Create agent scaffold: src/nodes/analyst_agents/benchmark

[matvix90]

Description
Add a new agent to handle the creation and evaluation of complex benchmarks.
For example, if the create_portfolio agent creates a portfolio composed of 70% stocks and 30% bonds with a set of different ETFs, it would be interesting if the benchmark agent created a benchmark like this:

• 1 global equity ETF for a total of 70% of the benchmark
• 1 bond ETF for a total of 30% of the benchmark

Why
Centralize benchmark logic for portfolio comparisons and reproducible performance tests.

Acceptance criteria

• Directory created: src/nodes/analyst_agents/benchmark
• Includes scaffold: README.md, init.py, agent.py (interface), tests/test_benchmark_agent.py
• Agent supports: loading benchmarks, running evaluations, returning metrics

Testing

• Unit tests verify agent can load a sample benchmark and return expected metric keys.

[Mmadan128]

can you assign this issue to me

[RangerCreaky]

Hi @matvix90 I have worked similar on the previous task of benchmarks, so can I pick this up?

[matvix90]

@RangerCreaky since you now have some experience with the project, you'd like to try coordinating with @Mmadan128 , creating a sub-project and dividing the development work.
What do you think?

[RangerCreaky]

@matvix90 Sure.. I would definitely like to co-ordinate!
@Mmadan128 will it be ok with you?

[Mmadan128]

@RangerCreaky sure it is fine with me too

[matvix90]

@RangerCreaky @Mmadan128 ok divide the tasks and decide how to proceed.
The issue is yours.

[RangerCreaky]

Hey @matvix90 sorry I was away for a few days.
So I am thinking of approaching it in the below way:

Currently we are taking a benchmark from the user right..
I was thinking we could keep it as is.

Then the portfolio given by create_portfolio will be given to the new create_benchmark agent, along with the chosen benchmark by the user. And the benchmark agent is supposed to evaluate whether the chosen benchmark is good or not (or if it could be mixed with other benchmarks..)

Or would you like to redact the benchmark flag in cli and just have the agent?

[matvix90]

Hi @RangerCreaky @Mmadan128, since the project's goal is to make investing as easy as possible, I would do the following:

• Create a CLI flag (set-benchmark) that, when enabled, allows the benchmark to be inserted, as is currently the case. Otherwise, the benchmark is created by the create_benchmark agent (default).
  I want the create_benchmark agent to be the default because we assume an inexperienced investor might not know what a benchmark is.
• Create robust logic for how the agent should create the benchmark.