Airavat is in Private Alpha. Request early access at airavat.xyz
MahoutGetting Started

Getting Started

Mahout is a chat-first research workspace. You start with a question, Mahout streams its answer, and useful outputs can be saved as artifacts.

Open Mahout

  1. Log in to Airavat.
  2. Open the dashboard.
  3. Select Mahout from the sidebar.

If the sidebar item is missing or disabled, your account has not been enabled for Mahout.

Start a session

Click New Chat or use the empty chat panel to send your first message.

A session keeps its message history, generated artifact candidates, and saved artifacts together. Use separate sessions for separate research threads so the context stays clean.

Reopen previous sessions

Use the History tab to reopen older sessions. Each row shows the session title, last update time, latest preview, model, effort level, and message count.

Open a previous session when you want to continue the same research thread. Start a new chat when the asset, time window, or decision changes materially.

See History for the full session workflow.

Pick a model and effort level

Mahout exposes model and effort controls when they are available for your account.

ControlHow to use it
ModelPick the model you want Mahout to use for the current session or turn. You can also set model preferences for your account.
EffortUse lower effort for quick summaries and higher effort for deeper analysis, multi-step reasoning, or backtests.

Higher effort can take longer and may use more of the daily model budget. See Models, Skills & Limits for details.

Send a good first prompt

Start with:

  • the asset, macro theme, filing, or question you care about
  • the time window
  • the decision you are trying to inform
  • any constraints, such as “use only public data” or “focus on the last two quarters”

Examples:

Analyze SOL over the last 60 days. Focus on trend, volatility, liquidity, and events that could change the next two-week setup.
Pull the latest SEC EDGAR filings for this company and summarize what changed versus the prior filing.
Backtest a simple 50/200 SMA crossover on BTC daily bars from 2020 through 2024 and return the performance table plus chart.

Read the streaming response

Mahout streams progress as it works. You may see:

  • answer text
  • tool activity
  • data-source references
  • artifact candidates
  • final response status

Wait for the final response before treating the answer as complete.

Continue the thread

Follow-up prompts can refine the same analysis:

  • “Show the same result as a table.”
  • “Use a shorter lookback.”
  • “Compare this to ETH.”
  • “Save the chart.”
  • “Rerun with a higher effort model.”

If the thread starts mixing unrelated topics, create a new session.

Save useful outputs

When Mahout produces an artifact candidate, save it if you want to keep it beyond the chat response. Saved artifacts are available from the Artifacts view.

See Artifacts for supported artifact types and rerun behavior.

Set up an agent

When a research question turns into a rule you want watched continuously — “tell me when X happens” — ask Mahout to turn it into an agent. Mahout drafts and compiles the rule, can preview it against historical time, and deploys it on a schedule with in-app notifications when it fires.

See Research Workflows for the full agent workflow.