gemini-cli - 💡(How to fix) Fix Track Claude provider path operational risks from review [1 participants]

• Failure mode is not explicitly handled, so the caller can get an opaque runtime failure instead of a structured provider error.

TL;DR

Switching to stdin or temp-file handoff for large Claude prompts and symmetrically recording fallback-provider outcomes are likely fixes.

Guidance

• To address the risk of hitting OS argument-length limits, consider modifying the fusion_gate.py script to pass large prompts via stdin or a temporary file instead of as CLI arguments.
• Implement error handling for the case where the OS argument-length limit is exceeded to provide a structured provider error instead of an opaque runtime failure.
• Record fallback-provider execution, health, and usage outcomes in the same manner as the primary provider to ensure accurate auditing and prevent hiding repeated failures.
• Review the current logging and auditing mechanisms to ensure they can handle the additional data from fallback-provider outcomes.

Example

import subprocess

# Instead of passing the prompt as an argument
# subprocess.run(['claude', prompt], ...)

# Pass the prompt via stdin
with subprocess.Popen(['claude'], stdin=subprocess.PIPE, stdout=subprocess.PIPE) as process:
    output, _ = process.communicate(input=prompt.encode())

Notes

The above example assumes that the claude command supports reading input from stdin. If not, a temporary file handoff may be necessary. Additionally, the exact implementation details may vary depending on the specific requirements and constraints of the fusion_gate.py script.

Recommendation

Apply workaround: Switch to stdin or temp-file handoff for large prompts and record fallback-provider outcomes symmetrically, as this addresses the identified operational risks and provides a more robust solution.