Skip to content

Fix: Resolve Agentic failure loop with improved json parsing #1541

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 4 commits into from
Oct 22, 2024

Conversation

ahgraber
Copy link
Contributor

Fixes #1538

  • Add final instruction for better json schema adherence and reduce "filler" text
  • Add extract_json function to identify json by pairs of [] or {}
  • Add unit tests for extract_json

Notes:
In my testing, these changes vastly reduce the number of times the 'repair' agent is triggered. However, this does not fix all parse errors:

  1. In some cases, json is valid json, but keys are invalid
  2. In some cases, especially with more complex Pydantic models (e.g., Themes, which is composed of List[Theme]), the Pydantic json schema includes $defs. As a result, some LLMs will return $defs or include "$schema": "https://json-schema.org/draft/...; either of these additional keys will cause the Pydantic parser to fail.

@dosubot dosubot bot added the size:L This PR changes 100-499 lines, ignoring generated files. label Oct 20, 2024
@shahules786 shahules786 requested a review from jjmachan October 21, 2024 14:06
Copy link
Member

@jjmachan jjmachan left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

thanks a lot @ahgraber for this PR ❤️
will be releasing it today

@jjmachan jjmachan merged commit 8deca92 into explodinggradients:main Oct 22, 2024
16 checks passed
@ahgraber ahgraber deleted the fix/output-json-parsing branch October 26, 2024 15:47
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
size:L This PR changes 100-499 lines, ignoring generated files.
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Expensive agentic failure loop
2 participants