Skip to content
Draft
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,15 @@

1. **Select URLs** — read [reference-document-links.md](reference-document-links.md). Select only the URLs relevant to the user's request and Step 1 result. If unsure, select all.
2. **Fetch** — `web_fetch` each selected URL. Extract content as markdown.
- **Timeout**: Allow at most 30 seconds per URL fetch. If a fetch times out or fails, skip it and continue with the remaining URLs.
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Just curious on the cases that timeout happen? Because from my previous perf test, web_fetch avg time is 0.55s. And I did not encounter timeout case before.

Image

#14740 (comment)

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm fine with holding off on this pr until we have more data/feedback. This came up when I was working on the evals with copilot and it was analyzing how long it took to process a request, however I wouldnt say there's enough data to strongly push to get this into the skill right now.

- **Total budget**: Complete all fetching within 2 minutes. If the budget is exceeded, stop fetching and proceed with whatever content has been collected.
3. **Search** — find content matching a query derived from the user's request and Step 1 result. Choose the most effective local search tool available.
4. **Iterate** — if initial results are insufficient, refine the query or fetch additional pages/URLs until the information satisfies the query.
4. **Iterate** — if initial results are insufficient, refine the query or fetch additional pages/URLs until the information satisfies the query. Respect the total time budget.
5. **Return** — provide the extracted guidance to the caller.

## Fallback
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

could we put the fallback part to thttps://github.com/Azure/azure-sdk-tools/blob/main/.github/skills/azure-typespec-author/references/authoring-plan.md?
And let the agentic search instructions focusing on agentic search itself.

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I saw there is a fall back added here: https://github.com/Azure/azure-sdk-tools/pull/15840/changes#diff-ca1df672aae30ef5d37ee71419674b77f4813b1169025e86a4c2b50baf1e3f16R16
could we remove the fallback part in agentic search?


If all URL fetches fail or the total budget expires with no useful content:
- Proceed without agentic search results.
- Rely on the MCP tool (`azsdk_typespec_generate_authoring_plan`) and built-in knowledge.
- Do NOT block the workflow waiting for unreachable URLs.