Real-time web search for engineering references and troubleshooting.
Web Search gives the AI access to live internet results so it can find up-to-date information, troubleshoot solver errors, and ground its answers in real sources. Multiple search providers work behind the scenes to deliver the most relevant results.
How it works
The AI selects the best search approach based on your query. For broad technical questions, it uses search-augmented generation that synthesizes results into grounded answers with source citations. For targeted factual lookups, it uses structured extraction to pull relevant snippets directly.
Automatic error recovery
One of the most important uses of Web Search is late-stage troubleshooting escalation during simulation recovery. Here is how it works:
Solver encounters an error
The simulation fails with an error message (e.g., a divergence, boundary condition mismatch, or mesh quality issue).
Local context gathering runs first
The AI evaluates the failed command, checks the exact logs, and gathers targeted context from skills, local references, internal knowledge, memory, and external docs before editing files.
Web search escalates only when needed
If local context and official docs are still insufficient, the AI can search the web for the exact error message, solver version, or edge-case behavior. The web search is flag-gated and used as a late fallback, not the default first step.
Search-informed diagnosis feeds a targeted retry
The AI combines web findings with internal evidence, applies a narrow fix to the relevant file, and compares the retry outcome against prior attempts. If the failure remains unclear, it can delegate focused troubleshooting to agent("error-diagnostician").
Transparent process
When web search is triggered during error recovery, you see the search query and source citations in the chat so you know exactly what information the AI is using.
Standalone usage
You can also use web search directly in conversation by asking the AI to look something up:
"Search for the latest OpenFOAM v2406 release notes."
"Find benchmark data for turbulent flow over a flat plate."
"Look up the recommended y+ values for the k-omega SST model."
The AI decides whether to use search based on the question. You can explicitly request it by saying "search the web for..." if the AI does not trigger it automatically.
Source citations
Search results include source citations displayed with:
Favicon of the source website
Title of the page or post
URL linking to the original source
Citations appear inline in the AI's response so you can verify the information and read the full source if needed.
Contextual grounding
Search results are not simply copied into the response. The AI uses them as grounding context -- it reads the sources, extracts the relevant parts, and integrates them into an answer tailored to your specific simulation problem. This means you get targeted advice, not generic search snippets.
Combine with Deep Research
For quick lookups, Web Search is fast and cheap. For thorough investigations, use Deep Research which runs multiple parallel searches and produces a synthesized report.