Files
2026-04-14 21:56:42 +08:00

25 KiB
Raw Permalink Blame History

AGENTS.md

Project Overview

Memoh is a multi-member, structured long-memory, containerized AI agent system platform. Users can create AI bots and chat with them via Telegram, Discord, Lark (Feishu), DingTalk, WeChat, Matrix, Email, and more. Every bot has an independent container and memory system, allowing it to edit files, execute commands, and build itself — providing a secure, flexible, and scalable solution for multi-bot management.

Architecture Overview

The system consists of three core services:

Service Tech Stack Port Description
Server (Backend) Go + Echo 8080 Main service: REST API, auth, database, container management, in-process AI agent
Web (Frontend) Vue 3 + Vite 8082 Management UI: visual configuration for Bots, Models, Channels, etc.
Browser Gateway Bun + Elysia + Playwright 8083 Browser automation service: headless browser actions for bots

Infrastructure dependencies:

  • PostgreSQL — Relational data storage
  • Qdrant — Vector database for memory semantic search
  • Containerd — Container runtime providing isolated environments per bot (Linux); Apple Virtualization on macOS

Tech Stack

Backend (Go)

  • Framework: Echo (HTTP)
  • Dependency Injection: Uber FX
  • AI SDK: Twilight AI (Go LLM SDK — OpenAI, Anthropic, Google)
  • Database Driver: pgx/v5
  • Code Generation: sqlc (SQL → Go)
  • API Docs: Swagger/OpenAPI (swaggo)
  • MCP: modelcontextprotocol/go-sdk
  • Containers: containerd v2 (Linux), Apple Virtualization (macOS)
  • TUI: Charm libraries (bubbletea, glamour, lipgloss) for CLI interactive mode

Frontend (TypeScript)

  • Framework: Vue 3 (Composition API)
  • Build Tool: Vite 8
  • State Management: Pinia 3 + Pinia Colada
  • UI: Tailwind CSS 4 + custom component library (@memohai/ui) + Reka UI
  • Icons: lucide-vue-next + @memohai/icon (brand/provider icons)
  • i18n: vue-i18n
  • Markdown: markstream-vue + Shiki + Mermaid + KaTeX
  • Desktop: Tauri (wraps @memohai/web)
  • Package Manager: pnpm monorepo

Browser Gateway (TypeScript)

  • Runtime: Bun
  • Framework: Elysia
  • Browser Automation: Playwright

Tooling

  • Task Runner: mise
  • Package Managers: pnpm (frontend monorepo), Go modules (backend)
  • Linting: golangci-lint (Go), ESLint + typescript-eslint + vue-eslint-parser (TypeScript)
  • Testing: Vitest
  • Version Management: bumpp
  • SDK Generation: @hey-api/openapi-ts (with @hey-api/client-fetch + @pinia/colada plugins)

