Using the chat

What you can do with the agent: search, analyse, visualise, and create reports through conversation

Use this page to learn what to ask the agent, so a question about a run ends with an answer and evidence, not an afternoon in the logs.

The agent is available throughout the platform. Ask questions, run analysis, build visualisations, and create reports, all in plain language. The same agent also answers in Slack and on Jira tickets.

Where to find it

Home page: the chat is front and centre
Chat panel: click Alloy AI on the right edge of the page (the chat icon in the new design). The panel is titled Agent. It is on every page except Home, which has the chat built in, and Mesh Storage.
Ask AI buttons: on missions and reports, plus the Ask AI about "…" option at the top of Search results

The chat knows where you are. The dotted badge above the input shows the current page, which is included with your next message. Ask a question on the Missions page and it references your current filters. Ask on a report and it can change that report.

Choose the context

Current page: the dotted badge follows you as you navigate and is always included. It cannot be removed.
Other missions and reports: click @, or type it in the input, to search by name or ID. References you add stay selected when you navigate or reopen the panel; starting a new chat clears them.
Selected text: select text inside a chat (on Home or in the chat panel), then click Add to chat to include it.
Continue a task: a clear follow-up such as "apply that title" keeps the agreed target, even after you navigate elsewhere. A new request normally uses the current page. Name the report or mission when you want a different target, or say "keep this in chat" for a response without editing it.

What you can do

Start from a field report or a failed test, the way it reached you:

"The field team says QP-07 stopped mid-patrol at about 14:30 yesterday. What happened, and have we seen this before?"
"This test run failed the docking check. Compare it with the last five passing runs and tell me what's different"
"Did anything change in the robot's logs in the minute before the emergency stop?"

The agent finds the run and the moment, compares it with other runs, and answers with the metrics, events and timestamps behind its findings. The engineering decision stays with your team.

Find data across your mission library:

"Find moments where two robots were at the same waypoint but reported different terrain classifications"
"Can you find any joint temperature spikes that correlate with the gait stumbles?"
"Find all perimeter patrols that completed without intervention and summarise the stats"
"Search for images where the robot encountered standing water"

The agent searches log messages by meaning and images by what is visible. Describe what you're looking for in words.

Run analysis on your data, including SQL queries and Python code:

"Walking speed should drop from 1.2m/s to 0.3m/s when the robot enters a confined space. Find cases where that transition didn't happen"
"Describe path-planning failures and how they lead to missed waypoints on the inspection route"
"Track CPU, GPU, RAM, and joint motor temperatures per robot per mission. Identify spikes and how they trend week to week"

The agent runs SQL and Python on your data in a secure sandbox.

Create interactive visualisations for reports:

"Plot where the leg servo errors begin and how they accumulate over the patrol"
"Show me a map of all inspection routes from this week"
"Build a weekly comparison of system health metrics per robot by serial number"
"Make a timeline of key events during this inspection run"

Visualisations can be embedded in reports for sharing.

Create and refine custom reports:

"That stair-climb failure you found for QP-07 during the warehouse inspection. Make a report about what happened"
"Create a report on how joint calibration drift builds up across back-to-back patrols"
"Remove the battery section and focus on navigation"

See Creating a Report for more.

Set up checks that run on every new mission:

"Create a scenario to flag any patrol where battery drops below 20% before returning to dock"
"Set up a scenario for leg motor fault events"

See Scenarios for more.

Open the Replay Viewer at exactly the moment you're discussing, with useful panels already laid out. Be as vague or as specific as you want:

"Show me the altitude drop around 14:32 on yesterday's patrol"
"Open the replay where the obstacle avoidance kicked in"
"Show me battery and motor-current time series with the map on top and camera5 in the bottom left"

Save the layouts you want to reuse. Once a layout works for the question you're asking, save it. Next time you ask something similar, the agent can reopen it instead of building a new one.

Analyse attached files

Attach files directly in chat to analyse them alongside your missions: CSV or text telemetry, GNSS files such as RINEX, RTCM3 and SBF, ZIP bundles, documents and images. For example: "Summarise the satellites and observation gaps in this RINEX file" or "Compare these odometer counts with the flight telemetry."

The agent can parse these files, summarise their contents, plot measurements, and export analysis tables. Attachments stay in the chat for 7 days. To keep recordings permanently, create a mission from them; when you attach recordings, the agent offers to. Parsing GNSS observations does not itself calculate an RTK or PPK position. When you compare files, give the timestamp conventions, units, and any known synchronisation offset.

Uploads are limited to 250 MB per file, 1 GB total, and 10 files per message. For large files, ask about a specific time window.

Create a mission from chat

A mission groups one recording, one test run or a set of related recordings. Ask the agent to turn MCAP files into a mission, for example "Make this a mission called Hotfire 3", or accept its offer when you attach recordings. The mission is processed exactly like one from Missions → Upload, and you are recorded as its uploader.

MCAP only. Up to 20 .mcap files per mission, each with a different name. The agent checks each file first: a valid MCAP becomes a mission straight away. A damaged one (for example cut off mid-recording) is repaired, and the agent tells you what was wrong and how many messages it kept. A file that is not an MCAP, or has no usable messages, is refused with the reason.
Other formats. For CSV, JSON or other telemetry, the agent can convert the data to MCAP first. It shows you the files, topics, row counts and time range, and creates the mission only after you confirm.
Metadata. Attach a metadata.json or tell the agent details such as the vehicle, serial number or site, and it writes them into the MCAP, where they become mission filters. Notes go in the mission description.
One request, one mission. Retrying the same request does not create a duplicate.

The new mission appears at the top of the Missions list within a few seconds, and fills in as processing finishes. Processing usually takes a few minutes; the agent tells you when larger recordings will take longer. Creating missions from chat is available in web chat, not in Slack.

Reuse saved scripts and documents

Your team can save analysis scripts and documents in the Scripts library tab of your organisation settings. Ask the agent to find one by name when you want to reuse it, for example: "Find our battery health script."

Tips

Iterate: start broad, then refine with follow-ups
Use the context: on a report, ask the agent to change sections; from Search, select Ask AI to explore a topic
Ask for the evidence: ask which missions, times and queries an answer relies on, or ask the agent to open Replay at the moment it cites