feat: web search fallback chain + loop-guard rework (Closes #467 #544 #571)#574
Merged
Merged
Conversation
This was referenced Jun 27, 2026
Contributor
🔎 Lint report:
|
| Rule | Count |
|---|---|
invalid-argument-type |
3 |
invalid-assignment |
2 |
unsupported-operator |
1 |
First entries
tools/web_tools.py:1019: [invalid-assignment] invalid-assignment: Invalid subscript assignment with key of type `Literal["provider"]` and value of type `bool | str | None | Any` on object of type `dict[str, str | int]`
tests/tools/test_web_providers.py:304: [invalid-assignment] invalid-assignment: Object of type `() -> dict[Unknown, Unknown]` is not assignable to attribute `_load_web_config` of type `def _load_web_config() -> dict[Unknown, Unknown]`
tests/tools/test_web_providers.py:252: [invalid-argument-type] invalid-argument-type: Method `__getitem__` of type `bound method str.__getitem__(key: SupportsIndex | slice[SupportsIndex | None, SupportsIndex | None, SupportsIndex | None], /) -> str` cannot be called with key of type `Literal["search_backend_fallback_chain"]` on object of type `str`
tests/tools/test_web_providers.py:252: [invalid-argument-type] invalid-argument-type: Method `__getitem__` of type `Overload[(i: SupportsIndex, /) -> Unknown, (s: slice[SupportsIndex | None, SupportsIndex | None, SupportsIndex | None], /) -> list[Unknown]]` cannot be called with key of type `Literal["search_backend_fallback_chain"]` on object of type `list[Unknown]`
tests/tools/test_web_providers.py:247: [unsupported-operator] unsupported-operator: Operator `in` is not supported between objects of type `Literal["search_backend_fallback_chain"]` and `str | dict[Unknown, Unknown] | list[Unknown] | ... omitted 37 union elements`
tests/tools/test_web_providers.py:252: [invalid-argument-type] invalid-argument-type: Method `__getitem__` of type `Overload[(i: SupportsIndex, /) -> str, (s: slice[SupportsIndex | None, SupportsIndex | None, SupportsIndex | None], /) -> list[str]]` cannot be called with key of type `Literal["search_backend_fallback_chain"]` on object of type `list[str]`
✅ Fixed issues (1):
| Rule | Count |
|---|---|
invalid-method-override |
1 |
First entries
tests/tools/test_web_providers.py:385: [invalid-method-override] invalid-method-override: Invalid override of method `extract`: Definition is incompatible with `WebSearchProvider.extract`
Unchanged: 6220 pre-existing issues carried over.
Diagnostics are surfaced as warnings — this check never fails the build.
This was referenced Jun 27, 2026
- _search_with_fallbacks now tries the configured fallback chain when the primary backend reports failure OR when it returns empty results and a fallback chain is configured. Empty results with no fallback chain remain success=True so real 'no hits' outcomes are not turned into provider-dead errors (#467 rework). - DDGS provider surfaces provider-dead only when empty results occur and a fallback chain exists. - Update fallback-chain tests to match the new semantics and keep regression coverage. - Carry forward prior loop-guard / tool-diagnostics improvements from #544. Closes #467 Closes #544 Closes #571 Co-Authored-By: Hermes Evolution <evolution@hermes.ai>
Lexus2016
force-pushed
the
evolution/issue-467-544-571-rework-v3
branch
from
4dd96f3 to
97d671f
Compare
The #467 same-query short-circuit for spiral-prone idempotent tools (web_search / web_extract / search_files) was functionally inert: the dedup identity was computed by `_tool_result_arg_hash`, which parsed the tool RESULT with `json.loads` to recover input args. But tool results do not carry the input args, and web_search / web_extract outputs are XML-wrapped in `<untrusted_tool_result>`, so the parse always raised and the arg hash was always None -> the short-circuit never triggered. With zero test coverage, CI stayed green over a dead feature. Fix: derive the identity hash from the assistant tool-call INPUT args (`tool_calls[].function.arguments`) instead of the result. New `_tool_call_arg_hash` canonicalizes JSON-string OR dict args (key-order insensitive), hashes unparseable string args verbatim, and returns None on missing args (fail-safe: no spurious short-circuit). The generic repeat/spiral detection is untouched. Adds TDD coverage in tests/agent/test_loop_guard.py: same-query fires (string args, dict args, key-order insensitive, web_extract), and negatives (different queries do not short-circuit; varied queries still hit the generic spiral nudge at the repeat threshold).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Automated evolution PR implementing the web-search fallback chain and loop-guard/tool-diagnostics improvements selected for this cycle.
Scope:
_search_with_fallbacksusessearch_backend_fallback_chain; falls back on failure or on empty results when a chain is configured.search_backend_fallback_chaindefault.Closes #467
Closes #544
Closes #571
Co-Authored-By: Hermes Evolution evolution@hermes.ai