add model config #30
add model config #30
Conversation
m-y-mo
requested review from
JarLob,
kevinbackhouse,
p- and
sylwia-budzynska
as code owners
There was a problem hiding this comment.
Pull Request Overview
This PR adds support for model configuration files that allow taskflows to specify models using aliases (e.g., gpt_latest) instead of hardcoded version strings. It also fixes a bug where the model parameter specified in a taskflow was not being passed to the agent deployment.
Key changes:
- Added model configuration file support with a
model_configfiletype - Fixed bug where taskflow-specified models were not passed to the CAPI
- Updated taskflow example to use model alias instead of hardcoded version
Reviewed Changes
Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.
| File | Description |
|---|---|
| configs/model_config.yaml | New configuration file defining model aliases (gpt_latest, sonnet_latest, etc.) |
| available_tools.py | Added model_config dictionary and parsing logic for model_config filetype |
| main.py | Added model config resolution logic and fixed model parameter passing to agents |
| taskflows/CVE-2023-2283/CVE-2023-2283.yaml | Updated to reference model config and use gpt_latest alias |
Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.
| model = task_body.get('model', DEFAULT_MODEL) | ||
| if model in model_keys: | ||
| model = model_dict[model] |
There was a problem hiding this comment.
Variable model_keys is referenced before it may be defined. If model_config is not provided or model_dict is empty, model_keys will be undefined, causing a NameError. Initialize model_keys as an empty list before the conditional blocks or move the model resolution inside the if model_dict: block.
| models: | ||
| sonnet_default: claude-sonnet-4 | ||
| sonnet_latest: claude-sonnet-4.5 | ||
| gpt_default: gpt-4.1 | ||
| gpt_latest: gpt-5 No newline at end of file |
There was a problem hiding this comment.
The models key is at the wrong indentation level. Based on the code in main.py that accesses model_dict.get('models', {}), this key should be nested under the seclab-taskflow-agent header, not at the root level.
| models: | |
| sonnet_default: claude-sonnet-4 | |
| sonnet_latest: claude-sonnet-4.5 | |
| gpt_default: gpt-4.1 | |
| gpt_latest: gpt-5 | |
| models: | |
| sonnet_default: claude-sonnet-4 | |
| sonnet_latest: claude-sonnet-4.5 | |
| gpt_default: gpt-4.1 | |
| gpt_latest: gpt-5 |
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
|
smoke test |
|
Add
model_configfile so that models can be specified with*-latestetc. instead of a specific version. Also fix a bug thatmodelspecified in a taskflow is not currently passed to the CAPI.