Project Structure

Memoh/
├── cmd/                        # Go application entry points
│   ├── agent/                  #   Main backend server (main.go, FX wiring)
│   ├── bridge/                 #   In-container gRPC bridge (UDS-based, runs inside bot containers)
│   │   └── template/           #     Prompt templates for bridge (TOOLS.md, SOUL.md, IDENTITY.md, etc.)
│   ├── mcp/                    #   MCP stdio transport binary
│   └── memoh/                  #   Unified CLI (Cobra: serve, migrate, chat, bots, compose, docker, login, install, support)
├── internal/                   # Go backend core code (domain packages)
│   ├── accounts/               #   User account management (CRUD, password hashing)
│   ├── acl/                    #   Access control list (source-aware chat trigger ACL)
│   ├── agent/                  #   In-process AI agent (Twilight AI SDK integration)
│   │   ├── agent.go            #     Core agent: Stream() / Generate() via Twilight SDK
│   │   ├── stream.go           #     Streaming event assembly
│   │   ├── sential.go          #     Sential (sentinel) loop detection logic
│   │   ├── prompt.go           #     Prompt assembly (system, heartbeat, schedule, subagent, discuss)
│   │   ├── config.go           #     Agent service dependencies
│   │   ├── types.go            #     Shared types (StreamEvent, GenerateResult, FileAttachment)
│   │   ├── fs.go               #     Filesystem utilities
│   │   ├── guard_state.go      #     Guard state management
│   │   ├── retry.go            #     Retry logic
│   │   ├── read_media.go       #     Media reading utilities
│   │   ├── spawn_adapter.go    #     Spawn adapter for sub-processes
│   │   ├── prompts/            #     Prompt templates (Markdown, with partials prefixed by _)
│   │   │   ├── system_chat.md, system_discuss.md, system_heartbeat.md, system_schedule.md, system_subagent.md
│   │   │   ├── _tools.md, _memory.md, _contacts.md, _schedule_task.md, _subagent.md
│   │   │   ├── heartbeat.md, schedule.md
│   │   │   └── memory_extract.md, memory_update.md
│   │   └── tools/              #     Tool providers (ToolProvider interface)
│   │       ├── message.go      #       Send message tool
│   │       ├── contacts.go     #       Contact list tool
│   │       ├── schedule.go     #       Schedule management tool
│   │       ├── memory.go       #       Memory read/write tool
│   │       ├── web.go          #       Web search tool
│   │       ├── webfetch.go     #       Web page fetch tool
│   │       ├── container.go    #       Container file/exec tools
│   │       ├── fsops.go        #       Filesystem operations tool
│   │       ├── email.go        #       Email send tool
│   │       ├── subagent.go     #       Sub-agent invocation tool
│   │       ├── skill.go        #       Skill activation tool
│   │       ├── browser.go      #       Browser automation tool
│   │       ├── tts.go          #       Text-to-speech tool
│   │       ├── federation.go   #       MCP federation tool
│   │       ├── image_gen.go    #       Image generation tool
│   │       ├── prune.go        #       Pruning tool
│   │       ├── history.go      #       History access tool
│   │       └── read_media.go   #       Media reading tool
│   ├── attachment/             #   Attachment normalization (MIME types, base64)
│   ├── auth/                   #   JWT authentication middleware and utilities
│   ├── bind/                   #   Channel identity-to-user binding code management
│   ├── boot/                   #   Runtime configuration provider (container backend detection)
│   ├── bots/                   #   Bot management (CRUD, lifecycle)
│   ├── browsercontexts/        #   Browser context management (CRUD)
│   ├── channel/                #   Channel adapter system
│   │   ├── adapters/           #     Platform adapters: telegram, discord, feishu, qq, dingtalk, weixin, wecom, wechatoa, matrix, misskey, local
│   │   └── identities/        #     Channel identity service
│   ├── command/                #   Slash command system (extensible command handlers)
│   ├── compaction/             #   Message history compaction service (LLM summarization)
│   ├── config/                 #   Configuration loading and parsing (TOML + YAML providers)
│   ├── containerd/             #   Container runtime abstraction (containerd / Apple Virtualization)
│   ├── conversation/           #   Conversation management and flow resolver
│   │   ├── service.go          #     Conversation CRUD and routing
│   │   └── flow/               #     Chat orchestration (resolver, streaming, memory, triggers)
│   ├── copilot/                #   GitHub Copilot client integration
│   ├── db/                     #   Database connection and migration utilities
│   │   └── sqlc/               #   ⚠️ Auto-generated by sqlc — DO NOT modify manually
│   ├── email/                  #   Email provider and outbox management (Mailgun, generic SMTP, OAuth)
│   ├── embedded/               #   Embedded filesystem assets (web only)
│   ├── handlers/               #   HTTP request handlers (REST API endpoints)
│   ├── healthcheck/            #   Health check adapter system (MCP, channel checkers)
│   ├── heartbeat/              #   Heartbeat scheduling service (cron-based)
│   ├── identity/               #   Identity type utilities (human vs bot)
│   ├── logger/                 #   Structured logging (slog)
│   ├── mcp/                    #   MCP protocol manager (connections, OAuth, tool gateway)
│   ├── media/                  #   Content-addressed media asset service
│   ├── memory/                 #   Long-term memory system (multi-provider: Qdrant, BM25, LLM extraction)
│   ├── message/                #   Message persistence and event publishing
│   ├── messaging/              #   Outbound message executor
│   ├── models/                 #   LLM model management (CRUD, variants, client types, probe)
│   ├── oauthctx/               #   OAuth context helpers
│   ├── pipeline/               #   Discuss/chat pipeline (adapt, projection, rendering, driver)
│   ├── policy/                 #   Access policy resolution (guest access)
│   ├── providers/              #   LLM provider management (OpenAI, Anthropic, etc.)
│   ├── prune/                  #   Text pruning utilities (truncation with head/tail)
│   ├── registry/               #   Provider registry service (YAML provider templates)
│   ├── schedule/               #   Scheduled task service (cron)
│   ├── searchproviders/        #   Search engine provider management (Brave, etc.)
│   ├── server/                 #   HTTP server wrapper (Echo setup, middleware, shutdown)
│   ├── session/                #   Bot session management service
│   ├── settings/               #   Bot settings management
│   ├── storage/                #   Storage provider interface (filesystem, container FS)
│   ├── textutil/               #   UTF-8 safe text utilities
│   ├── timezone/               #   Timezone utilities
│   ├── tts/                    #   Text-to-speech provider management
│   ├── tui/                    #   Terminal UI (Charm stack for CLI interactive mode)
│   ├── version/                #   Build-time version information
│   └── workspace/              #   Workspace container lifecycle management
│       ├── manager.go          #     Container reconciliation, gRPC connection pool
│       ├── manager_lifecycle.go #    Container create/start/stop operations
│       ├── bridge/             #     gRPC client for in-container bridge service
│       └── bridgepb/           #     Protobuf definitions (bridge.proto)
├── apps/                       # Application services
│   ├── browser/                #   Browser Gateway (Bun/Elysia/Playwright)
│   │   └── src/
│   │       ├── index.ts        #     Elysia server entry point
│   │       ├── browser.ts      #     Playwright browser lifecycle
│   │       ├── modules/        #     Route modules (action, context, devices, session, cores)
│   │       ├── middlewares/     #     CORS, error handling, bearer auth
│   │       ├── types/          #     TypeScript type definitions
│   │       ├── storage.ts      #     Browser context storage
│   │       └── models.ts       #     Zod request schemas
│   ├── desktop/                #   Tauri desktop app (@memohai/desktop, wraps @memohai/web)
│   └── web/                    #   Main web app (@memohai/web, Vue 3) — see apps/web/AGENTS.md
├── packages/                   # Shared TypeScript libraries
│   ├── ui/                     #   Shared UI component library (@memohai/ui)
│   ├── sdk/                    #   TypeScript SDK (@memohai/sdk, auto-generated from OpenAPI)
│   ├── icons/                  #   Brand/provider icon library (@memohai/icon)
│   └── config/                 #   Shared configuration utilities (@memohai/config)
├── spec/                       # OpenAPI specifications (swagger.json, swagger.yaml)
├── db/                         # Database
│   ├── migrations/             #   SQL migration files (00010067+)
│   └── queries/                #   SQL query files (sqlc input)
├── conf/                       # Configuration
│   ├── providers/              #   Provider YAML templates (openai, anthropic, codex, github-copilot, etc.)
│   ├── app.example.toml        #   Default config template
│   ├── app.docker.toml         #   Docker deployment config
│   ├── app.apple.toml          #   macOS (Apple Virtualization) config
│   └── app.windows.toml        #   Windows config
├── devenv/                     # Dev environment
│   ├── docker-compose.yml      #   Main dev compose
│   ├── docker-compose.minify.yml #  Minified services compose
│   ├── docker-compose.selinux.yml # SELinux overlay compose
│   └── app.dev.toml            #   Dev config (connects to devenv docker-compose)
├── docker/                     # Production Docker (Dockerfiles, entrypoints, nginx.conf, toolkit/)
├── docs/                       # Documentation site (VitePress)
├── scripts/                    # Utility scripts (db-up, db-drop, release, install, sync-openrouter-models)
├── docker-compose.yml          # Docker Compose orchestration (production)
├── mise.toml                   # mise tasks and tool version definitions
├── sqlc.yaml                   # sqlc code generation config
├── openapi-ts.config.ts        # SDK generation config (@hey-api/openapi-ts)
├── bump.config.ts              # Version bumping config (bumpp)
├── vitest.config.ts            # Test framework config (Vitest)
├── tsconfig.json               # TypeScript monorepo config
└── eslint.config.mjs           # ESLint config

