AinoWork is an AI-native collaborative platform — a developer-oriented AI workbench. It integrates AI agent conversations with Docker-sandboxed code execution, delivering a complete workflow from dialogue to delivery.
Multi-iteration agent mode that breaks down complex tasks into verifiable goals with automatic progress tracking, retry logic, and structured execution loops. Includes GoalStatus lifecycle management, optimistic locking for race-condition-free round acquisition, and goal snapshots in conversation responses with detail panel integration.
New llm_billing module for comprehensive token usage tracking and visualization. Features stacked bar charts with 3-layer monochromatic gradient (cached tokens, cache-miss input tokens, and output tokens), per-organization usage dashboards, user-level drill-down for administrators, and member self-service usage pages. Records token consumption per conversation, model, and user. SSE summarizer token attribution for title and round summary generation.
Full-featured Model Context Protocol server management: CRUD operations with dedicated two-column layout, automatic tool discovery with live test and preview, JSON import for Claude Desktop format, badge-based tool display, and structured config with JSON Schema validation. MCP connections run inside the Docker sandbox for improved network isolation, with tool discovery and invocation delegated to sandbox-resident scripts.
New conversation memory module providing SessionDB-based browsing of agent conversation history via an AinoTree directory view. Groups sessions by project and schedule task, with batch delete support at project and schedule levels, plus session-task linking for unified navigation.
Round-level rollback that can undo the last completed round via SQLite rewind. Soft-deletes undone messages and rounds in PostgreSQL, resets the conversation state, and restores the composer with the undone user input for easy retry. Confirmation is provided by an AinoPopconfirm on the round divider.
Batch delete for schedule task executions via the API, and standardized confirmation dialogs for deletion operations.
- Supermemory-server multi-architecture Docker image build (amd64 + arm64)
- Docker build scripts updated with
--localflag and improved usage instructions - Sandbox MCP service discovery consolidated into a unified
sandbox_infoentry point - Hermes Agent platform adapter customized with AinoWork-specific identity and tooling
- GitHub (English) and Gitee (Chinese) issue templates for bug reports and feature requests
- Database migrations consolidated: post-v1.0.1 changes merged into a single v1.1.0 migration
- Partial unique indexes (
WHERE deleted_at IS NULL) added to core tables for soft-delete safety
- Streaming content rendering lag reduced via
useDeferredValuefor deferred updates during rapid SSE delta events - DeepSeek prefix cache preserved by separating volatile context from the system prompt, keeping the system prompt byte-identical across sessions
- Branch selector in the skill marketplace repository modal now uses debounced search with epoch-based stale response guards to eliminate race conditions
| Image | Tag | Pull Command |
|---|---|---|
| ainowork-backend | 1.1.0, latest |
docker pull registry.cn-zhangjiakou.aliyuncs.com/oinone/ainowork-backend:1.1.0 |
| ainowork-frontend | 1.1.0, latest |
docker pull registry.cn-zhangjiakou.aliyuncs.com/oinone/ainowork-frontend:1.1.0 |
| ainowork-sandbox | 1.0.1, latest |
docker pull registry.cn-zhangjiakou.aliyuncs.com/oinone/ainowork-sandbox:latest |
| alpine-socat | 1.8.1.1 |
docker pull registry.cn-zhangjiakou.aliyuncs.com/oinone/alpine-socat:1.8.1.1 |
| postgres | 16-alpine |
docker pull registry.cn-zhangjiakou.aliyuncs.com/oinone/postgres:16-alpine |
| redis | 8-alpine |
docker pull registry.cn-zhangjiakou.aliyuncs.com/oinone/redis:8-alpine |
126 commits since v1.0.1.
Full Changelog · License: AGPL-3.0