From eb5fe204d3271214120669ad26887a465d5c0d5f Mon Sep 17 00:00:00 2001 From: Gohar Anwar Date: Wed, 20 May 2026 20:41:37 +0500 Subject: [PATCH] bump default request timeout 30s -> 60s Agent execution involves LLM generation that routinely takes >30s in staging; the prior default caused test_execute_agent_query to fail with 408 client timeouts while the server completed the request successfully in ~57s. The new default matches the existing budget asserted by test_execute_agent_response_time (< 60_000ms). Other endpoints' SLA assertions all remain well under 60s (auth < 5s, get/list corpus < 5s, indexing < 10s, rag summary < 30s, query < 5s), so the bump only buys headroom for agent calls. Co-Authored-By: Claude Opus 4.7 (1M context) --- README.md | 2 +- utils/config.py | 9 +++++++-- 2 files changed, 8 insertions(+), 3 deletions(-) diff --git a/README.md b/README.md index b96609b..c877914 100644 --- a/README.md +++ b/README.md @@ -102,7 +102,7 @@ python run_tests.py --profile core -p 4 |----------|-------------|----------| | `VECTARA_API_KEY` | Personal API key | Yes | | `VECTARA_BASE_URL` | API URL (default: `https://api.vectara.io`) | No | -| `VECTARA_TIMEOUT` | Request timeout in seconds (default: 30) | No | +| `VECTARA_TIMEOUT` | Request timeout in seconds (default: 60) | No | | `VECTARA_LLM_NAME` | LLM model name for generation | No | | `VECTARA_GENERATION_PRESET` | Generation preset name | No | | `OPENAI_API_KEY` | OpenAI key for BYOL LLM tests (regression only) | No | diff --git a/utils/config.py b/utils/config.py index ab94ce9..4e1c30a 100644 --- a/utils/config.py +++ b/utils/config.py @@ -27,8 +27,13 @@ def base_url(self) -> str: @property def request_timeout(self) -> int: - """Get request timeout in seconds.""" - return int(os.environ.get("VECTARA_TIMEOUT", "30")) + """Get request timeout in seconds. + + Default 60s — agent execution involves LLM generation which routinely + runs > 30s in staging; matches the explicit budget asserted by + test_execute_agent_response_time. + """ + return int(os.environ.get("VECTARA_TIMEOUT", "60")) @property def max_retries(self) -> int: