[BUG] Pipeline-as-Toolkit input is truncated when called from Agent

[epamLDadayan]

Title: Pipeline-as-Toolkit input is truncated when called from Agent

Description:
When a pipeline is added as a toolkit to an agent, the input string is mishandled and only a single character is passed to the pipeline node. Running the same pipeline directly works as expected.

Steps to reproduce:

1. Create a pipeline (Dummy_Pipeline) with a single node that uppercases the entire input string and returns “Uppercase: ”.
2. Run the pipeline directly with input “hello world” and verify the output.
3. Add the same pipeline as a toolkit to an agent (Dummy_Agent).
4. Execute the agent with the input “hello world” so it triggers the pipeline.
5. Check the final output.

Actual Result:
The agent-triggered run returns “Uppercase: D” (single-character), indicating the pipeline received a truncated input.

Expected Result:
The agent-triggered run should return “Uppercase: HELLO WORLD”, identical to the direct pipeline execution.

[marara-h]

@epamLDadayan Verified on Stage for Claude 2 Sonnet model

do not upper case the text at running pipeline for GPT-5, 4 mini

[SamvelSimonyan]

verified on stage