Features Overview

A complete tour of what LexiChat can do for you — out of the box, no cloud required.

chat Chat & Streaming Responses

Core

Every response streams token-by-token directly from the local Ollama server — there's no buffering and no waiting for a full response before you start reading. The chat interface renders Markdown (headings, bold, code blocks, lists) as it arrives.

  • check_circle Full Markdown rendering with syntax-highlighted code blocks
  • check_circle Copy individual messages or entire conversations
  • check_circle Full conversation history maintained across the session
  • check_circle New Chat button to clear context and start fresh
  • check_circle Stop generation mid-stream

folder_open File Tools

Built-in tools

LexiChat can read, write, and list files on your machine. You control which folders it has access to in Settings — by default it can only access directories you explicitly allow.

read_file
Read any file in your allowed directories. Great for asking the AI to summarise, translate, or analyse a document.
write_file
Have the AI draft and save a document, create a script, or save output to a file. You choose the destination folder.
list_directory
Browse folder contents so the AI can find the right file to read or reference in its answer.
lock
Sandboxed access: File tools are strictly limited to the directories you configured. The AI cannot read or write outside those paths — any attempt is blocked at the Rust layer.

image Vision / Image Analysis

Requires a vision-capable model

Attach images directly to your message using the paperclip icon. With a vision model like llava or gemma3, the AI can describe, analyse, extract text from, or answer questions about the image.

  • check_circle Supports JPEG, PNG, WebP, GIF
  • check_circle Multiple images per message
  • check_circle Images are sent locally to Ollama — never uploaded to any cloud
  • check_circle Preview thumbnails shown before sending

person Profiles & Personas

Settings → Profiles

Profiles let you instantly switch the AI's persona, default model, system prompt, and parameter defaults. Create a profile for each context you work in:

💻
Developer
Technical system prompt, code-optimised model, long context window, Precise style
✍️
Writer
Creative writing persona, higher temperature, longer responses, Creative style
📊
Analyst
Data-focused prompt, file access enabled, Precise style, medium responses
🌍
Translator
Multilingual system prompt, balanced style, translation-focused model

Each profile stores its own system prompt, allowed file directories, default model, and chat parameter defaults. Switch profiles instantly from the top bar.

tune Model & Parameter Control

Chat params button · Settings → Defaults

Fine-tune how the AI responds with a two-tier parameter system — simple presets for everyday use, raw numbers for power users.

Tier 1 — Quick presets
Response Style
Precise — factual, low creativity
Balanced — default
Creative — higher variance
Response Length
Short — concise answers
Medium — balanced
Long — comprehensive
Auto — model decides
Memory
Standard — model default
Extended — 8 K context
Tier 2 — Advanced (click "Advanced settings…")
Temperature0–2
Top-P0–1
Top-K1–100
Repeat Penalty0.5–2
Seedinteger
Context size (num_ctx)tokens
Max tokens (num_predict)tokens
System prompt overridetext

menu_open System Tray & Background Operation

All platforms

LexiChat runs as a persistent background application on all platforms. Closing the window hides it rather than quitting — the app continues running in the menu bar (macOS), system tray (Windows), or notification area (Linux).

macOS
Icon in the menu bar, top right. Left-click toggles the window.
Windows
Icon in the system tray notification area. Same click behaviour.
Linux
Icon in the system tray, desktop environment permitting.
Why this matters for scheduled jobs: Because the app stays running when you close the window, scheduled jobs fire at their configured times even while you're doing other things. The tray icon tooltip shows how many active jobs are scheduled.

schedule Scheduled Jobs

Automation

The Jobs panel (clock icon in the toolbar) lets you create prompts that run automatically. Each job is profile-specific — it carries a snapshot of the profile's tools so it runs correctly regardless of which profile you have active in the main chat.

Schedule types

Daily — runs once per day at a specific HH:MM time.
Weekly — runs once per week on a chosen day and time.
Interval — runs every N hours. New jobs with this type trigger within 30 seconds of being saved.
Manual — only runs when you click ▶ Run. Useful for on-demand recurring tasks.

Profile & tool snapshot

When you select a profile for a job, the form shows a three-section tool panel — Built-in tools, OpenAPI services, and MCP servers. You can deselect any tool you don't want this job to use. Clicking Save Job embeds a snapshot of your selections. The snapshot timestamp is shown on each job row; an orange ⚠ stale warning appears after 7 days, prompting you to refresh.

Run history & debug trace

Every run is recorded in the History tab with its start time, finish time, duration, and status. Expanding a run reveals the full Debug trace — each agent step, every tool call with its arguments and result, and the final output rendered as Markdown. A copy button lets you grab the output directly.

Read the full Scheduled Jobs guide →

save Save Responses

Export

Every completed assistant response has a ··· menu button next to the Copy button. Click it to save the response to a file in one of three formats:

.txt
Plain text — the raw Markdown source as typed by the model.
.pdf
PDF document with word-wrapped text layout.
.docx
Word document with full Markdown formatting preserved — headings, bold, italic, bullet lists, code blocks, and blockquotes all become proper Word styles.

A native OS save dialog opens so you can choose the destination and filename. Scheduled job outputs can also be auto-saved to a file — set the path in the job's Output File field.

shield Privacy by Design

Core principle

LexiChat is built around local-first principles. Your conversations, files, and data stay on your machine.

check_circle
Models run locally via Ollama — no API keys required, no cloud inference
check_circle
Conversations are not stored or logged anywhere beyond your RAM/session
check_circle
Web search uses DuckDuckGo with no tracking or user profiling
check_circle
File access is sandboxed to explicit directories you choose — no silent reads