Broken tool calls, hallucinating loops, memory leaks, bad prompts I diagnose and fix Hermes-based agents fast so you can ship.
WHAT I OFFER
- Infinite reasoning loops where the agent thinks forever and never acts
- Memory and context window overflow in long conversations
- System prompt / instruction following failures
- RAG retrieval producing irrelevant or hallucinated answers
- Multi-agent coordination issues (wrong agent routing, missed handoffs)
- Slow inference and performance bottlenecks on Ollama / vLLM
- Hermes function schema mismatches causing silent failures
- Agent producing inconsistent outputs on repeated identical inputs
- Broken integrations with external APIs, databases, or vector stores