Lagan Chat

Most research work in Ishvana goes through structured tools — the Research Pipeline for query-based synthesis, Search for semantic lookup, Deep Analysis for single-page understanding. These are all fine when you know what you’re looking for. But sometimes you don’t know yet. You’re at the start of a research session with a vague sense that something needs to be explored, and you want to talk about it with someone who’s good at research before you start structuring the work. Lagan Chat is that conversation. It’s a freeform chat interface with Lagan — Ishvana’s research specialist — where you ask questions, get answers, follow up, explore tangents, and generally use her as a research collaborator instead of a search tool. It’s where research sessions happen rather than where research tasks happen.

The tab is the conversational counterpart to every other tool in the Research module. You use Search when you know the answer exists in your knowledge base. You use the Pipeline when you have a specific research question. You use Lagan Chat when you want to think out loud with someone who’s going to push back, clarify, and occasionally run a tool on your behalf mid-conversation.

What Lagan Chat is

A regular chat interface with Lagan. You type messages, she responds, you keep going. Responses stream in real time. Every turn shows a toolbar with per-turn actions:

Copy. Copy Lagan’s response to clipboard.
Send to Lore. Turn Lagan’s response into a Legendry entry.
Regenerate. Have Lagan try the response again.
Edit and resend. Modify your own message and regenerate from that point.

The interface auto-scrolls to the newest message as the conversation grows. Under the input, a strip shows the tools Lagan has used in the current conversation — if she’s searched your knowledge base, consulted WorldKnowledge, fetched a web page, or run analysis on content, the strip shows the action so you can see what she’s been doing behind the scenes.

The conversation history persists across sessions. Closing Ishvana and reopening doesn’t lose the chat — you pick up where you left off.

What Lagan can do mid-chat

The chat interface isn’t just pure conversation. Lagan has access to the same tools she has in the Pipeline, and she can invoke them autonomously based on what the conversation needs:

Knowledge base search. “Do I have anything in my notes about medieval shipbuilding?” triggers a semantic search across your bookmarks and research history. Lagan reports what she found.
Lore search. “What characters are from Grimnest?” triggers a search across your Legendry. She returns the matches.
Web research. “Can you look up how long it takes to sail from Port Royal to Nugao?” triggers fresh web research via the browser and Wikipedia integration.
Page analysis. “What does this page I have loaded say about political systems?” triggers content analysis on the currently-loaded browser page.
Specialist consultation. “Is this historically accurate?” can trigger consultation with the WorldKnowledge agent.
Research bookmarks. “Save this conversation as a research note” triggers a save action.

The tool use happens transparently. You don’t tell her “search for X” — you ask a question, and if her answer needs a search, she runs one. The tools-used strip below the input shows you which tools fired so you can verify or explore further.

Why conversational research is different

The structured research tools (Pipeline, Search) work best when you have a clear question. Lagan Chat works best when you don’t, and the value is in the back-and-forth that helps you find the question.

A typical Lagan Chat session might look like:

You: “I’m writing a scene set in a medieval port city and I want to get the details right. What should I be thinking about?”
Lagan: “A few broad categories matter: the economics of the port (what’s traded, where it goes), the social structure (who lives there, who visits), the layout (where the docks are, where merchants live, where sailors drink), the governance (who runs it, how disputes are settled). What’s the closest historical reference you have in mind?”
You: “Something like 13th-century Venice but smaller. Maybe a regional port.”
Lagan: “Got it. For a regional port in that era, the biggest simplifiers are: no formal city council (usually a dominant merchant family), most trade is regional not international, and the ‘maritime district’ is usually one specific block rather than spread across the city. Do you want me to pull up some sources on Venetian port operations that scale down to regional cities?”
You: “Yes, but also I want to know whether the port would have a customs system.”
Lagan: runs research “Here’s what I found…”

Each turn is short, conversational, and targeted. Lagan is helping you figure out what you actually need before diving into sources. This is the mode that’s hardest to get from a search-first tool — conversational exploration is structurally different from query-based lookup.

Keeping the conversation scoped

Long Lagan Chat conversations can drift. You start asking about port cities, you end up discussing the history of the Hanseatic League, you circle back to maritime law, you dive into shipbuilding, you lose track of the original question. This is fine for genuine exploration but can become unproductive if you’re trying to get to a specific answer.

Two features help:

Edit and resend. If a conversation branch is going wrong, you can edit your own earlier message and resend it. Everything after that point gets replaced with the new response, pruning the tangent.
Send to Lore. When you reach a useful conclusion, save it as a Legendry entry before the conversation drifts. The saved entry becomes a research note you can reference later, even if the rest of the conversation gets lost.

Use these proactively. Don’t wait until the conversation has spiraled — when you’ve got a useful answer, save it, and if the conversation takes a bad turn, edit the bad prompt and redirect.

Context and persistence

Every Lagan Chat conversation exists in the current project’s chat history. Switching projects loads that project’s chat history. The histories are separate — your fantasy novel’s Lagan conversations don’t mix with your mystery series’ conversations.

Within a single conversation, Lagan has the full history as context. She remembers what you’ve already discussed and can refer back to it. Old turns eventually get trimmed by Etherforce’s context compression when the history gets too long, but the compression preserves recent turns and summarizes older ones so the conversation stays coherent.

If you want to start a new conversation without Lagan remembering the previous one, click New Conversation. The old one is saved to history but doesn’t affect the new one’s context.

When to use Lagan Chat vs. other tabs

A few rules of thumb:

Use Lagan Chat when:

You don’t know what your question is yet.
You want to think out loud.
You want Lagan to ask you follow-up questions.
You’re exploring a new topic area.
You want a conversational record of your research thinking.

Use the Research Pipeline when:

You have a specific research question and want structured output.
You want the answer synthesized into a summary you can save.
You need relevance scoring and source citations.

Use Search when:

You know exactly what you’re looking for in your existing knowledge base.
You need results fast with minimal ceremony.

Use Deep Analysis when:

You have a specific page in front of you and want to understand it deeply.

Different tools for different stages of research. Lagan Chat is for the exploration stage.

What Lagan Chat isn’t

Not a general-purpose chatbot. Lagan’s system prompt is tuned for research. She won’t help you with prose-craft — that’s Hawken’s job. She won’t answer mechanics questions — that’s GameMaster’s job. Her scope is research, and she redirects out-of-scope requests to the right specialist.
Not a long-term memory system. Lagan remembers the current conversation but not previous conversations. If you need persistent research knowledge across sessions, save it as Lore entries or research bookmarks.
Not a free service. Lagan runs on the LLM engine configured in Settings → Models. Each message costs tokens. Long conversations cost more than short ones. The Cost Tracking panel shows you what your research sessions are costing if you want to monitor.

What’s next

Research Pipeline The structured alternative for query-based research.

Lagan (agent page) Deeper documentation on Lagan herself — her capabilities, tools, and limits.

Deep Analysis When a specific page needs close reading instead of a conversation.