Bug Description
agents do not properly store context history in their analysis. analysis is often not there or minimal, which, when truncating responses to preserve tokens, causes issues. would be better to force them to ALWAYS include detailed analysis and repeat / store learned info (or to force this process by logging all analysis blocks and feeding this in regardless of truncation or soemthing.
Steps to Reproduce
Expected Behavior
Actual Behavior
Screenshots
Environment
- OS:
- Python Version:
- Cairn Version/Commit:
- Model Provider:
Additional Context
Possible Solution
Bug Description
agents do not properly store context history in their analysis. analysis is often not there or minimal, which, when truncating responses to preserve tokens, causes issues. would be better to force them to ALWAYS include detailed analysis and repeat / store learned info (or to force this process by logging all analysis blocks and feeding this in regardless of truncation or soemthing.
Steps to Reproduce
Expected Behavior
Actual Behavior
Screenshots
Environment
Additional Context
Possible Solution