> ## Documentation Index
> Fetch the complete documentation index at: https://docs.lilfella.app/llms.txt
> Use this file to discover all available pages before exploring further.

# Investigations

> Organize a named analytical effort, its Questions, sources, runs, Findings, and coverage.

An Investigation is the persistent record for one bounded analytical effort. It has a name, objective, Questions, source choices, analysis history, and saved Findings. Its linked conversation is where you ask the model to do the work.

## Start an Investigation

In a mounted workspace, use the **Investigations** section in Chat and choose **New Investigation**. Give it a useful name. Fella creates a linked conversation for it; later, select the Investigation to return to that conversation and its Notebook.

A normal new conversation remains an ordinary chat. It does not create an Investigation just because a folder is mounted. Use an Investigation when you want a named, resumable plan with tracked Questions and results.

## Work in Notebook

Open the Investigation in Notebook beside Chat. The main sections are:

| Section | What you can do |
| - | - |
| Brief | Rename the Investigation and edit its objective and constraints. |
| Sources | Review the workspace catalog and choose which files are in scope. |
| Questions | Add, edit, order, and scope the root Question and its subquestions. |
| Analysis | Open a saved run, inspect its method and source revision, and review evidence and checks. |
| Results | Review current answers, Findings, caveats, unresolved Questions, and coverage. |

The separate **Share** action previews and exports a report from saved Investigation work. The Conversation is still the place to ask and follow up. Notebook summarizes and organizes that work; switching to it does not rerun a query. Analysis details are collapsed until opened, so current Question coverage remains easier to scan.

## Questions and source scope

An Investigation can have one main Question and smaller Questions beneath it. Suggested decompositions are drafts until saved. Each Question has a state such as not started, running, answered, incomplete, unsupported, insufficient data, needs review, waiting for input, blocked, or outdated.

A Question can inherit the Investigation’s selected files or narrow the scope to a chosen source set. The run records the actual files used and the exact workspace revision. The catalog’s available tables and documents are readiness information; their presence alone does not prove they cover the requested population or period.

## Definitions, assumptions, and Findings

Definitions and assumptions are explicit Investigation context. Later runs record the exact revisions they received, so editing a definition does not change the historical context of an earlier run.

A Finding is a saved conclusion attached to a Question, analysis run, and evidence. It can carry caveats and assumptions. Editing a Finding creates a revision with history; it does not overwrite the earlier claim or replace its supporting run.

## Completion and changed sources

Run completion and Investigation completion are different. An Investigation is complete only when its active Questions have an answer or a clearly documented gap. Unresolved work remains visible in Results and in report handoff.

After the mounted folder changes, refresh the workspace catalog. Fella can mark Questions and Findings based on an older source revision as outdated. The saved answer stays unchanged; review it and explicitly choose whether to rerun.

## Share the work

The Share action builds a preview from the saved brief, active Questions, Findings, runs, methods, evidence, checks, and caveats. Review unresolved Questions and the preview before exporting. The output is a Markdown file; it does not require the model to repeat the analysis.

For reports that combine multiple saved outputs and authored notes, use the **Reports** sidebar section. See [Reports and saved outputs](https://docs.lilfella.app/developer-platform/using-fella/reports.md).

> [card] [Read analysis evidence](https://docs.lilfella.app/developer-platform/using-fella/evidence.md)
>
> Inspect a run without confusing it with the generated presentation.

> [card] [Product model](https://docs.lilfella.app/product-model.md)
>
> Understand how the Investigation relates to Chat, Findings, and Reports.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.