Add Dapr's Conversation API as LLM inference provider #8
Conversation
Signed-off-by: yaron2 <[email protected]>
Signed-off-by: yaron2 <[email protected]>
elena-kolevska
left a comment
elena-kolevska
left a comment
There was a problem hiding this comment.
In the meantime I opened a PR with Conversation API support in the python SDK (dapr/python-sdk#787). If we can get this approved soon I believe it will be easier to just use DaprClient, it handles all kinds of things for us out of the box (tls, retries, healthcheck, all supported env variables...).
Can we get this is for 1.15? |
|
Yes, it's last moment, but we're still on time. The PR is ready for review (with full test coverage and docs). I'd like to have at least one more person give it a review and get to an approval. |
Signed-off-by: yaron2 <[email protected]>
|
Updated with the usage of the Dapr Client |
elena-kolevska
left a comment
elena-kolevska
left a comment
There was a problem hiding this comment.
Just two nits, but overall lgtm
dapr_agents/llm/dapr/client.py
Outdated
| """ | ||
| Initializes and returns the Dapr Inference client. | ||
| """ | ||
| config: DaprInferenceClientConfig = self.config |
There was a problem hiding this comment.
The config var doesn't seem to be used.
dapr_agents/llm/dapr/chat.py
Outdated
| prompt_template = Prompty.to_prompt_template(prompty_instance) | ||
|
|
||
| # Extract the model configuration from Prompty | ||
| model_config = prompty_instance.model |
Signed-off-by: yaron2 <[email protected]>
^ Title