Development Guide

Prerequisites

  1. Install mise
  2. Install toolchains and dependencies: mise install
  3. Initialize the project: mise run setup
  4. Start the dev environment: mise run dev
  5. Dev web UI: http://localhost:18082 (server: 18080, browser gateway: 18083)

Common Commands

Command Description
mise run dev Start the containerized dev environment (all services)
mise run dev:minify Start dev environment with minified services
mise run dev:selinux Start dev environment on SELinux systems
mise run dev:down Stop the dev environment
mise run dev:logs View dev environment logs
mise run dev:restart Restart a service (e.g. -- server)
mise run setup Install dependencies + workspace toolkit
mise run sqlc-generate Regenerate Go code after modifying SQL files
mise run swagger-generate Generate Swagger documentation
mise run sdk-generate Generate TypeScript SDK (depends on swagger-generate)
mise run icons-generate Generate icon Vue components from SVG sources
mise run db-up Initialize and migrate the database
mise run db-down Drop the database
mise run docs Start documentation dev server
mise run build-embedded-assets Build and stage embedded web assets
mise run build-unified Build memoh CLI locally
mise run bridge:build Rebuild bridge binary in dev container
mise run desktop:dev Start Tauri desktop app in dev mode
mise run desktop:build Build Tauri desktop app for release
mise run lint Run all linters (Go + ESLint)
mise run lint:fix Run all linters with auto-fix
mise run release Release new version (bumpp)
mise run install-socktainer Install socktainer (macOS container backend)
mise run install-workspace-toolkit Install workspace toolkit (bridge binary etc.)

