Chat

Chat is the conversational interface where you actually talk to Ishvana’s agents. It’s the tab you open when you don’t want to use a structured feature like the editor or the Plot Studio — you just want to ask a question, brainstorm something, or have a discussion with an agent that actually knows your world. The critical thing about Chat in Ishvana versus a generic chatbot is that the agents here are lore-grounded: before they answer any question about your project, they search your Legendry, they read your active document, they pull in relevant context, and they respond based on the actual state of your world. Ask about a character and the agent pulls that character’s Legendry entry. Ask about a plot decision and the agent sees your Plot Studio. Ask about prose craft and Hawken responds with your manuscript’s voice in mind. That’s the thing you can’t get from a generic chatbot, and it’s the reason Chat is more useful than a separate window running ChatGPT next to your draft.

This page covers the Chat interface — conversations, agents, personas, stances, lore references, Writers’ Room mode, and the streaming response system that shows you what the agent is actually doing while it works.

Multi-thread conversations

Chat supports multiple independent conversation threads per project. Each thread is its own conversation with its own history — no cross-contamination between threads.

• Create threads — start a new conversation at any time with the New Thread button.
• Switch threads — click any thread in the sidebar to load its full message history.
• Delete threads — remove a thread and all its messages. Confirmation required.
• Thread sidebar — resizable left panel showing all conversations with titles and message counts.

Threads let you keep different conversation contexts separate. A “brainstorm plot for book three” thread shouldn’t contaminate a “research medieval siege tactics” thread. Start a new thread when the topic shifts; switch back when you want to resume an earlier one.

Agent selector

Three agents are directly selectable from the Chat tab, each with a distinct domain and visual identity:

• Ishvana — the orchestrator and your main assistant. Warm, insightful, world-aware. Good for general questions, lore exploration, brainstorming, and any question where you’re not sure which specialist is right.
• Hawken — the writing coach. Prose feedback, style analysis, line-level edit suggestions, scene-level craft questions. Reads your prose and gives you advice on it. Doesn’t draft your book for you.
• Lagan — the research and knowledge specialist. External research, pattern analysis, smart bookmarks, anything that involves pulling information from outside your own Legendry.

Each agent has its own icon and color in the message stream so you can tell at a glance who responded to what. For deeper documentation on each, see Ishvana, Hawken, and Lagan on their dedicated agent pages. GameMaster, WorldKnowledge, and Lorekeeper aren’t directly selectable from the Chat tab — they live in their own specialized interfaces, though Ishvana can delegate to them when the conversation requires it.

Persona system

Personas change how the agent frames its responses. Same agent, same knowledge, different voice and emphasis. Nine personas are available:

Persona	What it does
Default	Ishvana’s core personality — warm, insightful, helpful
Lore Scholar	Deep project lore expertise; digs into worldbuilding detail
Creative Editor	Writing improvement and editorial feedback
World Historian	Systematic worldbuilding with historical framing
Gritty Fantasy Author	Realistic fantasy perspective, raw and grounded
Lore Keeper	Continuity and consistency maintenance focus
Critical Reviewer	Constructive critique and analysis
Brainstorm Facilitator	Creative idea generation and divergent thinking
Technical Writer	Clear documentation and structural clarity

Personas shape voice, expertise stays the same

A Lore Scholar doesn’t know more about your world than the Default persona does — they share the same access to your Legendry. What changes is how they frame the response. A Lore Scholar digs deep and cites sources. A Gritty Fantasy Author grounds everything in physical detail and visceral consequence. Pick the persona that matches the mode you want to think in.

Stance system

Stances control response style independently from the persona. Seven stances available:

Stance	Approach
Brainstorming	Expansive, creative, idea-generating
Critical	Analytical, questioning, constructive critique
Supportive	Encouraging, positive, confidence-building
Analytical	Logical, structured, detail-oriented
Concise	Brief, to-the-point, efficient
Exploratory	Curious, open-ended, investigation-focused
Summarizing	Condensing, organizing, clarity-focused

Mix and match any persona with any stance for precise control over the agent’s voice. A Lore Scholar in Brainstorming stance produces wildly different output from a Lore Scholar in Critical stance — same underlying knowledge, completely different framing.

