CLI Help

Quick Reference

• Entrypoint: gemini-run.py
• Primary Argument: request (string)
• Output: Validated plan or execution results

Commands

python3 gemini-run.py "[request]"

Executes a complete agentic workflow for the given request.

Arguments:

• request: The natural language instruction for the orchestrator.

Options:

• --db-path: Path to the SQLite memory database (Default: storage/memory.db).
• --log-file: Path to the orchestrator log (Default: observability/orchestrator.log).

Workflow Outputs

When you run a command, the system outputs status updates for each phase:

1. [LOAD_MEMORY]: Context retrieval.
2. [DEBATE_START]: Reasoning rounds.
3. [EXECUTION_START]: DAG task processing.
4. [WORKFLOW_END]: Final results and audit.

Troubleshooting

Error: 429 Resource Exhausted

This means the primary model (Gemini) has reached its quota. The system will automatically attempt fallback to Mistral. No action needed unless both fail.

Error: JSONDecodeError

Commonly caused by malformed LLM responses. The system uses a multi-pass robust parser to mitigate this. If it persists, try refining your prompt for clarity.

---

See Also

• Deployment Guide
• Workflow Management