Forum Discussion
Knowledge retrieval on the GitHub Copilot harness suddenly slow (~1:25 min/turn)
Since about a week we're seeing severe knowledge-retrieval latency on a Copilot Studio agent (GitHub Copilot harness, SharePoint sources) — ~1:25 min per answer. No config changes on our side. Same source on the standard harness: ~20 s.
We captured a HAR of a slow turn and found this in the SSE stream:
- knowledge_search_sharepoint completes in ~11 s (durationMs: 10774) — search is not the problem.
- But instead of snippets, the tool result says: "The full documents have been downloaded to /app/uploads/" — 9 complete files, including large PDFs. 8 of 10 results contain no snippet at all.
- It then instructs the model: "Read each relevant file IN FULL before responding — every part, from the start to the end of the file."
- The remaining ~70 s is the model reading those files before answering. More sources = more full documents = longer turns.
Oddly, the tool's own description promises the opposite: "Results are search snippets … call KnowledgeRetrieve with the ReferenceId to get the full document."
Anyone else seeing this since ~mid-August? Known issue or intended GA behavior?
2 Replies
Your trace separates the stages: SharePoint search finishes in about 11 seconds, while the remaining delay follows delivery of complete files to the GitHub Copilot harness. That makes indexing or permissions an unlikely primary cause. Microsoft documents that SharePoint knowledge can add latency, but not that every result must be read in full, so treat this as a harness behavior change or service regression until Microsoft confirms it. Compare both harnesses using the same prompt, user, sources, and file count. Then reduce the source set and temporarily remove the largest PDFs to show how latency scales. Preserve the HAR/SSE event, UTC timestamps, agent ID, environment ID, region, publication version, and exact response instruction. Check service health and republish once, but do not repeatedly rebuild the source. Open a Copilot Studio support case with the 11-second retrieval and 70-second file-reading segments; only Microsoft can confirm whether that internal instruction is intended.
In my case the search takes 11 seconds the other 80 seconds the AI is reading entire PDF manuals cover to-cover because the retrieval no longer returns snippets
