Web-search connector (Google CSE) + extract_contacts step — enrichment (#99) - #111
Merged
Conversation
… enrichment (#99) Enriching a lead with its socials/email needs web search, which the generic http connector (#95) can't cleanly cover: the search API key must live in its OWN vault slot, separate from the http connector's key. So this is a first-class connector. - web-search connector (auth:"token", grantModel:"user", read-only, no tokenEnv → connectorClient reads user_api_keys provider "web-search"). Provides one web_search(query,[num],[cx]) → results[{title,link,snippet}] tool wrapping the Google Custom Search JSON API. The KEY comes from the vault and is injected onto the request URL only — never in inputs/schema/output. The CSE id (cx) is NOT secret: a tool input, else the WEB_SEARCH_CX env default. Every call goes through safeFetch (#95 SSRF guard), not a re-implemented fetch. Per-call result cap (max 10, default 5). - extract_contacts step (steps.ts): pure, no I/O — scans web_search results for the first Instagram profile URL, Facebook page URL, and email (mailto: preferred), skipping non-profile/share/login paths → {instagram, facebook, email, precision:"best-effort"}. Accepts raw rows or the {results:[…]} envelope. - Dispatched through runRegistryTool so audit/grant/scope come for free. Proof (acceptance): given a business name+suburb, web_search + extract_contacts populate instagram/facebook/email as pure config (test with a mocked CSE response). Unit tests: web_search shape, key-from-vault-not-leaked (incl. no key input in schema + no key echoed on upstream error), endpoint wiring, num cap, cx override, SSRF path; extract_contacts link/email extraction + non-profile/share filtering + field subset. api tsc clean; api suite 864 passing (842 baseline + 22 new, 0 regressions). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This was referenced Aug 1, 2026
Closed
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
2 participants
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.
Closes #99. Part of #94 / #84. Builds on #86 (connectorClient) + #95 (safeFetch/vault) + #96 (step catalog).
Adds a
web-searchconnector (auth:token,grantModel:user, read-scoped) exposingweb_search(query,[num],[cx]) → results[{title,link,snippet}]over the Google Custom Search JSON API. The API key is read from its own vault slot (user_api_keysproviderweb-search, separate fromhttp) via connectorClient — never in inputs/schema/output (asserted, incl. no echo on upstream error). Thecx(CSE id, not secret) is a tool input orWEB_SEARCH_CXenv default. Wire viasafeFetch(SSRF-guarded); dispatched throughrunRegistryToolfor audit/grant;numcapped at 10.Adds a pure
extract_contacts(items,[fields]) → {instagram,facebook,email,precision:'best-effort'}step (no I/O — reads text web_search already fetched) that pulls instagram/facebook/mailto from results.Acceptance proof: a test composes
web_search→extract_contacts(both via the registry) to populate instagram/facebook/email from a business name+suburb as pure config.Notes: Google CSE only (not SerpAPI — acceptance doesn't need both); extracts from snippets/links (keeps the step pure, no per-page fetch), precision surfaced in output.
Verify:
tscclean · 864 api tests pass (+22, 0 regressions) · biome clean.🤖 Generated with Claude Code