Implemented web search.

[ChenZiHong-Gavin]

The current integrated search capabilities include:
web: Google and Bing;
kg: Wikipedia;
db: UniProt—repository of protein sequence and functional information, offering the manually reviewed UniProtKB/Swiss-Prot and the computationally annotated UniProtKB/TrEMBL datasets, covering protein function, sub-cellular localization, interactions, post-translational modifications, and disease associations.

When raw data are missing internally, we can use web search, leveraging the above engines and databases to gather and verify information broadly.

[Copilot]

Pull Request Overview

This PR enhances GraphGen's search functionality by implementing a comprehensive web search system. The update replaces the previous Wikipedia-only search with a multi-source search capability including Google, Bing, and database sources like UniProt.

Key changes:

• Replaces single Wikipedia search with multi-source search architecture supporting web (Google/Bing) and knowledge graph (Wikipedia) sources
• Refactors search configuration from boolean flag to structured config with enabled status and search types
• Adds new search client implementations for Google, Bing, and UniProt APIs

Reviewed Changes

Copilot reviewed 19 out of 28 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
webui/translation.json	Updates UI translation strings, reorders some entries
webui/app.py	Removes web_search boolean parameter, improves code formatting and structure
resources/examples/keywords_demo.txt	Adds new example file with biological keywords
graphgen/utils/file.py	Creates new utility function for reading different file formats
graphgen/utils/init.py	Updates imports to include new file utility
graphgen/operators/search_wikipedia.py	Removes old Wikipedia search implementation
graphgen/operators/search/web/search_google.py	Implements Google search functionality
graphgen/operators/search/web/search_bing.py	Implements Bing search functionality
graphgen/operators/search/search_all.py	Creates unified search orchestrator
graphgen/operators/search/kg/search_wikipedia.py	Refactored Wikipedia search implementation
graphgen/operators/init.py	Updates exports to use new search_all function
graphgen/models/search/web/google_search.py	Google Search API client implementation
graphgen/models/search/web/bing_search.py	Bing Search API client implementation
graphgen/models/search/kg/wiki_search.py	Updates Wikipedia search with num_results parameter
graphgen/models/search/db/uniprot_search.py	UniProt database search client implementation
graphgen/models/init.py	Updates imports for new search clients
graphgen/graphgen.py	Refactors search configuration and integrates new search system
graphgen/generate.py	Updates to use new search config and file reading utility
graphgen/configs/graphgen_config.yaml	Changes configuration format for search settings

[Copilot]

The filter_keys method is being called with all_nodes_names but this method typically filters for existing keys, not missing ones. For search functionality, you likely want entities that don't already exist in storage, which would require a different approach or method.

Suggested change

	new_search_entities = await self.full_docs_storage.filter_keys(
	all_nodes_names
	)
	existing_keys = await self.full_docs_storage.get_keys()
	new_search_entities = [
	node_name for node_name in all_nodes_names if node_name not in existing_keys
	]

[Copilot]

If trafilatura.fetch_url fails or returns None, the subsequent trafilatura.extract call will fail with an unclear error. Consider adding error handling and validation.

Suggested change

	content = trafilatura.fetch_url(first_result["link"])
	content = trafilatura.fetch_url(first_result["link"])
	if content is None:
	logger.warning("Failed to fetch content for URL: %s", first_result["link"])
	return None

[Copilot]

If trafilatura.fetch_url fails or returns None, the subsequent trafilatura.extract call will fail with an unclear error. Consider adding error handling and validation.

Suggested change

	content = trafilatura.fetch_url(first_result["url"])
	content = trafilatura.fetch_url(first_result["url"])
	if content is None:
	logger.warning("Failed to fetch content from URL: %s", first_result["url"])
	return None