All-in-One Agent Platform for Global × China Enterprises with unified connector model bridging global SaaS and China-native systems, featuring dynamic DAG planning, progressive disclosure meta-tools, visual workflow editor, and multi-channel delivery.
FIM One is an AI Agent orchestration platform designed for multinational enterprises, solving the "dual ecosystem" fragmentation between global SaaS (Slack, Teams, Salesforce, Workday) and China-native systems (Feishu, WeCom, DingTalk, DM, KingbaseES, etc.). Through standardized connector bridge layers and BaseChannel abstraction, the Agent core remains agnostic to underlying system differences.
The platform offers three delivery modes—Standalone (independent AI assistant), Copilot (embedded UI iframe/widget), and Hub (central cross-system orchestration)—sharing the same Agent core. The planning engine supports dynamic DAG (LLM-generated dependency graphs at runtime, asyncio concurrency, auto-replanning up to 3 rounds) and ReAct modes, with AUTO_ROUTING for automatic selection. The Agent Harness includes ContextGuard (5-layer token budget management), progressive disclosure meta-tools (80%+ token reduction), and self-reflection loops. The Hook system runs deterministic logic outside the LLM loop, with FeishuGateHook as the first implementation for approval gating.
Workflow capabilities include a visual drag-and-drop editor with 12 node types (React Flow v12) and JSON import/export. File processing supports visual-aware PDF/DOCX/PPTX parsing with intelligent text-vs-scan page detection. The RAG pipeline integrates Jina embedding + LanceDB + hybrid retrieval + Reranker with inline citations. Tool execution supports Python/Node.js/Shell with optional Docker sandbox isolation.
Platform features include multi-tenancy (JWT + org isolation), Marketplace (publish/subscribe Agents, connectors, knowledge bases, skills, workflows), global SOPs, 6-language UI (EN/ZH/JA/KO/DE/FR), and dark/light themes. Feishu channel is currently implemented; Slack/Teams/WeCom/email are on the v0.9 roadmap. Licensed under FIM One Source Available License—internal use and modification allowed; multi-tenant SaaS hosting and competitive use prohibited.
Architecture Overview
| Layer | Technology |
|---|---|
| Backend | Python 3.11+, FastAPI, SQLAlchemy, Alembic, asyncio |
| Frontend | Next.js 14, React 18, Tailwind CSS, shadcn/ui, React Flow v12 |
| AI / RAG | OpenAI-compatible LLMs, Jina AI (embed + search), LanceDB |
| Database | SQLite (dev) / PostgreSQL (production) |
| Messaging | BaseChannel abstraction, Fernet encryption, HMAC signature verification |
| Infrastructure | Docker, uv, pnpm, SSE streaming |
Quick Start (Docker)
git clone https://github.com/fim-ai/fim-one.git
cd fim-one
cp example.env .env
# Edit .env: set LLM_API_KEY
docker compose up --build -d
# Visit http://localhost:3000, create admin account on first launch
Connector Ecosystem
- Global SaaS: Slack, Microsoft Teams, Salesforce, Workday (via API connectors)
- China-native systems: Feishu, WeCom, DingTalk (via messaging channels + API connectors)
- Databases: PostgreSQL, MySQL, Oracle, SQL Server, DM, KingbaseES, GBase, Highgo
- Protocols: MCP Server, OpenAPI spec import
Typical Use Cases
- Legal & Compliance: Auto-extract contract clauses, compare versions, flag risks, trigger approvals
- IT Operations: Agent pulls multi-system logs, performs root cause analysis, pushes diagnostics via IM
- Business Operations: Scheduled data summaries to group chats + natural language ad-hoc queries
- Developer Integration: Import API docs to quickly generate authenticated tools