Retrying Context creation once

[jamesbraza]

This PR fixes an issue where, if the LLM poses bad JSON, it would crash gather_evidence. Now:

1. First the Context creation is retried, with details on the prior failure
2. Upon second failure, the Context is ditched, with a helpful logger.exception

It also takes care to index all LLM costs across retries.

[dosubot]

Related Documentation

Checked 1 published document(s). No updates required.

^{How did I do? Any feedback?}  

[Copilot]

Pull Request Overview

This PR implements retry logic for Context creation when LLMs produce malformed JSON, preventing crashes in the gather_evidence function.

• Adds retry mechanism for Context creation with failure details passed to second attempt
• Creates proper error handling that abandons Context creation after two failures with helpful logging
• Ensures all LLM costs are tracked across retry attempts

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.

File	Description
src/paperqa/docs.py	Updates result handling to process lists of LLM results and filter out None contexts
src/paperqa/core.py	Implements retry logic with new exception class and refactored function structure

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[whitead]

Any ideas for a test? Maybe a mock llm that throws bad JSON on first call and valid JSON on second?

[jamesbraza]

Any ideas for a test? Maybe a mock llm that throws bad JSON on first call and valid JSON on second?

Sure, I just created patches to test this in both non-JSON and JSON mode in test_paperqa.py, both JSON failures and score failures

[dosubot]

Documentation Updates

Checked 1 published document(s). No updates required.

^{How did I do? Any feedback?}