alikon-art/DeterminFlow
DeterminFlow is pitching a production AI workflow runtime built around versioned nodes, retries, recovery, audit trails, and narrower agent context.
The project says it has already been validated in a real AI novel-writing pipeline at 笔枢写作, where one completed task used 11 separate model sessions and 176,584 tokens. It claims node-level context isolation could cut token use by about 70% to 89% versus a long-chain single-agent setup, based on its workflow ledger and estimated repeated-context overhead. The framework includes Agent, Script, Approval, and Subprocess nodes, with FastAPI, a React console, cron automation, WebSocket events, checkpoints, and plugin packaging. GitHub · LLM repos' note.
The project says it has already been validated in a real AI novel-writing pipeline at 笔枢写作, where one completed task used 11 separate model sessions and 176,584 tokens. It claims node-level context isolation could cut token use by about 70% to 89% versus a long-chain single-agent setup, based on its workflow ledger and estimated repeated-context overhead. The framework includes Agent, Script, Approval, and Subprocess nodes, with FastAPI, a React console, cron automation, WebSocket events, checkpoints, and plugin packaging. GitHub · LLM repos' note.
score 4