GitHub Issue: Cron job fails to expand ${ENV_VAR} from .env file

Bug Description

When running a cron job without specifying a model, environment variables (e.g., ${HERMES_MODEL}, ${OPENROUTER_FALLBACK_MODEL}) from .env are not expanded. The literal string ${VAR} is passed to the API, causing a 400 error.

Steps to Reproduce

1. Set variables in ~/.hermes/.env:

   HERMES_MODEL=some-model
   OPENROUTER_FALLBACK_MODEL=some-fallback-model
   

2. Create a cron job WITHOUT specifying model:

   hermes cron create --name "test" --schedule "0 3 * * *" --prompt "test"

3. Run: hermes cron run <job_id>

4. Check logs: ~/.hermes/logs/agent.log

Expected Behavior

${HERMES_MODEL} should expand to some-model, and ${OPENROUTER_FALLBACK_MODEL} to some-fallback-model.

Actual Behavior

Logs show:

Fallback activated: ${HERMES_MODEL} → ${OPENROUTER_FALLBACK_MODEL}
...
Error: '${OPENROUTER_FALLBACK_MODEL}' is not a valid model ID

Root Cause (Speculation)

The .env file uses ${VAR} syntax, but the cron execution path may not apply expansion. The cli.py has "Expand ${ENV_VAR}" logic, but it may not be applied in cron context.

Workaround

Specify model explicitly (but loses flexibility):

hermes cron create ... --model "some-model" --provider "openrouter"

Environment

• Hermes Version: v0.11.0
• OS: Linux
• Python Version: 3.x

Additional Context

• Cron jobs run with cronjob, messaging, and clarify toolsets disabled
• The issue also affects cron jobs that rely on other environment variables defined in .env
• Similar to issue [Bug]: required_environment_variables are not passed through correctly for remote-backed sessions (daytona, docker, etc.) #3433 (environment variable handling in remote sessions)

Labels

bug, cron, environment-variables