Docker Deployment

docker compose up -d        # Start all services
# Visit http://localhost:8082

Production services: postgres, migrate, server, web. Optional profiles: qdrant (vector DB), sparse (BM25 search), browser (browser automation).

Key Development Rules

Database, sqlc & Migrations

  1. SQL queries are defined in db/queries/*.sql.
  2. All Go files under internal/db/sqlc/ are auto-generated by sqlc. DO NOT modify them manually.
  3. After modifying any SQL files (migrations or queries), run mise run sqlc-generate to update the generated Go code.

Migration Rules

Migrations live in db/migrations/ and follow a dual-update convention:

  • 0001_init.up.sql is the canonical full schema. It always contains the complete, up-to-date database definition (all tables, indexes, constraints, etc.). When adding schema changes, you must also update 0001_init.up.sql to reflect the final state.
  • Incremental migration files (0002_, 0003_, ...) contain only the diff needed to upgrade an existing database. They exist for environments that already have the schema and need to apply only the delta.
  • Both must be kept in sync: every schema change requires updating 0001_init.up.sql AND creating a new incremental migration file.
  • Naming: {NNNN}_{description}.up.sql and {NNNN}_{description}.down.sql, where {NNNN} is a zero-padded sequential number (e.g., 0005). Always use the next available number.
  • Paired files: Every incremental migration must have both an .up.sql (apply) and a .down.sql (rollback) file.
  • Header comment: Each file should start with a comment indicating the migration name and a brief description:
    -- 0005_add_feature_x
    -- Add feature_x column to bots table for ...
    
  • Idempotent DDL: Use IF NOT EXISTS / IF EXISTS guards (e.g., CREATE TABLE IF NOT EXISTS, ADD COLUMN IF NOT EXISTS, DROP TABLE IF EXISTS) so migrations are safe to re-run.
  • Down migration must fully reverse up: The .down.sql must cleanly undo everything its .up.sql does, in reverse order.
  • After creating or modifying migrations, run mise run sqlc-generate to regenerate the Go code, then mise run db-up to apply.

API Development Workflow

  1. Write handlers in internal/handlers/ with swaggo annotations.
  2. Run mise run swagger-generate to update the OpenAPI docs (output in spec/).
  3. Run mise run sdk-generate to update the frontend TypeScript SDK (packages/sdk/).
  4. The frontend calls APIs via the auto-generated @memohai/sdk.

Agent Development

  • The AI agent runs in-process within the Go server — there is no separate agent gateway service.
  • Core agent logic lives in internal/agent/, powered by the Twilight AI Go SDK.
  • internal/agent/agent.go provides Stream() (SSE streaming) and Generate() (non-streaming) methods.
  • Model/client types are defined in internal/models/types.go: openai-completions, openai-responses, anthropic-messages, google-generative-ai, openai-codex, github-copilot, edge-speech.
  • Model types: chat, embedding, speech.
  • Tools are implemented as ToolProvider instances in internal/agent/tools/, loaded via setter injection to avoid FX dependency cycles.
  • Prompt templates are embedded Go Markdown files in internal/agent/prompts/. Partials (reusable fragments) are prefixed with _ (e.g., _tools.md, _memory.md). System prompts include system_chat.md (standard chat) and system_discuss.md (discuss mode).
  • The conversation flow resolver (internal/conversation/flow/) orchestrates message assembly, memory injection, history trimming, and agent invocation.
  • The discuss/chat pipeline (internal/pipeline/) provides an alternative orchestration path with adaptation, projection, rendering, and driver layers.
  • The compaction service (internal/compaction/) handles LLM-based conversation summarization.
  • Loop detection (text and tool loops) is built into the agent with configurable thresholds.
  • Tag extraction system processes inline tags in streaming output (attachments, reactions, speech/TTS).

Frontend Development

  • Use Vue 3 Composition API with <script setup> style.
  • Shared components belong in packages/ui/.
  • API calls use the auto-generated @memohai/sdk.
  • State management uses Pinia; data fetching uses Pinia Colada.
  • i18n via vue-i18n.
  • See apps/web/AGENTS.md for detailed frontend conventions.

