Chat
Chat is the main working surface in Engenerate. It combines grounded retrieval, citations, attachments, plans, notebooks, and optional capabilities in one place.
What Chat Is For
Use chat when you want to:
- ask grounded questions about project material
- compare sources and inspect citations
- create or refine a plan for a larger task
- execute work and build outputs
- save useful results to notebooks
Analyze, Plan, and Do
The mode selector in the prompt footer helps you match the conversation to the kind of work you are doing.
| Mode | Best For |
|---|---|
| Analyze | Search, compute, and draft — grounded questions, analysis, comparisons, and in-chat drafts |
| Plan | Creating or refining a structured work plan before execution |
| Do | Executing work, following a plan, and assembling deliverable-ready output |
Use Analyze when you are still learning what is in the project or drafting an answer in place. Use Plan when the task has multiple steps or needs a clear structure. Use Do when you are ready to carry out the work.
Chat Tabs
Chats are organized as tabs, and each tab is an independent conversation.
- Create a new conversation from the + button in the chat tab bar
- Let Engenerate auto-name the tab after the first exchange
- Keep separate threads for different questions, scenarios, or report tracks
The + Menu
The + button beside the prompt is the main control center for context and optional capabilities.
Add photos or files
Use attachments for one-turn context:
- Photos and images (JPG, PNG) for visual questions or figure references — pasting a screenshot works too
- Documents (PDF, Word, PowerPoint) for one-off supporting material
Attachments help with the current turn, but they are not automatically indexed as project knowledge.
Attachments use the same list of accepted formats as document upload. CSV and plain text attachments are temporarily unavailable while we rework how unprocessable file types are handled; Excel workbooks are not accepted.
Context
The Context section controls what the model can draw from during the conversation.
| Item | What It Does |
|---|---|
| Resources | Search processed project resources and pin specific files |
| Notebooks | Pin notebook content from the current workspace |
| Libraries | Pin organization or Engenerate library references |
| Context settings | Open retrieval settings for search scope, mode, result count, score threshold, and navigation filtering |
When you pin a resource, notebook, or library document, that selection becomes part of the working context for future turns.
Capabilities
The Capabilities section exposes optional chat behaviors and tools.
| Capability | What It Does |
|---|---|
| Web search | Pull live web results into the turn (Quick or Thorough depth) |
| Research | Use a deeper research flow for harder questions |
| Review Rubrics | Choose a rubric for document review, or Author with AI (create/edit in Do) — see Review Rubrics for details |
| Knowledge Modules | Apply a saved guidance package for a discipline or workflow — see Knowledge Modules for details |
| Output Templates | Shape the final response with a reusable outline, or Create/Edit in Chat (author in Do). When a template is selected, the footer icon opens a read-only preview with Remove / Edit / Edit in Chat. See Output Templates for details. |
Some capabilities may only appear when they are enabled for your account or workspace, or when you are working inside a project.
Context Usage
The prompt footer includes a context usage meter. Open it to see how much of the available context window is being used and how that usage is split between input and output.
This is especially helpful when you are pinning large notebooks, adding multiple sources, or switching to larger models.
Retrieval Settings
The retrieval settings help you tune how Engenerate searches your knowledge sources.
| Setting | What It Does |
|---|---|
| Search Scope | Search only the current project or all of your projects |
| Search Mode | Use Vector Only for semantic search or Hybrid for semantic plus keyword matching |
| Results (Top K) | Control how many candidate matches are considered |
| Min Score | Filter out weaker matches |
| Exclude Navigation Blocks | Ignore low-value structural content such as headers, footers, and table-of-contents text |
For most grounded work, the defaults are a good starting point. Switch to Hybrid when exact terminology, codes, or identifiers matter.
Models
Engenerate offers a flat, user-selectable list of models spanning OpenAI, Anthropic, and Google — there is no free tier vs. paid tier split. Every model draws from the same usage balance.
Manage which models are available to you and choose a default in Settings > General > Models, where per-model pricing is also shown.
In chat, the model selector lists your active models plus your default. Each model has a Thinking toggle and an Effort level (Low, Medium, High, Extra High, or Max) that you can adjust per turn to trade off response depth against speed.
Conversation Actions
Assistant responses support several actions directly in chat.
Copy
Click Copy to choose Copy (this response) or Copy Full Chat. Full-chat copies the entire conversation, with each turn labeled User Message or Response and a single trailing References list. Full-chat is disabled while that response is still streaming.
Send to Notebook
Click Send to Notebook to choose Send to Notebook (this response) or Send Full Chat to Notebook. Full-chat writes the whole conversation with the same User Message / Response headings. Full-chat is disabled while that response is still streaming.
Citations
Grounded responses include citations that link back to the underlying source. Selecting a citation opens the referenced material in context so you can verify the answer.