8000 fix: add LLM schema preparation utilities and enhance schema processi… by tonmanna · Pull Request #327 · presenton/presenton · GitHub
[go: up one dir, main page]
Skip to content

fix: add LLM schema preparation utilities and enhance schema processi… #327

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.

Sign up for GitHub

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

Jump to bottom
Open
tonmanna wants to merge 3 commits into
base: main
Choose a base branch
from
Open

fix: add LLM schema preparation utilities and enhance schema processi… #327

tonmanna wants to merge 3 commits into from

Conversation

@tonmanna
Copy link
@tonmanna tonmanna commented Oct 20, 2025 •
edited
Loading

Fix: Resolve Google Gemini "schema exceeds maximum nesting depth" error

Problem

Google Gemini API was rejecting structured output requests with a schema validation error:

google.genai.errors.ClientError: 400 INVALID_ARGUMENT
'A schema in GenerationConfig in the request exceeds the maximum allowed nesting depth.'

This occurred because our presentation schemas had deeply nested structures that exceeded Google Gemini's limits:

  • Max nesting depth: ~5 levels
  • Schema complexity: Constraints like minLength, maxLength, minItems, maxItems create "too many states"

Solution

Implemented a schema processing pipeline that:

  1. Flattens $ref references to reduce indirection
  2. Removes metadata fields (e.g., title)
  3. Strips problematic constraints:
    • String: minLength, maxLength, pattern, format
    • Array: minItems, maxItems
    • Number: minimum, maximum
    • Other: enum, const, default, examples
  4. Limits nesting depth to provider-specific maximum

Changes

New Files

  • servers/fastapi/utils/llm_utils.py - Central utility for LLM schema preparation with provider-specific helpers:
    • prepare_schema_for_google() (max_depth=4)
    • prepare_schema_for_openai() (max_depth=5)
    • prepare_schema_for_anthropic() (max_depth=5)

Modified Files

  • servers/fastapi/utils/schema_utils.py - Added schema simplification functions
  • servers/fastapi/services/llm_client.py - Applied schema processing to all providers
  • servers/fastapi/api/v1/ppt/endpoints/presentation.py - Fixed unawaited coroutine warnings

Provider Settings

Provider Max Depth Remove Constraints Debug Logging
Google 4 ✅ Yes ❌ No
OpenAI 5 ✅ Yes ❌ No
Anthropic 5 ✅ Yes ❌ No

Result

Schemas are now automatically processed before being sent to LLM providers, preventing validation errors while maintaining the necessary structure for structured output generation.

@tonmanna tonmanna mentioned this pull request Oct 20, 2025
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@tonmanna
0