[franciscomaestre]

Adds Serper.dev as a web search backend.

What

• New `serper` provider in `WebSearchTool`, following the existing provider pattern (keenable/exa).
• POSTs to `https://google.serper.dev/search\` with the `X-API-KEY` header and maps the `organic` results into the shared result format.
• Key resolved from `config.api_key` or the `SERPER_API_KEY` env var; falls back to DuckDuckGo when no key is set (so it stays concurrency-safe).
• `429` returns a rate-limit message; other HTTP errors are surfaced like the other providers.

Tests

Added in `tests/tools/test_web_search_tool.py`, mirroring the keenable tests:

• success (asserts endpoint, header, payload)
• env-var API key
• no-key fallback to DuckDuckGo
• HTTP error (403)
• rate limit (429)

Docs

• Added a Serper config example in `docs/configuration.md`.
• Listed `serper` in the `tools.web.search` provider table.

Config

```json
{
"tools": { "web": { "search": { "provider": "serper", "apiKey": "${SERPER_API_KEY}" } } }
}
```

[chengyongru]

Automated PR review by nanobot. This is not a human maintainer review or approval.

Adds a serper provider (Serper.dev Google Search API) to the web search tool, following the existing keenable/exa pattern.

What's done well:

• Follows the established provider pattern faithfully: _effective_provider with key check → DuckDuckGo fallback, _search_serper method, _format_results for uniform output.
• HTTP error handling is thorough — 429 rate-limit gets a user-friendly message, other HTTP errors include the status code, and unexpected exceptions are caught.
• Test coverage mirrors the keenable tests: success, env-var key, no-key fallback, HTTP error (403), and rate limit (429). 5 new tests.
• Docs include a config example and the provider table is updated.

Observations:

• No issues found. The PR is well-scoped and consistent with existing code patterns.

[chengyongru]

Automated PR review by nanobot. This is not a human maintainer review or approval.

Side effect analysis:

Adds Serper.dev (Google Search API) as a new web.search provider. Follows the existing provider pattern identically to keenable/bocha/olostep:

• _effective_provider(): checks config+env key, missing key → DuckDuckGo fallback (same pattern as keenable)
• _search_serper(): POST to https://google.serper.dev/search with X-API-KEY header + JSON payload {"q": query, "num": n}
• Response parsing: organic[].{title→title, link→url, snippet→content} matches standard Serper response format
• Error handling: 429 triggers specific rate-limit message, other HTTP errors return status code, generic Exception caught as catch-all
• Docs: provider list table updated, config example added
• Tests: 6 test cases covering key fallback, env key, env missing, http errors, rate limiting, happy path

No correctness, security, or performance concerns.

[franciscomaestre]

Rebased onto current main — conflicts resolved and all 45 web-search tests pass. Ready for another review whenever you have a moment. Thanks!

[franciscomaestre]

Friendly ping 🙂 — this one's been green for a while: CI passes on all matrix targets, both automated reviews found no correctness/security/performance concerns, and it's already rebased onto current main with all web-search tests passing. Whenever a maintainer has a moment for a final look/approval, I'd appreciate it. Thanks!

[Re-bin]

LGTM. Serper is added as a normal web search provider with configuration docs and structured error handling for failed/rate-limited searches. CI is green.