fix: move plugin load before backend resolve in web_extract

[kyan001]

Extract should load plugins before checking registry, same as search. Now custom extract plugins can't be found on cold start.

What does this PR do?

Fix init order in web_extract_tool: load plugins before checking registry, matching the pattern already used by web_search_tool. Without this, custom extract plugins can't be found on cold start.

Related Issue

Fixes #73192

Changes Made

• tools/web_tools.py: moved _ensure_web_plugins_loaded() before _get_extract_backend() in web_extract_tool

How to Test

1. Configure a custom-plugin-based extract backend (e.g. web.extract_backend: my-extractor)
2. Start a fresh Hermes session and call web_extract(urls=[...]) as the first web operation
3. Content is extracted successfully instead of returning "search-only backend" error

[teknium1]

Thanks for the focused fix. The current main still calls _get_extract_backend() before _ensure_web_plugins_loaded() in tools/web_tools.py:857-866; this differs from the working search ordering at tools/web_tools.py:678-685 and can reject a registry-backed custom backend during a cold start.

Problems

• No regression test accompanies the reorder. The closest test, tests/tools/test_web_providers.py:249-332, checks only that discovery was eventually called (mock_hook.called at line 327); it does not establish that discovery precedes backend resolution or that a custom configured provider is used.

Suggested changes

• Add or strengthen a cold-start custom-provider test in tests/tools/test_web_providers.py: have discovery register the configured custom extract provider and assert that provider performs the extraction. That test should fail with the current-main order and pass with this patch.

Automated hermes-sweeper review.

[GottZ]

This was generated by AI during triage.

Summary

One PR addresses #73192. #73183 moves plugin loading ahead of extract-backend resolution, directly correcting the reported cold-start registry lookup failure.

Related pull requests

• fix: move plugin load before backend resolve in web_extract #73183 best fix — (+1/-2) — keep open with a salvage path: The diff moves _get_extract_backend() after _ensure_web_plugins_loaded(), so a configured plugin-backed extractor can be registered before resolution on the first call. Consistent with the maintainer-bot keep_open review on fix: move plugin load before backend resolve in web_extract #73183, the salvage path is to add a cold-start regression test in tests/tools/test_web_providers.py that registers the configured custom provider during discovery and verifies that it performs extraction.

Suggested consolidation

Keep #73183 open with the concrete salvage path identified by the maintainer-bot review: add the cold-start custom-provider regression test that fails under the current ordering and passes with this reorder. No competing or duplicate PRs are present.

Complex graph

flowchart LR
    classDef open fill:#dbeafe,stroke:#1d4ed8,color:#1e3a8a
    classDef merged fill:#dcfce7,stroke:#15803d,color:#14532d
    classDef closed fill:#e5e7eb,stroke:#6b7280,color:#1f2937
    classDef unverified fill:#f3f4f6,stroke:#9ca3af,color:#374151
    classDef best stroke-width:3px,stroke:#b45309
    classDef target stroke-width:3px,stroke:#4338ca
    I73192(["issue #73192 (open)"])
    P73183["PR #73183 (open)"]
    P73183 -->|best fix| I73192
    class I73192 open
    class P73183 open
    class P73183 best
    class P73183 target
    click I73192 "https://github.com/NousResearch/hermes-agent/issues/73192"
    click P73183 "https://github.com/NousResearch/hermes-agent/pull/73183"

Loading

Graph: solid arrow = fixes / best fix, dashed arrow = partial or unverified (see edge label); boxed group = PRs duplicating each other; amber border = best fix; indigo border = target; gray node = closed (state tag in the node label).

Cross-PR triage: Reviewed 1 pull request and 1 issue in this complex. Each diff was read against this issue; Assessment working set: <1 kB of PR diffs, 4 kB of issue/PR text, 2 verify verdicts. verdicts reflect diff content, not PR titles. Part of an automated triage batch.