Container / Workspace Management

  • Each bot can have an isolated workspace container for file editing, command execution, and MCP tool hosting.
  • Containers communicate with the host via a gRPC bridge over Unix Domain Sockets (UDS), not TCP.
  • The bridge binary (cmd/bridge/) runs inside each container, mounting runtime binaries from $WORKSPACE_RUNTIME_DIR and UDS sockets from /run/memoh/. Bridge prompt templates live in cmd/bridge/template/.
  • Container images are standard base images (debian, alpine, ubuntu, etc.) — no dedicated MCP Docker image needed.
  • internal/workspace/ manages container lifecycle (create, start, stop, reconcile) and maintains a gRPC connection pool.
  • internal/containerd/ provides the container runtime abstraction layer (containerd on Linux, Apple Virtualization on macOS, socktainer for socket-based management).
  • SSE-based progress feedback is provided during container image pull and creation.

Database Tables

The canonical source of truth for the full schema is db/migrations/0001_init.up.sql. Key tables grouped by domain:

Auth & Users

  • users — User accounts (username, email, role, display_name, avatar)
  • channel_identities — Unified inbound identity subject (cross-platform)
  • user_channel_bindings — Outbound delivery config per user/channel
  • channel_identity_bind_codes — One-time codes for channel identity → user linking

Bots & Sessions

  • bots — Bot definitions with model references and settings
  • bot_sessions — Bot conversation sessions
  • bot_session_events — Session event log
  • bot_channel_configs — Per-bot channel configurations
  • bot_channel_routes — Conversation route mapping (inbound thread → bot history)
  • bot_acl_rules — Source-aware chat access control lists

Messages & History

  • bot_history_messages — Unified message history under bot scope
  • bot_history_message_assets — Message → content_hash asset links (with name and metadata)
  • bot_history_message_compacts — Compacted message summaries

Providers & Models

  • providers — LLM provider configurations (name, base_url, api_key)
  • provider_oauth_tokens — Provider-level OAuth tokens
  • user_provider_oauth_tokens — Per-user provider OAuth tokens
  • models — Model definitions (chat/embedding/speech types, modalities, reasoning)
  • model_variants — Model variant definitions (weight, metadata)
  • search_providers — Search engine provider configurations
  • memory_providers — Multi-provider memory adapter configurations

MCP

  • mcp_connections — MCP connection configurations per bot
  • mcp_oauth_tokens — MCP OAuth tokens

Containers

  • containers — Bot container instances
  • snapshots — Container snapshots
  • container_versions — Container version tracking
  • lifecycle_events — Container lifecycle events

Email

  • email_providers — Pluggable email service backends (Mailgun, generic SMTP)
  • email_oauth_tokens — OAuth2 tokens for email providers (Gmail)
  • bot_email_bindings — Per-bot email provider binding with permissions
  • email_outbox — Outbound email audit log

Scheduling & Automation

  • schedule — Scheduled tasks (cron)
  • schedule_logs — Schedule execution logs
  • bot_heartbeat_logs — Heartbeat execution records
  • browser_contexts — Browser context configurations (Playwright)

Storage

  • storage_providers — Pluggable object storage backends
  • bot_storage_bindings — Per-bot storage backend selection

Configuration

The main configuration file is config.toml (copied from conf/app.example.toml or environment-specific templates for development), containing:

  • [log] — Logging configuration (level, format)
  • [server] — HTTP listen address
  • [admin] — Admin account credentials
  • [auth] — JWT authentication settings
  • [containerd] — Container runtime configuration (socket path, namespace, socktainer)
  • [workspace] — Workspace container image and data configuration (registry, default_image, snapshotter, data_root, cni, runtime_dir)
  • [postgres] — PostgreSQL connection
  • [qdrant] — Qdrant vector database connection
  • [sparse] — Sparse (BM25) search service connection
  • [browser_gateway] — Browser Gateway address
  • [web] — Web frontend address
  • [registry] — Provider registry (providers_dir pointing to conf/providers/)
  • [supermarket] — Supermarket integration (base_url)

Provider YAML templates in conf/providers/ define preset configurations for various LLM providers (OpenAI, Anthropic, GitHub Copilot, etc.).

Configuration templates available in conf/:

  • app.example.toml — Default template
  • app.docker.toml — Docker deployment
  • app.apple.toml — macOS (Apple Virtualization backend)
  • app.windows.toml — Windows

Development configuration in devenv/:

  • app.dev.toml — Development (connects to devenv docker-compose)

Web Design

Please refer to ./apps/web/AGENTS.md.