fix: nightly hardening - http empty-content error surfacing

[mouse-value-add]

Problem

The HTTP provider could return a successful with empty content when upstream endpoints returned unsupported payload shapes or structured error payloads. This made failures look like normal empty generations and hid root-cause diagnostics.

Approach

• Added to normalize common error fields (, , , including nested ).
• In , if extracted content is empty:
  • raise a descriptive error when an upstream error payload is present
  • otherwise raise an explicit unsupported-shape error with top-level keys
• Added focused unit tests for error extraction behavior.

Verification

• Ran:
  • ============================= test session starts ==============================
    platform darwin -- Python 3.12.13, pytest-9.0.3, pluggy-1.6.0
    rootdir: /Users/mouse/.openclaw/workspace/PromptLens
    configfile: pyproject.toml
    plugins: cov-7.1.0, anyio-4.13.0
    collected 7 items

tests/test_http_provider_empty_content_guard.py ... [ 42%]
tests/test_http_provider_response_parsing.py .... [100%]

=============================== warnings summary ===============================
promptlens/models/tools.py:14
/Users/mouse/.openclaw/workspace/PromptLens/promptlens/models/tools.py:14: PydanticDeprecatedSince20: Support for class-based config is deprecated, use ConfigDict instead. Deprecated in Pydantic V2.0 to be removed in V3.0. See Pydantic V2 Migration Guide at https://errors.pydantic.dev/2.13/migration/
class ToolParameter(BaseModel):

promptlens/models/test_case.py:10
/Users/mouse/.openclaw/workspace/PromptLens/promptlens/models/test_case.py:10: PydanticDeprecatedSince20: Support for class-based config is deprecated, use ConfigDict instead. Deprecated in Pydantic V2.0 to be removed in V3.0. See Pydantic V2 Migration Guide at https://errors.pydantic.dev/2.13/migration/
class TestCase(BaseModel):

promptlens/models/test_case.py:66
/Users/mouse/.openclaw/workspace/PromptLens/promptlens/models/test_case.py:66: PydanticDeprecatedSince20: Support for class-based config is deprecated, use ConfigDict instead. Deprecated in Pydantic V2.0 to be removed in V3.0. See Pydantic V2 Migration Guide at https://errors.pydantic.dev/2.13/migration/
class GoldenSet(BaseModel):

promptlens/models/config.py:100
/Users/mouse/.openclaw/workspace/PromptLens/promptlens/models/config.py:100: PydanticDeprecatedSince20: Support for class-based config is deprecated, use ConfigDict instead. Deprecated in Pydantic V2.0 to be removed in V3.0. See Pydantic V2 Migration Guide at https://errors.pydantic.dev/2.13/migration/
class RunConfig(BaseModel):

promptlens/providers/google.py:8
/Users/mouse/.openclaw/workspace/PromptLens/promptlens/providers/google.py:8: FutureWarning:

All support for the google.generativeai package has ended. It will no longer be receiving
updates or bug fixes. Please switch to the google.genai package as soon as possible.
See README for more details:

https://github.com/google-gemini/deprecated-generative-ai-python/blob/main/README.md

import google.generativeai as genai

-- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
================================ tests coverage ================================
______________ coverage: platform darwin, python 3.12.13-final-0 _______________

Name Stmts Miss Cover

promptlens/init.py 3 0 100%
promptlens/main.py 3 3 0%
promptlens/cli.py 169 169 0%
promptlens/exporters/init.py 6 6 0%
promptlens/exporters/base.py 15 15 0%
promptlens/exporters/csv_exporter.py 26 26 0%
promptlens/exporters/html_exporter.py 38 38 0%
promptlens/exporters/json_exporter.py 16 16 0%
promptlens/exporters/markdown_exporter.py 77 77 0%
promptlens/judges/init.py 3 3 0%
promptlens/judges/base.py 15 15 0%
promptlens/judges/llm_judge.py 71 71 0%
promptlens/judges/parser.py 107 107 0%
promptlens/judges/prompts.py 30 30 0%
promptlens/loaders/init.py 4 4 0%
promptlens/loaders/base.py 14 14 0%
promptlens/loaders/json_loader.py 23 23 0%
promptlens/loaders/yaml_loader.py 34 34 0%
promptlens/models/init.py 4 0 100%
promptlens/models/config.py 41 0 100%
promptlens/models/result.py 60 13 78%
promptlens/models/test_case.py 25 0 100%
promptlens/models/tools.py 74 34 54%
promptlens/providers/init.py 3 0 100%
promptlens/providers/anthropic.py 52 32 38%
promptlens/providers/base.py 23 6 74%
promptlens/providers/factory.py 21 10 52%
promptlens/providers/google.py 47 28 40%
promptlens/providers/http.py 84 32 62%
promptlens/providers/openai.py 57 36 37%
promptlens/providers/you.py 57 39 32%
promptlens/runners/init.py 2 2 0%
promptlens/runners/runner.py 96 96 0%
promptlens/utils/init.py 1 0 100%
promptlens/utils/cost.py 15 11 27%
promptlens/utils/diff.py 25 25 0%
promptlens/utils/retry.py 21 15 29%
promptlens/utils/timing.py 24 13 46%

TOTAL 1386 1043 25%
Coverage HTML written to dir htmlcov
======================== 7 passed, 5 warnings in 1.51s =========================

• Result: all tests passed (7 passed).

Risks

• Some custom endpoints may intentionally return empty content for valid calls. Those now return explicit errors instead of silent empties.

Rollback Plan

• Revert this PR commit to restore previous permissive behavior for empty content responses.