feat: add ability to specify guardrails model #135

[bahdan111]

Applicable issues

Resolves #135

Description of changes

Add separate llm_model_config to OutOfScopeConfig for guardrails, allowing independent model configuration from the supreme agent.

Changes:

• statgpt/common/schemas/channel.py: Added llm_model_config field to OutOfScopeConfig with default LLMModelConfig
• statgpt/app/chains/out_of_scope_checker.py: Updated to use out_of_scope.llm_model_config
• configurations/clients/sample/channels.yaml: Added example config

Checklist

• Title of the pull request follows Conventional Commits specification
• Deployed and tested in a Review environment.

By submitting this pull request, I confirm that my contribution is made under the terms of the MIT license.

[Fedir-Yatsenko]

Looks good

[bahdan111]

/deploy-review

GitHub actions run: 21940016646
Environment URL: review-environment | pipeline

[bahdan111]

tested, changed the model via config and made sure change propagated

[navalnica]

tested, changed the model via config and made sure change propagated

how did you make sure changes propagated?

[bahdan111]

tested, changed the model via config and made sure change propagated

how did you make sure changes propagated?

sent question and then checked grafana logs for review env

[navalnica]

sent question and then checked grafana logs for review env

both guardrails (new LLM) and agent logs (old LLM)?

[bahdan111]

sent question and then checked grafana logs for review env

both guardrails (new LLM) and agent logs (old LLM)?

yeah, so checked that default model works, without changing the config, and then changed the config to a new model