Fixing gen_answer failover leaving raw_answer blank

[jamesbraza]

#1073 was suboptimal in that if the EmptyDocsError route was taken, the PQASession's raw_answer/answer_reasoning would be left unpopulated without patching in-place. This PR:

• Keeps the specialization in the returned tool response message
• While allowing the code to populate raw_answer/answer_reasoning in a normal flow

[Copilot]

Pull Request Overview

This PR fixes an issue where the EmptyDocsError failover in gen_answer was leaving raw_answer blank in the PQASession. The changes ensure that when no papers are available, the answer generation process still populates these fields through the normal flow while maintaining specialized error messages in tool responses.

Key changes:

• Removes early return from gen_answer tool when no docs are present
• Updates answer generation logic to handle empty contexts with specific error messages
• Refactors test to properly verify tool response content

Reviewed Changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File	Description
src/paperqa/agents/tools.py	Removes early EmptyDocsError check to allow normal flow processing
src/paperqa/docs.py	Updates empty context detection and error message formatting
src/paperqa/prompts.py	Adds constant for measuring empty context length
tests/test_agents.py	Refactors test to spy on tool responses instead of mocking dependencies

---

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

[jamesbraza]

Given empty context_str is 11-chars, we couldn't hit this case actually

[jamesbraza]

Note that state.session.raw_answer was left untouched here. This is fine in normal gen_answer use, but is not okay in the truncation's gen_answer failover. So this is a bug this PR fixes

[mskarlin]

lgtm

[dosubot]

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

^{How did I do? Any feedback?}