Skip to Content
Tares

Can I ask Tares what happened during a disk filling up straight from the console?

Yes, Tares has an assistant called Ask built into the console, and you can type a question about a disk filling up and it investigates your data. It shows each tool call as it works and ends with a verdict. Ask needs a model provider on the instance, and it runs on the instance’s default provider, which the chat names.

How it works

Ask is a chat loop that runs on the Tares daemon, with up to 10 tool rounds per question. Its tools are thin calls to the daemon’s own read endpoints, so it sees what the API exposes and nothing more.

For a question about a disk filling up, Ask can use these tools:

  • List sources and connectors, and describe any catalog object, including its schema, entities, freshness and samples.
  • Profile a source’s fields and list its entities.
  • Read recent events.
  • Use read to get a correlated timeline, across every source or only one project’s.

The answer comes from the data already in your instance. If the events around the disk filling up are in a source, Ask can read them and lay them out on a timeline. If a source looks stale or a label is missing, the same read tools answer that from live health and field coverage.

The model is the provider’s default. On Anthropic that is TARES_AGENT_MODEL, which is claude-sonnet-4-6 unless you set it. The same providers also power Tares agents. For the wider picture of watching and investigating incidents, see AI SRE.

Set it up

  1. Make sure Tares is running. On a local install, start it with the command below.
tares up
  1. Give the instance a model provider in one of two ways. Set ANTHROPIC_API_KEY in the daemon’s environment before the daemon starts, or open the console and add a provider under Settings, Model providers.
  2. Check which provider is the default. Ask runs on the instance’s default provider, and the chat tells you which one it is using.
  3. Open Ask in the sidebar, or press ⌘K from anywhere in the console.
  4. Type your question, for example: What happened during the disk filling up, and which service was affected first?
  5. Watch the tool calls as Ask works through them, then read the verdict it reaches. Open the Ask page for the full list of what it can do.

What you get

  • An investigation of the disk filling up from inside the console, with no external client to wire up first.
  • Every tool call visible while Ask works, so you can see which sources and events it read and how it got to its verdict.
  • A correlated timeline across all your sources, or only a project’s, so the disk events sit next to whatever else was happening.
  • Debugging help when the answer is thin. Ask can tell you whether a source is not ingesting, what looks sparse and why a label is missing.
  • Optional drafts of catalog changes. Ask has two proposal tools, propose_labels and propose_trigger, that show cards you Apply or skip. A trigger you apply from Ask goes to the Default project. The proposals change nothing until you apply them, and Apply goes through the same validated management APIs the console forms use.

When something else fits

  • Use your own agent, such as Claude Code, Cursor or your own service, when you want the same read surface inside your own workflow, plus subscriptions. Connect it to Tares over MCP. Ask lives inside the console and reads this one instance.
  • If you want to set up a whole project from a goal rather than ask questions, use the guided setup instead of Ask.

Still stuck? Get help

Last updated on