The combinations aren’t all equally useful. Gritty Fantasy Author + Concise is a great stance for quick scene suggestions. Lore Scholar + Analytical is ideal for deep research questions. Creative Editor + Critical is the combo you reach for when you want actual editorial feedback instead of encouragement.

@-mentions for lore references

Type @ in the message input to reference lore entries directly:

• Fuzzy search by title as you type after the @.
• Navigate the dropdown with arrow keys; select with Enter or Tab.
• Up to 8 matching entries shown at a time.
• Selected entries appear as pinned tags above the input area.
• Pinned entries give the agent direct access to those specific entries’ content for grounding its response.

Mentions are the fastest way to give the agent explicit context. Instead of writing “remember that character Kent Musa from my fantasy series,” you type @kent and the character’s full Legendry entry is available to the agent. The response quality goes up measurably when you use mentions, because the agent has the canonical entry in front of it instead of relying on a search to surface the right material.

Writers’ Room mode

Writers’ Room is for multi-perspective brainstorming

Toggle Writers’ Room mode to have all three agents respond to the same message simultaneously, each from their own specialist perspective. Ishvana synthesizes, Hawken handles prose craft, Lagan handles research depth. Three angles on the same question, side by side. For the conceptual breakdown of what Writers’ Room actually does differently, see the dedicated page.

Toggle Writers’ Room from the Chat toolbar to get responses from all three agents simultaneously:

• Each agent sees the same message but responds from their own perspective.
• Responses stream in real-time with per-agent bubbles.
• Layout options — sequential (one after another), columns (side by side), or threaded (grouped by conversation flow).

Writers’ Room is the mode you reach for when you’re not sure which specialist should answer the question, or when you genuinely want three different framings to compare. It costs roughly 3x the tokens of a single-agent response, so it’s not something to enable for every message — use it deliberately for the hard questions where the multi-perspective view is worth the extra cost.

Streaming with thinking steps

Responses stream in real time with visible pipeline progress. You see what the agent is doing before the response starts arriving:

• Before the first token, a random lore snippet from your world may appear with source attribution. This is both a visual placeholder and a tiny delight — a reminder that the agent is reading your actual world data.
• Thinking steps appear as a checklist showing each pipeline stage with completion status and timing. Lore search, entity extraction, context synthesis, prompt construction, generation — each stage shows as it completes.
• Once tokens start arriving, the response streams with full Markdown rendering. Code blocks, lists, tables, emphasis — all rendered live as the response generates.
• Metadata captured with every response: which agent answered, which model was used, what sources the response drew from, a confidence score, total generation time, tokens per second.

The thinking steps are important for transparency. You’re not waiting in silence wondering if the agent is working — you see each pipeline stage complete and know exactly what’s happening under the hood.

Context panel

A resizable right-side panel showing:

• Persona and stance selectors — change the agent’s voice on the fly.
• Metadata for the selected message — agent used, model, confidence score, generation time, tokens per second, RAG sources that grounded the response, context quality rating.

Click any assistant message to view its metadata. Useful when a response surprises you and you want to see why — the RAG sources tell you which parts of your Legendry the agent actually read before responding.

Keyboard shortcuts

Key	Action
Enter	Send message
Shift+Enter	New line in input
@	Open lore mention autocomplete
Arrow Up/Down	Navigate mention dropdown
Enter / Tab	Select mention
Escape	Dismiss mention dropdown

Learn these. The chat workflow is keyboard-heavy, and clicking through menus for every mention slows everything down.

What’s next

Ishvana (agent) The orchestrator — your main assistant and the agent you talk to first.

Writers' Room The deeper explanation of the multi-agent simultaneous-response mode.

Pinning Context How pinning lore entries and documents works across conversations.

Councils Multi-round agent debates for deeper exploration of worldbuilding questions.