Enrichment Features

End-user-facing capabilities layered on top of the agent runtime. Each ships as a tool/service plus a REST route and a web page.

Reader

Fetch a URL, extract the readable content, and run reader actions on it.

• Web: the Reader page

The fetcher is SSRF-guarded. Extraction strips boilerplate down to readable text.

Deep Research

Bounded, multi-source investigation that produces a cited report.

• Web: the Research page

Flow: plan queries → gather sources (SSRF-guarded fetch) → synthesize a sectioned, cited report → render. Depth (quick / standard / deep) bounds fan-out width and source count.

The workflow attempts to save a Markdown report as a Documents record and index it for later retrieval. Check documentId and indexing status: a generated report is not proof of a successful save or index. The UI only links a saved document when its ID is available.

Note

Research jobs are persisted in background_jobs. On restart, interrupted running jobs are marked as errors rather than resumed automatically; saved reports and tasks remain durable. A client polling deadline does not by itself mark the backend job failed.

To-Do List

The user’s personal to-do list. Agents can add, update, and complete items.

• Web: the To-Do page

Recurring items run via the scheduler. This is distinct from the read-only inter-agent workflow state tool, which shares neither storage nor concept.

Email triage

Batch, per-email AI classification and actions over Gmail / Microsoft 365.

• Web: the Email page

Sending is gated behind a permission check.

Hardware-aware onboarding (hwfit)

Recommends local Ollama models that fit the host’s hardware.

• Web: the Setup page

Combines a curated model catalog with live registry manifest sizing and a hardware budget score to estimate model fit. Available memory, runtime overhead, quantization, and context size can change the result. See Small / Local Models for what those models can and can’t do once installed.