I Ran 6 n8n AI Workflows on a 1.8B Local Model: 192 Runs, 0 Broken JSON, 7 Fake Hot Leads

说明与用途

在本地 1.8 B CPU 模型上运行 6 个 n8n 工作流，JSON 输出均有效，但 4 个工作流产生了错误内容。守护节点捕获了许多错误，但并非全部。

问题

Running n8n AI workflows on a small local 1.8B model produced many content failures while JSON output remained valid.

结果

0/192 responses were invalid JSON; however, 4/6 workflows produced incorrect content (e.g., lead scoring contradictions, wrong invoice dates, fabricated FAQ answers). Guard nodes mitigated many but not all issues.

工作流

• 1. Configure n8n with local Ollama endpoint
• 2. Run each workflow 3 times per test case
• 3. Capture raw and guarded outputs via Python script
• 4. Evaluate failures per workflow
• 5. Apply guard nodes to validate content
• 6. Analyze results and propose fixes

证据

0 of 192 responses were invalid JSON.

来源

Score and tier disagreed in 13/18 raw runs.

来源

Wrong date 5/18, wrong total 3/18, made‑up invoice number 3/18

来源

Answered a question that isn't in the knowledge base with a confident "No, we don't" 3/3 times

来源