Agents et multi-agents

1flowbase

taichuy/1flowbase

An open-source AI gateway that allows local agent clients to publish fusion-style multi-model workflows as OpenAI and Claude-compatible virtual models, providing full observability into traces, tokens, latency, and costs.

Ouvrir le dépôt officiel ↗ Page d'accueil du projet ↗
★ 257Étoiles
⑂ 19Forks
24Problèmes ouverts
RustLangue
Apache-2.0Licence
Q@project.QualityScoreScore éditorial

Captures d'écran du projet

Capture d’écran de 1flowbase Capture d’écran de 1flowbase

Vue d’ensemble

1flowbase bridges the gap between LLM gateways and workflow builders. Instead of simply routing a request to a single model, it composes multiple model and tool nodes into a workflow-backed virtual model. Local AI agents and SDKs call a single standard model endpoint, while 1flowbase executes a complex workflow behind the scenes—such as mounting vision models for screenshots, running a fusion-style multi-model review panel, or verifying output structures—and returns the final answer with complete execution logs, tool callback traces, and token consumption metrics.

Fonctionnalités clés

  • Visual workflow editor
  • Multi-node workflow orchestration
  • Virtual model endpoint publishing
  • OpenAI Responses API compatibility
  • OpenAI Chat Completions API compatibility
  • Claude-compatible Messages API support
  • Streaming response support
  • Mounted LLM tools for multimodal and branch-model workflows
  • Fusion workflow template
  • Execution logs and tool callback traces
  • Application-level token consumption statistics
  • Prompt and model configuration version history

Prérequis, installation et démarrage rapide

Linux/macOS: curl -fsSL https://raw.githubusercontent.com/taichuy/1flowbase/main/scripts/shell/docker-deploy.sh | sh

Windows PowerShell: irm https://raw.githubusercontent.com/taichuy/1flowbase/main/scripts/powershell/docker-deploy.ps1 | iex

Windows CMD: powershell -NoProfile -ExecutionPolicy Bypass -Command "irm https://raw.githubusercontent.com/taichuy/1flowbase/main/scripts/powershell/docker-deploy.ps1 | iex"

Utilisation

After deployment, access the frontend at http://127.0.0.1:3100. Use the visual workflow editor to build a multi-model workflow. Once built, publish the workflow as a virtual model endpoint using the OpenAI or Claude-compatible APIs. Configure your local agent client (e.g., Claude Code, Cline) to point to the published endpoint. The client will interact with the endpoint as if it were a standard model, while 1flowbase runs the workflow behind the scenes. You can then inspect execution logs, tool callback traces, and token consumption directly in the 1flowbase dashboard.

Compatibilité des modèles et cas d’usage

Compatible with models accessible via OpenAI Responses, OpenAI Chat Completions, and Claude-compatible Messages APIs. Specific models mentioned in workflows include GLM-5.2, DeepSeek, GLM-5V-Turbo, Gemini, and GPT vision models.

Licence et notes sur les risques

Apache-2.0

Editorial verification 2026-08-02: repository URL, owner, description, license and repository statistics were reviewed. License metadata: Apache-2.0. README was fetched for the channel draft; re-check repository dependencies, releases and model terms before production use.

Publication et maintenance

Not stated in the repository metadata

Firecrawl

firecrawl/firecrawl

★ 161,1KTypeScript

LangChain

langchain-ai/langchain

★ 143,6KPython

RAGFlow

infiniflow/ragflow

★ 86,7KGo