MartBakler

Suppressing userwarnings because of pydantic having a protected namespace with model_. We can't use ConfigDict because that isn't compatible with lower pydantic versions

**Is your feature request related to a problem? Please describe.** Currently tanuki only supports OpenAI finetuning. We should also support anyscale **Describe the solution you'd like** Tanuki needs a way...

**Is your feature request related to a problem? Please describe.** We need to support reading and saving data to S3 rather than only disk support **Describe the solution you'd like**...

**Is your feature request related to a problem? Please describe.** We want to generalise the package by supporting local OSS models as students or teachers **Describe the solution you'd like**...

Currently validator.check_type does not support enum outputs types. If LLM outputs a valid integer, check_type outputs False where it should put True ``` class RequestIntent(Enum): REFUNDS = 1 PRODUCT_QUESTIONS =...

During application startup for the todolist example (examples/todolist/backend/main.py) the following error occurs Error `monkey-patch\src\assertion_visitor.py", line 254, in extract_output raise NotImplementedError(f"Function {func_name} not handled yet") NotImplementedError: Function TodoItem not handled yet`

Reduce the token count

5

Currently we are sending the full manual (API description, entities, manual for every tool, few-shot examples of tool usage) as the system prompt to the agents. This ends up blowing...

Currently we only support APIs in the LLM factory. We should include supporting VLLM and local models as well

Run evaluations with decreasing number of context tokens

5

We should run experiments to determine how much performance increases with less irrelevant tokens in the context window. This would help understand if and how much performance decreases due to...

We should be able to run multiple tasks one after another, following a curriculum learning approach