโโโโโโโโโโโโ โ โ โ Record โ โ Inject โ โ โ โ Query โ โ Results โ โ โ โโโโโโฌโโโโโ โโโโโโโโฒโโโโโโโ โ โ โ โ โ โ โผ โ โ โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโ โ โ โ AmbientMode Thread โ โ โ โ โ โ โ โ Standard: โ โ โ โ [Idle 30s] โโโบ [Prompt] โโโบ [Run] โโโบ [Store] โ โ โ โ โ โ โ Autonomous: โ โ โ โ [10s cooldown] โโโบ [Prompt] โโโบ [Run] โโโบ [Store] โ โ โ โฒ โ โ โ โ โ โโโโโโโโโ Loop until done โโโโโโ โ โ โ โ โ โ โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ โ โ โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Agent continues finding vulnerabilities and issues while you read the code.
Autonomous mode builds complete features while you're away.
Deep exploration of topics with accumulated findings.
Background checking of assumptions and edge cases.
Generate comprehensive docs across multiple iterations.
Autonomous test creation and execution cycles.
Each iteration is a full agent call. Standard mode adds 1-3 calls; autonomous can use 5-50+ calls per task.
| Mode | Typical Calls | Best For |
|---|---|---|
| Standard | 1-3 extra | Low-cost exploration & validation |
| Autonomous | 5-50+ calls | High-value complex tasks |
MAX_ITERATIONS if concerned[AMBIENT_DONE]| Command | Description |
|---|---|
ambient |
Toggle standard ambient mode (REPL command) |
auto / autonomous |
Toggle autonomous mode (REPL command) |
ambient_mode(action="start") |
Start standard ambient mode |
ambient_mode(action="start", autonomous=True) |
Start autonomous mode |
ambient_mode(action="stop") |
Stop ambient mode |
ambient_mode(action="status") |
Get current status and stats |
ambient_mode(action="configure", ...) |
Update settings at runtime |