DEEIX Chat is an open-source, deployable AI platform that puts multiple model providers behind one entry point. It combines multimodal chat, model routing, files and retrieval, MCP tools, usage billing, identity, audit, and operations in one application.
Where to find it#
- Web app: open your deployment URL. The default is
http://localhost:8080. - Desktop client: install from GitHub Releases. It can run a bundled local server or connect to an existing deployment. See Desktop App.
- User console: sign in and use the left navigation for New chat, Recent, Files, Knowledge bases, and Skills & Prompts.
- Admin console: open the user menu and choose Admin.
- Settings: open the user menu and choose Settings for General, Account, Chat, Subscription, and About.
What you work with#
DEEIX Chat runs as one Go service that serves the browser app and the API. You connect it to the model providers you choose, then manage routing, files, tools, billing, and identity from the admin console.
| Surface | What you do there | Guide |
|---|---|---|
| New chat | Send prompts, attach files, choose a model, configure parameters, select MCP tools and knowledge bases, and generate images or video. | New chat |
| Recent | Search, filter, star, archive, move, share, export, and delete conversations. | Recent |
| Files | Upload, preview, extract, search, rename, and delete files, and check storage usage. | Files |
| Knowledge bases | Group sources and select them in chat for retrieval. | Knowledge bases |
| Skills & Prompts | Maintain personal prompts and skills that insert into chat with /. | Skills & Prompts |
| Settings | Manage profile, appearance, chat defaults, account security, and subscription. | General |
| Admin console | Configure accounts, permission groups, upstreams, models, tools, billing, announcements, logs, moderation, login, conversation, and files. | Accounts |
Capabilities#
| Area | Capabilities |
|---|---|
| Conversations | Streaming replies, message branches, retry, edit, continue, fork, delete, like and dislike, temporary chats, voice input, message queue, conversation screenshots, sharing, and JSON export. |
| Models and routing | Platform models, upstream channels, route bindings, priority, weight, automatic circuit breaking, model kinds, vendor mapping, and context windows. |
| Protocols | OpenAI Responses and Chat Completions, OpenAI image generation and editing, Anthropic Messages, Google generateContent and image generation, Gemini interactions, xAI responses, images, video, and video extension, and OpenRouter chat, responses, and images. |
| Files and retrieval | Upload, preview, text extraction, OCR, storage quota, full-context injection, chunking, embeddings, semantic retrieval, and per-file retrieval switches. |
| Knowledge bases | Personal and built-in knowledge bases, platform sources, source availability, and vectorization. |
| Tools | MCP servers over Streamable HTTP, tool discovery and sync, tool enablement and ordering, per-message limits, image routing, native provider tools, and tool result rendering. |
| Rendering | Markdown, code, KaTeX math, Mermaid diagrams, HTML and SVG artifacts in a sandbox, interactive components, and an optional HTML visual layout prompt. |
| Context and memory | Message window, token budget, context compression, project and conversation context, AI preference memory, RAG evidence, and message process traces. |
| Billing and usage | Self-use, subscription, and usage billing; token, per-call, per-second, tiered, and cache pricing; Stripe and EPay; balances, subscriptions, redemption codes, usage logs, and statistics. |
| Identity and security | Username and email login, email registration and verification, password reset, 2FA/TOTP with recovery codes, OIDC/OAuth2 providers, active sessions, and account deletion. |
| Administration | Accounts, permission groups, upstreams, models, tools, billing, announcements, logs, content moderation, login and auth, conversation, files and retrieval, and knowledge bases. |
| Operations | Docker deployment, SQLite or PostgreSQL, in-memory or Redis cache, local or S3 storage, GeoIP, OpenTelemetry tracing, structured logs, a version endpoint, and Swagger in development. |
Clients#
The browser app is served by the same service that serves the API. The desktop client is built with Tauri 2 and can run a bundled local server or connect to a deployment. Web, desktop, and future clients share the platform-agnostic logic in packages/core and the generated API contract in packages/api-contract.
See Desktop App, Architecture, and Platform matrix.
Deployment#
DEEIX Chat ships Docker assets in deploy/. Three profiles cover a local SQLite install, an existing PostgreSQL and Redis deployment, and a single-machine stack with all three containers.
| Profile | Config file | Compose file | Dependencies |
|---|---|---|---|
| Lightweight | config.sqlite.example.yaml | docker-compose.sqlite.yml | SQLite + sqlite-vec + in-memory cache |
| Default | config.example.yaml | docker-compose.yml | External PostgreSQL + Redis |
| Full | config.full.example.yaml | docker-compose.full.yml | App + PostgreSQL + Redis |
Start with Quickstart for exact commands, then read Configuration for secrets, environment variables, and the admin console settings.