diagnose_failure

Diagnose ML training or deployment failures using knowledge base evidence. Returns diagnosis, fix steps, and prevention advice.

When to Use

Error diagnosis: Root cause analysis of errors, crashes, or unexpected behavior
Config issues: Debugging configuration problems or dependency conflicts
Unexpected behavior: Understanding why a framework behaves differently than expected

Parameters

Parameter	Required	Type	Description
`symptoms`	Yes	string	Description of the failure symptoms
`logs`	Yes	string	Relevant log output or error messages

Example

Tool: diagnose_failure
Symptoms: "QLoRA training OOM on A100 40GB with 7B model, batch size 4"
Logs: "RuntimeError: CUDA out of memory. Tried to allocate 2.00 GiB..."

Returns: Diagnosis (likely cause), fix steps (e.g., enable gradient checkpointing, reduce batch size), and prevention advice.

Getting Started

Connect to Your Agents

Tools

Architecture

Support

When to Use

Parameters

Example

​When to Use

​Parameters

​Example

When to Use

Parameters

Example