Skip to content

Chat with the agent

The Scry tab is your primary surface. Streaming responses, live tool results, multi-turn context.

Sending a message

Type in the composer at the bottom, tap Send. The agent will:

  1. Stream a short "thinking" indicator while picking tools
  2. Stream text + inline tool result cards as they arrive
  3. Settle into the final reply with action icons below it

You can also tap a suggestion chip on the empty state — those cycle on each new session. The shuffle button rotates them.

What's an "inline tool result card"

When the agent calls a tool that returns structured data, Scry doesn't print the raw JSON — it renders a typed card. For example:

  • ros_topic_hz returns rate + std-dev → renders a big-number card
  • fleet_overview returns the robot list → renders a fleet table
  • ros_node_info returns the publishers/subscribers → renders a tabbed panel
  • tf_tree returns the transform tree → renders a hierarchical view

The agent's prose adds context and anomaly callouts above the card; it never repeats what the card already shows.

Multi-turn

Conversations are persisted in Room (local SQLite). The agent sees the full conversation history on every turn, so follow-ups work:

You: what topics are on /odom*?
AI: [lists /odom, /odom_filtered]
You: what's the rate on the second one?
AI: [calls ros_topic_hz on /odom_filtered]

Conversations are per-robot. Switching robots in Fleets switches the chat surface too.

Action icons under each assistant reply

Every settled assistant message has a row of small icons:

Icon What
(scry mark) Subtle indicator that the agent finished this turn
Copy Copy the assistant's text to clipboard
Retry Re-run this turn — sends the prior user message again
/ Flag this reply (see Sending feedback)
⋮ More Edit message, fork conversation, delete from here

Editing and forking

  • Long-press a user message → Edit. You can rewrite your prompt and the agent re-runs from there, discarding everything after.
  • Long-press any message → Fork. Splits the conversation into a new session at that point, preserving the original.

Useful when you're exploring "what if I asked it differently."

Robot switching

The chat top-bar shows the active robot's name and live latency:

Scry ▾
deep-dell · 33ms

Tap the dropdown to switch robots. The chat surface swaps to that robot's history. Your in-flight composer text follows you.

Model picker

Tap the Scry chip at the top to change AI provider/model. The picker shows OpenRouter's 300+ models grouped by tier (Free, Cheap, Mid, Premium) plus any Ollama models discovered on the robot.

The selection is per-device, not per-conversation — switching models mid-conversation works fine.